On Wednesday, August 5, Axon reported its Q2 earnings. I was on holiday in Sardinia then, but of course, I want to know what’s going on with the companies you and I hold in our portfolios. That’s also because, as you probably know, I add money to my portfolio every two weeks and that’s why I want to know if the stock is attractive to add to my position. Here we go!

The Numbers

• Revenue: $904 million, up 35% YoY, beating the consensus by $28 million, a beat of 3.2%, which is pretty significant for such a big and widely followed company. It’s also the tenth consecutive quarter with revenue growth of more than 30%.

• Software & Services revenue: $398 million, up 36% YoY.

• Connected Devices revenue: $507 million, up 35% YoY. Within that category, Platform Solutions (in-car video, fixed cameras, drones, counter-drone, VR hardware) grew to $150 million, up 123% YoY.

• Dedrone: passed $100 million in revenue in a single quarter for the first time.

• Annual recurring revenue: $1.639 billion, up 39%.

• Net revenue retention: 126%, versus 124% a year ago.

• Future contracted bookings: $15.1 billion, up 41%.

• Adjusted gross margin: 62.9%, down from 63.3% a year ago, up from 61.6% last quarter.

• Adjusted EBITDA: $242 million, a 26.8% margin, up more than 40%.

• Operating income: $47 million, versus a loss of $1 million a year ago.

• Non-GAAP EPS: $1.88, versus $2.18.

• Free cash flow: negative $1 million, versus negative $115 million.

Axon also raised full-year revenue guidance from 30-32% to 32-34%, the second raise this year, and held the adjusted EBITDA margin guide at 25.5%.

This is an overview in the company’s earnings slide deck.

Earnings per share dropped while revenue grew 35%. That looks awful, right? It isn’t.

In Q2 2025, Axon’s pre-tax income was negative $38.9M. But it reported positive net income of $36.1M, because it booked a $75M tax benefit that quarter. So, you can’t really compare the quarters on this basis. This quarter, pre-tax income was $32.7 million, so much better than last year’s Q2 numbers.

You can also see the improvement in operating income. It came in at $47M this quarter, while there was still a $1M loss in Q2 2025.

I also like to see operating leverage. If you grow revenue by 35%, as Axon did, but the costs grow faster, you are not doing better with scale. But that’s not the case. SG&A (Selling, General and Administrative) grew 20% and R&D (Research and Development) 28%. So, there is operating leverage. President Josh Isner on the call:

A little over a year ago, we began flattening the organization to empower people closest to the work, reducing layers of friction and moving faster. We paired that organizational change with disciplined investment in AI tools for our teams. We didn’t rush to deploy every tool or ask people to use AI for the sake of using AI. We rolled it out deliberately, and we focused on where it can make our people more effective.

I like that, of course. Empowering the people closest to the work is a recipe for success.

In the last quarter, GAAP net income came in at $169M. But that was because of $189M of rising value of Axon’s investments. That’s a distortion, because the company doesn’t sell, but it has to adapt to the market value of its investments every quarter. Shopify is another company that has this.

Axon gives revenue per product line, but I prefer the Fiscal chart.

If you want a Fiscal subscription as well (highly recommended, I use it multiple times every single day), you can get 15% off through this link.

As you can see, the two legacy products, TASERS and body cameras (personal sensors) grew the least, at 21% (still very strong) and 3% respectively.

Software and services, mostly Evidence. com, was up 36% to almost $400M, and Platform Solutions was up a whopping 123% to $150M.

Last quarter I wrote “34% Growth Before Optionality.” What I meant was that Axon grew 34% while the new products contributed almost nothing. This quarter the optionality showed up just in time to offset the slowdown in body cameras.

But the body cameras are only in a temporary dip. Q1 is the worst quarter for new bookings for body cameras. As these are shipped a quarter later, that’s the effect you see. President John Isner said to expect camera shipments to grow by 20% to 30% QoQ in Q3 and probably closer to 30%.

The story is not that much different for TASER. Even if growth is still 21%, very respectable, it would have been higher without the seasonally weak Q1. In Q3, Axon booked the highest TASER unit volume ever. It also signed its biggest TASER order ever. President John Isner on the call:

That may sound crazy for a product category that has been around for three decades. Well, it’s real, and it speaks to how much opportunity remains.

I also want to talk about free cash flow because the company seems to be at a turning point there as well. It was still negative, $1M, but that was a negative $115M last year. That’s already a big improvement, but COO and CFO . It was negative $1 million against negative $115 million a year ago. A big improvement, and COO and CFO Brittany Bagley reaffirmed the full-year target of $450 million in free cash flow.

We are two quarters into the year and the mid-term score is negative $56M. In other words, for that $450 million target, Axon has to generate $506 million in free cash flow in the second half of the year. That’s a big acceleration. Q4 is seasonally Axon’s best quarter each year, so that won’t be different this year. Management also said that the inventory investments it made with the tariffs will moderate after this year.

In the first half of the year, Axon spent $145M to build out its inventory. That depressed FCF, but it was not even the biggest contributor. The biggest drain on FCF was receivables and contract assets, with $257M. Receivables are when the invoice is already out, but not paid yet. Contract assets are customers paying over the years they take the subscription for both hardware and software, for example. In both cases, Axon has already spent to deliver the goods, but it hasn’t received payment yet. FCF is, therefore, systemically suppressed.

So the numbers are excellent. But the most interesting stories of this quarter are not in the numbers.

Below: the protest movement that’s smashing cameras in 460 American cities (and why I think it’s actually good news for Axon), my Selling Rules, my proprietary Quality Score and Valuation. Don’t want to miss these? Subscribe now!

Subscribe now

Read more

TransMedics (TMDX) reported Q2 2026 on August 4. I was on holiday so I couldn’t do the analysis back then. That’s why I’m doing this article now. But sometimes, often, actually, some delay is good. It cuts out the noise and highlights the really important things.

The stock initially fell, but recovered and now it’s up about 10% since the earnings.

In my last update, in May, I gave a Valuation Score of 10/10 and I called it a STRONG BUY at $64. Since then, the stock is up more than 40%.

So, is it still a buy today? Let’s find out.

The Numbers

• Revenue: $189.9 million, +20.7% YoY, +9.2% QoQ.

• Product revenue: $111.2 million, +15.7% YoY, +3.0% QoQ.

• Service revenue: $78.8 million, +28.6% YoY, +19.4% QoQ.

• Gross margin: 59.6%, up 140bps QoQ.

• Adjusted operating margin: 13.6%, up from 10.4% in Q1, down from 23.2% a year ago.

• Adjusted EPS: $0.44 versus consensus of about $0.48.

• Cash: $472.7 million.

• Guidance: low end raised, now $737-757 million, 22-25% growth.

Here you have an overview from the Q2 2026 earnings call slide deck.

Source: Fiscal AI. You can find all transcripts there, earnings slides, the consensus estimates, financial data, charts, and so much more. Get a 15% discount through this link.

The Main Reason?

In the Q1 article, adjusted margins were 18.5%, or $112.2 million of adjusted operating income on $605 million of revenue. In May, the company guided for gross margins of about 16%, because of the investments it was making. But those margins are now seen as lower. CFO Gerardo Hernandez on the Q2 earnings call:

We now expect full-year adjusted operating margin, excluding the impact of PAD Aviation, of approximately 12.5%-14%. The range primarily reflects potential variability in revenue performance, while the reduction from our prior expectation primarily reflects the higher planned investment in OCS Kidney.

So margins went from 18.5% last year to a guidance of 16% for the whole year in May, and now to a guidance of about 13% for the whole year. The market doesn’t like that.

At Canaccord Genuity’s 46th Annual Global Growth Conference, CFO Gerardo Hernandez gave more context:

The way we should see or expect for the future years is we should not see an increase in operating expenses similar to the one that we had in 2026. This year was really the step up. In 2027, we will see some increase, probably more around the low teens, which certainly will help us to continue to expand our operating margin.

Hernandez added that there will be margin improvement next year, combined with a growth acceleration. On top of that, both Hernandez and founder and CEO Waleed Hassanein repeated their long-term goal of the old promise of about 30% operating margin at 10,000 transplants. To remind you, that goal has been set by management for the end of 2028 and it’s one of the Selling Rules I set.

In other words: 2026 is the investment year, but then it should get better next year and thereafter.

It may look like a detail, but TransMedics’ management raised the low end of the guidance. That detail is more important than you might think, because it points to a reacceleration in revenue growth. Let’s do the math together. In the first half of this year, revenue came in at $363.9 million. The guidance range of $737-757 million means that in the second half, the company has to have revenue of $373 to 393 million. If you know that in the second half of 2025, revenue was $304.6 million, that implies revenue growth of 22.5% to 29%. In the first half of this year, it was ‘just’ 20.9%. It’s good to see this, but management shouldn’t be shy about sharing this.

In my last update, I called TransMedics a STRONG BUY at $64. The stock is up more than 40% since.

Below: the numbers management shared about the company’s biggest potential market, why I’m removing a Selling Rule for the first time ever, why margins jumped when I expected them to fall, and whether the stock is still a buy after the run. Don’t want to miss the valuable insights? Subscribe now!

Subscribe now

Read more

As one of the last Potential Multibaggers, Rubrik (RBRK) reported its Q2 results on August 27. The company beat every single guided metric, raised full-year guidance for revenue, ARR, earnings, and free cash flow, and posted its first operating profit as a public company. But the stock is down 18.5% since the earnings.

To be honest, I’m a bit confused at this moment. It could be a buy-the-rumor, sell-the-event thing, based on my superficial look at the earnings. But that’s why I love this work so much. Together with you, I can find out what is really going on. Of course, we know that in the short term, the market is very fickle. Over the long term, the results matter and that’s what I want to assess. Is there real weakness here or not?

Let’s find out together.

Rubrik’s logo, source: the company

The Numbers

• Subscription ARR (annual recurring revenue): $1.661 billion, up 33% YoY. Net new subscription ARR of about $96 million, up 35% YoY.

• Subscription revenue: $407.2 million, up 37%.

• Total revenue: $427.3 million, up 38%, beating consensus by $31.3 million, or 7.5%. Very strong.

• Non-GAAP EPS: $0.20, much better than the consensus of $0.04.

• Subscription NRR: over 119%.

• Customers with $100K+ in subscription ARR: 3,084, up 23%. These customers now make up 88% of subscription ARR. Customers spending over $1 million grew more than 57%.

• Non-GAAP gross margin: 81%, versus 81.6% a year ago.

• Non-GAAP operating income: $33.5 million, or 8% of revenue, against a $4.4 million loss a year ago.

• Free cash flow: $65.7 million, up from $57.5 million, a 15% margin.

• Cash: $1.75 billion, against $1.13 billion of convertible debt.

Revenue growth of 38% is great and much better than the consensus. But as I already explained last quarter, it even understates things.

Rubrik used to be a hardware company, although it always intended to switch, but it was just easier to start.

Customers who gave up their right to a free hardware refresh received cloud credit as compensation, and when they use that credit, it appears as a one-time revenue bump. Rubrik calls this “material rights.” This compensation negatively affects revenue, of course, because those customers use credits they got for free. If you exclude the material rights, revenue grew 43%. That’s the third quarter in a row at 43%.

Source: Rubrik’s Q2 earnings slide deck

You can see the reported growth sliding from 51% at the peak last year, but the normalized line, excluding material rights, has been at 43% for three quarters. Very strong.

Net new subscription ARR accelerated to 35% growth. At $1.66 billion of ARR. Founder and CEO Bipul Sinha called it the tenth consecutive quarter of beating guidance since the IPO, and he was very proud of the acceleration:

And this quarter, we accelerated, yes, let me repeat, accelerated net new subscription ARR growth.

He’s right, although the acceleration is minimal, as you can see on this chart.

Rubrik had an operating profit for the first time as a public company. Non-GAAP operating income was $33.5 million, or 8% of revenue.

There were several contributors to this: non-GAAP sales and marketing dropped from 49% of revenue to 44%, R&D from 22% to 21% and general and administrative even dropped in absolute dollars, from $37.4 million to $34.6 million. If revenue then grows by 38%, you know that there’s operating leverage. Management also put this in the slides deck.

Free cash flow was $65.7 million, up 14% YoY. The margin came down from 19% to 15%. But the quarters differ quite a bit and so does the timing of big contracts, so let’s look at the first six months to smooth that out already a bit more. The H1 margin was 17%, up from 15% a year ago. The full-year guidance of $323 to $333 million of FCF means 36% to 40% growth in FCF.

I think you know by now I also like the Rule of 40. If you add FCF margin and revenue growth and it’s above 40, you have a good company. For Rubrik, you add up 15% FCF margin and 38% revenue growth, you get 53%. Without the material rights, that would even be at least 58%. Very strong, in other words.

All in all, very strong results overall. Which leads us to the next question.

Why Did The Stock Drop?

In the earnings slides deck, there’s a chart of cloud net new ARR.

The left panel shows the calculated cloud net new ARR sliding from $101 million in Q1 to $89 million this quarter, with year-over-year growth at minus 4%. Ouch, right? No, not really.

Rubrik has been migrating its legacy self-hosted customers to the cloud platform. Those customers were already paying Rubrik, but the moment they were switched to the cloud, their money got counted as new cloud ARR. That means it boosted the cloud new ARR for a long time.

But that migration is done now and therefore, the numbers look bad. In the right panel, the same two quarters with the migrations removed from both years show the real numbers: 20% growth, just like last quarter.

Cloud ARR is 89% of subscription ARR, exactly where it was last quarter. CFO Kiran Choudary said at the Analyst Day in June that it would stay around 90%.

The non-cloud business, which had been shrinking for years, is growing again. That’s good for margins. Founder and CEO Bipul Sinha explained where the demand comes from:

Given the geopolitics and given all the things that we are seeing around the world in terms of security, every country is now concerned about containing their supply chain and economic infrastructure. And as you can imagine, digital infrastructure and AI infrastructure is an important economic infrastructure. And that’s why folks are working on digital embassies and containing data, containing infrastructure in their own sovereign realm.

Let me recapitulate: the migrations ended, which looks optically bad, but it’s not. The growth that came on top of that, has a better margin and comes from governments, defense and regulated industries. Non-cloud is guided to contribute only low- to mid-single-digit net new subscription ARR in the second half, so this isn’t a meaningful growth engine but the higher margins are good.

So the first reason for the drop is superficial, not a structural weakness. But there’s a second reason. We also dive deeper into Rubrik’s unique position in cybersecurity and why AI is such a tailwind.

You also get my proprietary Potential Multibaggers Quality Score and the valuation and then you know if the stock is a buy or not.

Don’t want to miss it?

Subscribe now

Read more

Tempus AI (TEM) reported its Q2 2026 results on July 30.

The stock climbed steadily over the next two and a half weeks and was up about 24% by August 17. A week earlier, it was even up almost 35%.

Then, on Wednesday, it jumped again. This time, the reason was that Moderna and Merck published outstanding phase-3 data on their personalized cancer vaccine. We’ll see how Tempus is involved later in this article. This second jump means that Tempus AI’s stock is now up 47% since the earnings.

At the same time, it’s still down 35% from its high set in October 2025.

That’s why I prioritized this analysis. I want to know if Tempus AI’s stock is still attractive at this point or not. When it traded at $46.18 a few months ago, I gave it a 9/10 Valuation Score. The uncertainty was the only reason I didn’t give it the maximum score. How about now? Is the stock still attractive for long-term investors?

Let me go through the quarter first. At the end of the article, we’ll look at the Moderna breakthrough. Of course, we will also look at the Selling Rules, the PM Quality Score update and the stock’s valuation to decide if Tempus is still attractive now or if it is time to take profits after this move. So, first to the results.

The Numbers

Revenue: $382.5M, +21.6% YoY, beating the consensus by about $3M.

Diagnostics: $289.3M, +20%.

Data and Applications: $93.2M, +28%, with Insights, Tempus AI’s data licensing and AI modeling division, up 36%.

Hereditary: $107.4M, +5%.

Gross profit: $246.5M, +26%, a margin of 64.4%, up from 62% in Q2 2025.

Non-GAAP EPS: -$0.04, an $0.11 beat, up from -$0.22.

GAAP net income: $5.6M, versus a $42.8M loss a year ago. It was the first positive GAAP quarter ever.

Adjusted EBITDA: $8M, versus -$5.6M in Q2 last year.

Oncology volume: about 96,500 tests, +31%, accelerating from 28% in Q1.

MRD volume: about 9,000 tests, +38% QoQ.

Guidance: full-year guidance raised to $1.595-1.605B, which means about 25% growth YoY.

Here’s a summary from the earnings slides.

Tempus had its first quarter with a GAAP profit but I think you should ignore it, as it’s only a paper profit. Fair value changes accounted for $97.4M. Tempus already owned about 20% of Personalis and then announced that it will buy the company. The stock shot up because of that and the paper gains of the existing position have to be counted for GAAP. But this is not real profitability.

Loss from operations was -$75.9M, worse than -$61.8M a year ago, mostly because of stock-based compensation. If you strip that out and also the acquisition costs and amortization of intangibles, the non-GAAP loss from operations was -$2.7M versus -$17M last year. So, there is improvement, but with a big asterisk. Non-GAAP net loss was -$7.7M, versus -$37.3M last year.

Hereditary

Hereditary testing looks for cancer risks you were born with, not for a tumor you already have. For example, it can look for a BRCA mutation, which raises the chances of breast and ovarian cancer a lot and healthy women can be screened earlier if they know they have this mutation. Tempus got the biggest part of Hereditary with its Ambry Genetics acquisition in 2025.

But just like last quarter, this section is not doing great. It grew just 5% YoY to $107.4M. CEO Eric Lefkofsky said last quarter mid-teens growth would return in the back half of the year. The explanation is that Q2 2025 was abnormally strong, so the comps are very tough.

Tempus also launched two new tests in the hereditary branch. xH reads the entire genomes of patients with blood cancers such as leukemia and lymphoma. Rare Whole Genome does the same for people with rare diseases that are extremely hard to find. Often, those people already got many tests, and this genome test can finally give them the diagnosis they have been looking for.

Both have just launched and won’t fix the slow revenue growth in Hereditary soon. But they are promising nonetheless.

The ASP Boost

ASP means average selling price: what Tempus actually gets per test from insurers and Medicare. It’s now $1,850, up from around $1,740 in Q1.

There were two important developments this quarter. The first is that the FDA approved xT CDx in its tumor-only version. A tumor-only test reads the cancer alone, without also sequencing the patient’s healthy DNA to compare with. Tempus already had approval for the version that does the test that does both readings.

With this approval, now every solid-tumor test Tempus has can be sold as an FDA-approved test. The result is that Medicare used to pay $2,900 but now pays $4,500. That’s theory, because in reality, not every test gets the full price. Founder and CEO Eric Lefkofsky said you should expect an increase of about $200 per test, or roughly $85 million a year, starting in 2027.

The second development is the expected boost from xF. That is a liquid biopsy test that reads tumor DNA in the blood when you can’t use tissue. The FDA is now assessing the xF test, and CFO Jim Rogers said the company wants to set the price at $7,500 once approved. CEO Eric Lefkofsky explained the reasoning for this price:

And look, it’s an evolving market. Our assay is most comparable in terms of size, like literally size like megabases and size totality to Guardant’s recent assay that they got approved. And I believe their ADLT pricing is something like $8,300 or $8,400. And so we -- it would be very hard for us to go to the market with a almost identical, at least in terms of like size and complexity assay that’s radically less expensive.

Guardant Health’s similar test sets the reference price for other companies. That’s how it often goes in healthcare.

So, you can see that the ASP is currently $1,850. If you add $200 from xT CDx and $550 from xF, then you get $2,600. In other words, that’s 40% more revenue from tests Tempus already has.

I think this is why the stock jumped after the earnings release. Higher profits on the same tests cost almost nothing extra, and that means more gross profit. Right now, Tempus has about 386,000 oncology tests a year. $750 more per test means $290M of extra revenue even if there are no additional tests, which is very unlikely. The number of tests grew over 30% and that’s why management projects about $400 million extra revenue in 2028.

xF isn’t approved yet, though. Lefkofsky expects it in the market in the second half of 2027.

The Data Business

This is the most interesting division of the company.

A few months ago, I gave Tempus a 9/10 Valuation Score at $46.18. The stock is at $72.69 now, up 57%.

Below: how Tempus is connected to the Moderna cancer vaccine breakthrough everyone talked about this week, the reason pharma giants almost can’t leave Tempus anymore, the AstraZeneca test that was passed and if Tempus AI’s stock is still a buy after the jump.

Subscribe now

Read more

Today, just before the market closed, I added to the Forever Portfolio.

For newer Multis, that’s the portfolio with my real money.

This is a short article about what I bought and why. I already announced most of what I would buy in the Best Buys Now article, and I’m not going to write all the same things again here. As always, I put my money where my mouth is and I buy the Best Buys Now.

Let’s go over them one by one.

If you don’t have a subscription yet, take one now to see what I did today!

Subscribe now

Read more

Celsius Holdings (CELH) reported its Q2 2026 results on August 6, and the stock fell about 20% into the low $20s.

It recovered a big part of the losses, though.

But at the same time, the stock is still down more than 70% from its 2023 highs (when I didn’t own it yet).

I wrote about the Q1 results in June. The Celsius brand grew 6% then, and I explained why that number was better than it looked at first sight. Management was removing regional flavors to create a single national assortment within the PepsiCo distribution network. Every removal costs revenue, but the core 12-ounce cans were still growing double digits in convenience stores.

Before we look at the analysis of this quarter, it’s good to know what we’re talking about, so let’s check the numbers.

The Numbers

• Revenue: $817.9M, up 11% YoY, missing the consensus by about $52M. That’s a big miss of 6%.

• Adjusted EPS: $0.36, down 23% YoY, missing the consensus by $0.06. The biggest reason was a distributor termination charge, which is nothing to worry about.

• GAAP EPS: $0.14, down from $0.33 a year ago.

• North America revenue: $790.7M, up 11%.

• International revenue: $27.2M, up 10%, and down from $35.3M in Q1.

• Brand split: Celsius $387.0M (down 12% YoY), Alani Nu $364.4M (up 21%), Rockstar $66.5M.

• Gross margin: 48.1%, down from 51.5% a year ago and roughly flat versus the 48.3% in Q1.

• GAAP net income: $55.3M, down 45%. Net income attributable to common shareholders: $36.4M, down 57%. (The difference is that Pepsi owns a part of Celsius).

• Adjusted EBITDA: $184.2M, down 12%, a 22.5% margin against 28.4% last year and 25% last quarter.

The Celsius Brand Is Now Shrinking

The Celsius brand had $387M in net sales, down 12% year over year. Next to the earnings miss, it’s probably the reason the stock got hit so hard.

CFO Jarrod Langhans gave more insight:

Starting with brand CELSIUS, net sales were down approximately 12% year-over-year, while retail sales in tracked channels were down 2% in the second quarter.

I’m going to discuss those 2 numbers separately because they are telling you 2 different things. The scanner number, down 2%, is the consumer...

The gap to reported net sales came from 3 things: shipment timing related to inventory rebalancing, increased trade and promotional investment, and softness in the club channel.

Scanner data measures what shoppers really bought at the store, not what Celsius shipped. That distinction is important.

The revenue bridge in the deck breaks the 12% down: scanner growth was minus 2%, contra revenue took a 3% hit (contra revenue is the money that never reaches the top line, things like promotional spending and discounts), the gap between orders and scans was 7%, non-ready-to-drink energy products dropped 2%, and international added 2%.

The 7% orders-versus-scan difference is inventory rebalancing in the PepsiCo distribution system. As SKUs come out (a SKU is a single item, one specific flavor in one specific size), distributors buy more inventory, but then buy less a while later. That goes in cycles.

The 2% drop is the real Celsius brand slowdown. And, of course, that’s not great.

When asked whether the SKU rationalization was the right call and what he’d do differently, CEO John Fieldly said:

We went too deep on the CELSIUS rationalization. I think when you look at we’re entering the year, we could have done a much lighter job on that. But then there’s also puts and takes. The integration on Alani could have been further challenged, and Rockstar. Looking back, I definitely would have not cut as many SKUs within the organization through these commercial plans.

I like the honesty, and I think it’s worth more than superficial marketing answers you often get. Adjusting the course is the best thing you can do for any business and acknowledging you were wrong so fast after implementing the rationalization is a strength.

There’s a second self-inflicted wound that has consequences for revenue growth. CEO John Fieldly said they deliberately held back innovation on the Celsius brand while integrating Alani Nu and Rockstar, and that means the brand is now lapping six innovations from last year with nothing new to replace them. It’s normal you don’t do great then.

There’s a number that remains very positive, though: dollars per point of distribution rose about 16% from Q1 to Q2, despite about 7% fewer points of distribution.

CEO John Fieldly expects Q3 to look a lot like Q2 for the Celsius brand, with a return to growth in Q4. So, if you care about the short term, don’t expect Celsius to move anytime soon. The weird thing is that the analysts don’t see an acceleration in Q4 or Q1 2027 in their forecasts.

We also have to talk about Costco. In June I wrote that Costco’s Kirkland energy drink was scaring the market more than it was hurting the business, and I estimated the total damage at maybe 1% to 2% of sales. I still think that argument is mostly right. Over 60% of energy drink sales happen in convenience stores and gas stations where people buy one cold can, and nobody drives to Costco for a single drink.

But management has now named club softness as one of the three reasons the flagship brand fell 12%. That doesn't prove Kirkland caused it, and club channel weakness can come from a lot of places, including Celsius pulling its own SKUs. But it's still remarkable.

If the scanner number for the Celsius brand stays negative once the innovation returns in 2027, private label is probably doing real damage.

Below: the real Alani Nu growth number (it’s better than the reported number), the Selling Rules, Quality Score and Valuation Score.
And also what I’m doing with the stock now. Don’t want to miss this?

Subscribe now

Read more

Last quarter I wrote an article called “Is This The Explanation?” I put forward a hypothesis: that the high growth in Joint Business Plans and the weakness in headline revenue could be two different sides of the same coin, that The Trade Desk (TTD) was circumventing the ad agencies to the brands themselves, and that this explained both the numbers and Jeff Green’s uncharacteristic silence.

I have to take that thesis back today. Not because the quarter was bad, though it was awful, but because Green himself said the opposite during this of what I thought could be the reason.

The stock fell about 22.5% after the results were announced and kept sliding.

In July 2024, I gave The Trade Desk a Valuation Score of 3/10. Since then, the stock is down 86%, even though it initially continued to go up.

From the absolute top, it’s down even more, 90%. In the last two quarters, I was very critical of the company, and on top of that, the Quality Score deteriorated substantially. Unfortunately, this quarter is even worse.

The Numbers

Revenue came in at $715.1 million, up just 3% year over year, against a consensus of $751.4 million. That’s a miss by 5%, which is a lot for a revenue miss. And don’t forget that the consensus was already down because of the weak guidance in Q1.

But the consensus is not the number that matters the most here. The Trade Desk itself guided for “at least $750 million” three months ago, when Q2 was already well underway. They came in $35 million below their own minimum. That big revenue miss is a big part of the big price drop and I understand the market.

Profitability was not better. Non-GAAP EPS came in at $0.34, missing by $0.06. GAAP net income was $64.4 million, or $0.14 per diluted share, down 28.6% from $90.1 million. Income from operations fell 13%, from $116.8 million to $101.6 million.

Adjusted EBITDA was $241.3 million, a 33.7% margin, down from $270.8 million and 39% a year ago. So EBITDA declined 11% in absolute dollars, not just as a percentage. All these numbers look really bad.

The hit to profitability came partly from taxes. The company paid $48.7 million in taxes on $113.1 million of pretax income. That means a tax rate of 43%, against 32% a year ago.

Total operating expenses were $613.5 million, up 6% but excluding SBC, up 12%. This was mainly driven by platform operations, which were up 22% to $184.3 million. That is the data center and AI infrastructure build-out that has been going on for a few quarters now.

The Trade Desk repurchased $78 million of stock in Q2, with $269 million left on the authorization. That is about half the $164 million they bought in Q1. I don’t find this very confidence-inspiring. In the quarter the stock collapsed, with $1.5 billion in the bank and a lot of the buyback program left, they cut the buyback in half. Is this because of the new CFO or just because the company lost its faith in its own stock?

The company guided for revenue of “at least” $650 million. While the results were very weak, guidance was even worse. The consensus stood at $806.5 million. That’s a miss by a whopping 24%. For revenue, this is almost unprecedented. Adjusted EBITDA is seen at about $160 million.

In Q3, revenue came in at $739.4 million, so with guidance of $650 million that’s a YoY decline of about 12% and 9% below the quarter the company just reported. I can’t say anything else than that this guidance stinks. Very, very bad and disappointing.

Look at this list:

• Q1 2025: 25.4%.

• Q2 2025: 18.7%.

• Q3 2025: 17.7%.

• Q4 2025: 14.3%.

• Q1 2026: 11.7%.

• Q2 2026: 3.0%.

• Q3 2026 guidance: -12%.

This is just horrible.

And there is a detail that makes the 3% of this quarter even worse. The new CFO (another one), Nate Olmstead, said this on the call:

We also benefited from political spending related to the U.S. midterm elections during Q2.

So, this quarter, the 3% includes a midterm tailwind that did not exist a year ago. In other words, normally, it would have been even worse than 3%.

There is one more way to see how much has changed: the Rule of 40, revenue growth plus margin. A year ago this quarter, that was roughly 19 plus 39, or 58. This quarter, 3 plus 33.7, or about 37. On the Q3 guidance, it implies something in the low teens. Growth and margin used to be communicating vessels. Sometimes, there was more growth and that impacted the margins a bit and sometimes, there was a bit less growth, but then margins were generally better. Now, both go down together. Again, a very bad look.

Management leaned hard on macro. The word was used 14 times during the earnings call. There is definitely some truth in the macro comments. But in the same quarter, Magnite grew its revenue by 17% with CTV up 36%, Meta’s advertising revenue grew 27%, and Amazon’s ad services grew 26%.

I have argued before that Amazon and Meta are not the same as The Trade Desk. But it shows that digital advertising did not slow down to low single digits for everyone. Macro can explain why The Trade Desk is under pressure. It cannot, on its own, explain why The Trade Desk is under so much more pressure than the industry it operates in. Last quarter, Jeff Green explained that most of the company’s revenue comes from Fortune 500 brands that respond differently to macro shocks. That was a reasonable answer for a quarter with 12% growth. It’s not for 3% growth (including extra money from the midterms) and definitely not for the 12% drop guidance.

Customer retention stayed above 95%. It always does. And yet revenue grew 3% and is guided to fall 12%. Retention at The Trade Desk counts the number of companies, not dollars. It is not a dollar-based net retention rate.

There are three explanations for the lower spending:

One: clients cut their total advertising budgets, so spending on the platform fell along with everything else. That is cyclical and it means it will recover.

Two: clients kept their budgets but moved a smaller share of them through The Trade Desk.

Three: gross spend held up reasonably but take rate or revenue mix weakened, so less of that spend converted into revenue.

Management gave us no way to judge on what it is. Gross spend and take rate are only disclosed once a year.

Jeff Green said CPG (consumer packaged goods) and autos together are about 25% of the business and that roughly half of those clients are growing well. He also said the majority of the top 100 accounts are growing double digits, that clients outside the top 500 grew over 50% year to date, and that EMEA and APAC each grew almost 30%.

If you put that together, it looks very bad. If roughly 12% of revenue is the troubled part and the other 88% is genuinely growing at double digits, that troubled 12% has to be down something like 45% to 50% YoY to drag the total down to 3%. That is not cyclical softness but a total collapse. The alternative is that the healthy 88% is not really growing at double digits. One half of the story has to give.

Green said some version of “don’t overstate this” at least four times on the call:

This is more a cyclical issue with a handful of customers, and it’s not hard to look at their earnings and see that they’ve had some challenges. (...) please don’t walk away from this thinking that that’s affecting everyone. It’s actually just a couple of them.

When a CEO tells you four times not to worry, that’s pretty worrying in itself.

Below: I look deeper into the earnings, to understand all details, see a worrying trend that The Trade Desk is not responsible for, update the Quality Score and Valuation Score and ultimately tell you what I'm doing with my position: holding, adding or selling.

Want to find out?

Subscribe now

Read more

When I picked Axon (AXON) in early April, I wrote that I didn’t want the stock price to skyrocket soon, so I could build out my position calmly. Well, the market didn’t listen. The stock is up about 35% since the pick. But fortunately, I didn’t follow up my own advice, as you could see in my buys for the PM Forever Portfolio. I bought faster than I’m used to because I saw the opportunity.

I’m happy for every Multi who started buying with me. But a 35% paper gain in three months says nothing about whether the thesis is right. Stocks move on sentiment over the short term and on fundamentals over the long term. And the long term can feel really long. Sometimes, it’s 3 years, sometimes 5 or even 7.

And on May 6, Axon released its Q1 2026 earnings, the first quarter it reports as a Potential Multibagger.

In the weeks after the earnings, management also spoke at the William Blair, TD Cowen and J.P. Morgan conferences, and hosted a webinar with three police chiefs.

I went through all of the material. That’s why this is just the first article this week. It was a lot to go through. This article is the result of all the material. I’ll try to give you the best insights from all of the sources in this article.

Let’s start with the numbers.

Subscribe now

The Numbers

• Revenue: $807 million, up 34% YoY, beating the consensus by $27.7 million, or 3.6%. This was the ninth consecutive quarter of 30%+ growth.

• Software & Services revenue: $355 million, up 35%.

• Connected Devices revenue: $453 million, up 33%. Within Connected Devices, Platform Solutions (counter-drone, fixed cameras, drones, VR) was the strongest grower with 95% YoY growth.

• AI product revenue: up more than 700% YoY (from a small base, but still great growth). AI bookings up 140%.

• Dedrone: revenue up over 300% for hardware and software. Bookings up 500%.

• Annual recurring revenue: $1.5 billion, up 35%.

• Net revenue retention: 125%, up from 123% a year ago and the fifth straight year of Q1 improvement.

• Future contracted bookings: $14.3 billion, up 44%.

• International revenue: up more than 100%, now 20% of total revenue for the second quarter in a row.

• Non-GAAP EPS: $1.61 versus $1.47 last year, a beat by $0.01.

• Adjusted EBITDA: $201.6 million. 25% EBITDA margin.

• Adjusted gross margin: 61.6%, down from 63.6% a year ago.

• Free cash flow: negative $54.6 million, versus roughly break-even last year.

You can see some of the numbers on this earnings slide.

You probably noticed two ugly ducklings in that list: the negative free cash flow and the falling gross margin. Don’t worry, we will get back to this. But first, something in the income statement that most coverage of this quarter got wrong in the other direction: net income.

GAAP net income was $169.3 million, versus $88 million in Q1 last year. Almost a double. Great quarter for profitability, you’d think, right? But if you go to the reconciliation tables in the appendix, you see a line called “Other (income) loss, net” of $189 million. That’s mostly gains on Axon’s investments, so the stakes it holds in other companies. In other words, more than the entire GAAP net income this quarter came from investments going up in value.

The best profitability numbers are the $201.6 million in adjusted EBITDA at a 25% margin and the 75.8% adjusted gross margin in software. Those are earned by the business, not the company’s investments.

Non-GAAP EPS grew only 10% while revenue grew 34%. Interest expense jumped from $7.8 million to $28.6 million because of the $1.8 billion in senior notes raised for the Carbyne and Prepared acquisitions. That’s a conscious move and I don’t care too much about this. The acquisitions were worth it for the long term, in my opinion. I just wanted to point it out.

Free cash flow was negative $54.6 million. That looks bad, but it isn’t.

Q1 is always Axon’s weakest cash quarter because of bonuses, sales commissions, and one of the two semiannual interest payments are all reasons. But the biggest driver was an inventory buildout. CFO Brittany Bagley said that without the inventory investments, free cash flow would have been positive even with all those Q1 payments. president Josh Isner explained the logic behind the buildout on the call:

Given the expanding geopolitical risks, the competition for key components and the growing demand of our products worldwide, we are investing in inventory with durability in mind. We never want inventory to be the reason we cannot maximize our growth and impact.

When hardware companies start to pile up inventory, investors often get nervous but with Axon, it’s safe: its products don’t go obsolete. TASER life cycles are 15 years or so, body cameras definitely more than 5 years, and president Josh Isner mentioned at the J.P. Morgan conference that TASERs launched in 2010 still sell today. This proves that this inventory buildout is not a risk.

Axon said it didn’t suffer from the high memory costs because it had already bought before. This shows the power of having inventory. CFO Brittany Bagley said that the memory impact on gross margin is not even meaningful enough to mention it separately.

On top of that, the full-year free cash flow guidance stayed at approximately $450 million.

Guidance

Axon raised its full-year revenue guidance from 27-30% growth to 30-32%.

The adjusted EBITDA margin guidance stayed at 25.5%, versus 25% in Q1, so management expects operating leverage in the second half of the year.

I also liked that the capex guidance went down, from $185-215 million to $160-190 million. So Axon guides for faster growth and lower spending at the same time. That also makes the free cash flow target of around $450 million attainable, despite the inventory buildout.

Future contracted bookings grew 44% to $14.3 billion. Management expects to recognize 20% to 25% of that as revenue over the next twelve months. That’s $2.9 to $3.6 billion of revenue that’s already signed, against a full-year guidance of roughly $3.6 billion. In other words, most of this year is already under contract. As long as bookings grow more or less in line with revenue, the 30%+ growth can go on for years and I love to see that. It was another part of my investment thesis.

The AI Era Plan: From Curiosity To Default

In Part 1 of my deep dive, I called Draft One the crown jewel of Axon’s AI. And this shows. Revenue grew more than 700%, AI bookings grew 140%, and Josh Isner said that nearly all large domestic law enforcement agencies now include AI in their purchases.

Founder and CEO Rick Smith on the call:

Just yesterday, I hosted a collection of chiefs in San Francisco. Two years ago, when I queried the room who used AI, it was almost 0. Today, it’s 100% are using AI tools in their personal lives daily, but they know those tools can’t be used on government data.

So, every chief now uses AI privately but they are not allowed to use ChatGPT or other LLMs for evidence. But they can use Axon’s AI, as that has all the necessary compliance. That’s why these deals are getting bigger and bigger. There were mentions of contracts of $50 million, $150 milllon and even $200 million with big cities.

The economics of the AI bundle are great. The original Officer Safety Plan from 2017 cost $99 per officer per month, the OSP 10 Premium costs $370 and with the AI Era Plan on top, it’s $569. That means that the revenue per officer is up almost six times over the years. But still, the customers feel like the winners because new products like Axon Vision are added to the plan automatically. That’s how you get NRR of 125%.

It has to be mentioned, though, that adoption is not completely frictionless. In a customer webinar Axon hosted in June, Chief Darrell Lowe of Redmond, Washington said his prosecutor’s office simply won’t accept AI-generated reports, which makes Draft One unusable for his department. But that’s the kind of resistance that should be expected.

That same webinar also gave a strong data point to support Axon. Chief Doreen Jokerst of Overland Park, Kansas, said that her drone-as-first-responder program gets to scenes 26% faster than her officers.

And how does management think about its own position in this race? Josh Isner, in his typical style:

The Kentucky Derby was last week, the Belmont is in a few weeks, we look at ourselves like Secretariat at the Belmont. We want to be accelerating ahead of everybody in AI.

For the non-horse-racing readers: Secretariat won the 1973 Belmont by 31 lengths, still the record. Modesty is not Isner’s biggest talent. Execution is, fortunately.

I picked Axon in April. The stock is up 32% since then. I’m happy for every Multi who bought with me. If you didn’t, the big question is, of course, if it’s too late now.

Below: my answer, the growth engine almost nobody talks about yet, my proprietary Quality Score for Axon, and much more.

Don’t want to miss opportunities like this anymore?

Subscribe now!

Read more

Rubrik (RBRK) reported its fiscal Q1 2027 results (yes, they are a year ahead of the calendar) on June 4, but I hadn’t written about them yet. The stock dropped after the earnings, but that was not that surprising. It had already run up close to 90% from the April lows before the earnings.

Six days later, management held its first Analyst Day in Las Vegas, and after that, the market reacted positively. Of course, I include the additional insight I got from that day in this article.

Let’s start with the results.

Subscribe now

The Numbers

• Subscription ARR (annual recurring revenue): $1.57 billion, up 32% YoY. Net new subscription ARR of $103 million.

• Subscription revenue: $374 million, up 41%.

• Total revenue: $387.1 million, up 39% and beating the consensus by $20.75 million. That’s 5.7%, strong for a revenue beat.

• Non-GAAP EPS: $0.16, versus the consensus of -$0.03.

• Subscription NRR: approximately 120%.

• Customers with $100K+ in subscription ARR: 2,946, up 24%. Larger customers now make up 88% of subscription ARR. Customers spending $1 million or more grew over 50%.

• Cloud ARR: $1.39 billion, up 43%, now 89% of subscription ARR.

• Non-GAAP gross margin: 82.9%, up from 80.5% a year ago.

• Free cash flow: $74 million, more than double last year’s $33 million. FCF margin came in around 19%, up from 12%.

For companies growing so fast, it’s always good to look at the Rule of 40. If you add revenue growth and FCF margin together, the sum should be above 40 to be a strong company. Rubrik has a Rule of 58. That’s very strong.

Rubrik started as a backup and recovery company, first with hardware, then in the cloud. When you were attacked by ransomware, you can simply turn back the clock to the second before the hack. From there, it also expanded into identity security and, more recently, into securing the AI agents that enterprises use on their own systems.

With its new offerings, Rubrik is positioned where it should be: at the intersection of cybersecurity and agentic AI. A while ago, some believed that Anthropic’s Mythos model would end the company, but that’s definitely not the case.

But back to the past, because there’s something you should understand there. Rubrik used to sell physical boxes bundled with software. Customers got a free hardware upgrade every few years as part of the deal. Rubrik has been moving away from that for a while now, toward selling pure cloud software subscription, Rubrik Security Cloud, or RSC.

The old hardware bundles left behind a small accounting trace, though. Customers who gave up their right to a free hardware refresh got a cloud credit instead, and when they use that credit, it shows up as a one-time bump in revenue. Rubrik calls this “material rights.”

Material rights are shrinking fast, from $13.4 million a year ago to $8.5 million this quarter, and that shrinkage makes the reported revenue growth look softer than it really is. If you leave it out, then revenue grew 43%, not 39%. That’s even stronger.

At the Analyst Day, CFO Kiran Choudary said cloud ARR is now 89% of subscription ARR, up from roughly 60% a couple of years back, and much higher than the 80% long-term target Rubrik gave at its IPO.

He expects it to reach around 90% within four to six quarters and then to stay stable. For regulated and self-hosting customers, the legacy solution is still the best way to secure their data.

Source Analyst Day slides

I heard some investors claim that rising prices for servers, HDDs, flash, and DRAM are a problem for Rubrik. Founder and CEO Bipul Sinha denied this on the earnings call:

Rubrik has no real impact from the hardware market dynamics... the enterprise data protection part of our business is a relatively smaller part of our business now.

Rubrik doesn’t own the servers in a customer’s data center, nor does it own the cloud infrastructure. It mostly sells software. This reminds me of Cloudflare. Just like Cloudflare, Rubrik stays asset-light, without the huge capex spending. In these times, this is an overlooked competitive advantage.

The gross margin has now gone up five years in a row: 70% in FY23, then 77%, 78%, 82%, and 83% this quarter. And free cash flow margin nearly doubled in a year, from 12% to 19%.

This shows that the growth is combined with efficiency.

On the Analyst Day, the company also gave long-term targets.

These look all very good, although the “low single digits” for share dilution is a bit too vague for me.

Subscribe now

“The Cybersecurity Industry Is Dead”

That title is a direct quote from founder and CEO Bipul Sinha on the Analyst Day. His logic is that the old industry was built to prevent and detect attacks from human hackers using human identities. That world doesn’t exist anymore. Attackers use AI models that find vulnerabilities at machine speed, and a growing number of the “employees” in companies are AI agents, not humans. That means that a compromised agent isn’t a stolen password; it’s an insider attack. So:

You can’t respond by human speed, which is at best 10 bits per second, to an AI attacker, which is going million to billion bits per second. You will be out of business.

For a few weeks, the story was that this powerful model would kill the cybersecurity industry. But that’s very uninformed. It will definitely kill companies that still work in the old way, but the best have already adapted to AI and they will not be obsolete, but even more crucial.

Rubrik turned its own platform into an agent called Rubrik AI. It continuously monitors, builds a recovery runbook before an attack happens, and only brings in a human for the final yes or no. Chief Product Officer Anneka Gupta demoed this live at Capital Market Day, walking through a fictional ransomware attack in a Microsoft 365 environment in which the recovery plan was already built by the time the IT admin logged in.

Obviously, the cybersecurity market is not dead. Management put a $125 billion 2029 TAM (total addressable market) on the table for Rubrik.

Big TAM numbers often make me sigh deeply, but in this case, I think this is not outlandish.

Rubrik’s track record of how quickly each new product category has scaled after launch is also convincing.

Co-founder and CTO Arvind Nithrakashyap calls this the platform’s “complementary network effect”: each new product makes the existing ones more valuable, so the adoption compounds.

This is the breakdown of the revenue.

Chief Business Officer Mike Tornincasa runs the incubation team internally called Rubrik X. Products that came out of his team have grown at roughly 100% CAGR over five years, versus around 50% for the company overall, and now make up more than 40% of total company ARR.

To me, this also answers the question of how Rubrik can continue to grow so fast. At this scale, most slow down. Rubrik doesn’t.

The newest product from the Rubrik X team is Rubrik Flex. It’s a single contract that gives a customer access to the whole platform, and they can spend their commitment across whatever products they want. Announced at the Analyst Day. We have seen similar products at CrowdStrike (Falcon Flex).

The first Flex deal was a Global 2000 insurer that signed a contract for $6 million per year and over $30 million in total over five years. They started with three products and have room to grow into the rest.

Below:

The $50 million business Rubrik built from scratch in barely a year (and how it recovers customers 100x faster than Microsoft!)

Why "AI controlling AI agents" is the answer to the Mythos bear case

Five Selling Rules for Rubrik,

If the stock is still a buy after running 90% from the April lows.

And of course, much more.

Don't want to miss opportunities like this one again?

Grab your 20% discount now!

Read more

A few weeks ago, Topicus (TOI:CA) (TOITF) reported its Q1 2026 results. I started the earnings analysis last week, but I couldn't get it out and so, I'm finishing it on the plane on my way to Vail, Colorado.

As you probably know, I have a second service called Best Anchor Stocks, mostly run by Leandro. He is a Topicus shareholder and wrote a short earnings analysis for Best Anchor Stocks subscribers. This article took Leandro’s article as a basis, but I added some context, a different angle here and there, and a couple of extra things from the filings.

And as always, I’ll update the Selling Rules, the Quality Score and the valuation at the end.

The Numbers

Topicus grew revenue 23% year-over-year to €435.7 million, up from €355.6 million. This acceleration came with EBITA margin expansion, as expenses grew less than revenue, at 20% YoY.

Organic growth (growth that doesn’t come from acquisitions) was 5%, the average since Topicus became a public company.

Maintenance and other recurring revenue grew 7% organically this quarter, the median over all quarters.

You can see a more detailed breakout here.

That recurring revenue accounts for €316 million of the €436 million, or about 72.5% of revenue. It is the sticky, predictable, high-margin business. The fact that this grew by 7% organically is an indication that there is no negative impact from AI so far.

Licenses came in at €11.5 million (+23%), and hardware and other were €9.9 million (+35%), but those are not really meaningful for revenue. Professional services generated €98.2 million (+19%). Organic growth was slightly negative here, but I have no problem with that, because services are always the lowest-quality part of any VMS business.

Net income dropped by 21.4% to €55.1 million from €70.1 million a year ago, and diluted EPS dropped from €0.54 to €0.41.

This sounds like a problem, right? But it isn’t. It’s just accounting noise.

Topicus’ Q1 2025 net income was inflated by a €32.8 million non-cash gain on its Asseco shares. The company bought a large stake in Asseco in January 2025 at 85 PLN per share, but the stock was already trading higher than that, so it booked a paper gain. That’s what causes this drop, which is only accounting.

If you exclude that, pre-tax net income this quarter was €64.4 million, versus €47 million of underlying pre-tax income last year (after removing that €32.8 million). So, the normalized growth is 36%.

This is exactly why, with Topicus, you can’t just read the headlines. The cash flows are another example. The metric that matters most for Topicus is free cash flow available to shareholders (FCFA2S). That’s the cash Topicus can give to shareholders if it stopped with acquisitions after taking out everything that goes to minority partners and to Constellation, the parent. It’s the cleanest number on what the business actually earns and I’m happy the company always provides it.

Operating cash flow rose 3% to €280.5 million, and FCFA2S 2% to €165.4 million. After a 23% revenue quarter with expanding margins, you’d expect more than 2-3% cash growth, right? So why did it grow so little? There are two main reasons.

Behind the paywall, I look at that, Topicus’ enormous acquisition firepower, the three numbers that challenge the AI bear case, the hidden value in Asseco and, most importantly, whether a stock down 51.6% is now genuinely cheap. As always, I also update the Selling Rules, the PM Quality Score, the valuation and my Buy-Hold-Sell rating.

If you want the full value, subscribe now!

Subscribe now

Read more

Wow, what a trip this has been so far.

I picked CrowdStrike (CRWD) as a Potential Multibagger on June 29, 2020, almost six years ago exactly, at $98.10. That means it is almost a 7-bagger. But in typical Potential Multibaggers style, it has been very volatile and, I’m sure, will remain so.

This is how multibaggers go. Everyone wants to celebrate a multibagger, but very few have the stomach to hold on to stocks through the ups and downs.

Don’t be mistaken in thinking that CrowdStrike was an easy hold. It first became a threebagger in a relatively short time and then went back to square 1 in January 2023.

That means the stock was down 65% from the top. Of course, as always, the cynics were ready to chime in: to get back to even, the stock has to almost triple.

There are two problems with that.

The first false assumption is that everyone picked the absolute top. If you picked the absolute bottom, everyone would accuse you of cherry-picking, and they would be right. But if someone cherry-picks negatively, it just passes as if that’s normal. My irritation about this is one of my pet peeves.

Secondly, the percentages make it seem impossible. To be exact, the stock would have to win 186% to get back to the top. Including the stock price, that’s almost a triple. Sounds hard, right? But the stock just has to regain the lost dollar amount. Math is somewhat misleading here. Not factual, of course, but psychological. It makes it sound almost impossible or very hard. But if you look at the bottom, this often happens and usually very quickly.

Just look at what CrowdStrike did from January 2023.

As you can see, the stock is more than a 7-bagger from that point. I looked at when it exceeded its previous top. The poetic beauty of the facts is that it was exactly a year after that bottom.

As long as a company keeps performing fundamentally, the stock return will come. Sometimes it goes swiftly, sometimes you have to exercise years of patience. That’s the entrance ticket for outsized returns.

This is what Charlie Munger said about this.

But enough musings. It’s time for action. And Zack will help me. He will analyze the earnings, although I also contributed a few elements. Even when someone else writes, I still analyze, so I don’t lose touch with the stock.

After the analysis, I take over again for the Selling Rules, the PM Quality Score update and the Valuation to see if CrowdStrike is still attractive at this price, or if it would be wise to sell or trim. Take it away, Zack.

Hi Multis

Leading up to earnings, CrowdStrike went on a generational run, more than doubling from the AI-fear lows set earlier in the year.

The sentiment on X went from “CrowdStrike is going to be replaced by Claude!” to “CrowdStrike is overvalued!” in a matter of months.

CrowdStrike stock run(Fiscal.ai)

So what happened in the past 4 months? Did something fundamental change? Read on to find out.

Headline Numbers

Let’s first dive into the numbers.

Revenue grew to $1.39B, 26% increase year-over-year, exceeding estimates by 1.7%.

Annual Recurring Revenue (’ARR’) grew to $5.51B, growing 24% year-over-year.

As a reminder, for SaaS companies, ARR is often a better indicator than revenue because it more accurately reflects current and future obligations due to the nature of GAAP revenue recognition.

Net New ARR which came in at $256M, up 32% year-over-year indicating continual expansion of contracted revenue with existing and new business. A related note: both SGNL and Seraphic closed in Q1, contributing $7.8M of acquired net new ARR to this quarter.

Charting ARR over the past few quarters, we see the reacceleration of revenue continue after an initial slowdown in the quarters prior:

The growing ARR is also reflected in the increased module adoption that continues to march up and to the right:

ARR and adoption charts(Fiscal.ai)

There’s one more forward-looking number worth paying attention to: RPO or Remaining Performance Obligations. That’s revenue that is already under contract but that can’t be booked as revenue yet because CrowdStrike still has to deliver the services. RPO came in at $8.8B, up 29% year-over-year. That’s growing even faster than ARR and revenue. Probably, that’s a result of the bigger and longer contracts that CrowdStrike can sign.

Subscription Gross Margin came in at 78%, or 81% on a non-GAAP basis. Both were up by 1% compared to their respective quarterly numbers from last year.

Operating Income (Loss) was ($30.6M) compared to ($118.7M) in FY 2026. Non-GAAP income was $325M compared to $201M in FY2026. The biggest contributors to the swing in income were a reduction in expenses related to the July 19th incident (damages from the big outage), which dropped over 50% compared with last year, and reduced SBC (stock-based compensation) as a percentage of revenue.

While operating income is still in the negative on a GAAP basis (mostly due to SBC), we see that Operating Cash Flow continues to grow every quarter, reaching a record high of $591M, growing 54% year-over-year.

Net Income (Loss) was $27.8M compared to ($104.3M) in Q1 FY2026. Non-GAAP net income was $283.4M compared to $184.7M in Q1 FY2026. GAAP net income got an $18M boost from gains on strategic investments. CrowdStrike has a venture fund, the Falcon Fund, and probably one or a few of those investments did very well.

Earnings Per Share (diluted) on a GAAP basis was $0.11 compared to a loss of $0.42 in Q1 FY2026. Non-GAAP Earnings Per Share (diluted) was $1.10 compared to $0.73 in Q1 FY2026.

Free Cash Flow was $468.5M compared to $279.4M in FY2026 while maintaining a stellar 34% free cash flow margin.

Cash and cash equivalents stood at $4.55B. This rock-solid balance sheet continues to provide security and massive optionality via future acquisitions or buybacks.

In Q1, CrowdStrike repurchased $176 million in shares at an average price of $365.63, leaving approximately $1.3 billion remaining under their current authorization. Management noted they will remain opportunistic in returning capital while staying focused on capturing the significant growth opportunities ahead.

CrowdStrike also announced a 4-for-1 stock split, the company’s first as a public company. While stock splits are purely cosmetic, they can increase liquidity for both shares and options. Shares will begin trading on a split-adjusted basis on July 2, 2026.

Guidance

For Q2 FY2027, management guided:

• ARR: $5.793B – $5.795B (24% YoY growth),

• Net new ARR: $284M – $286M (28 – 29% YoY growth)

• Revenue: $1.436B – $1.442B (23% YoY growth)

For the full year FY2027:

• ARR: $6.531B – $6.556B (25% YoY growth); a slight raise.

• Net New ARR growth: raised to 27.7% at the midpoint which is a 520 basis point increase from prior guidance, implying $1.279B – $1.303B of net new ARR. This is a meaningful revision and signals that management sees future business expansion in the pipeline.

• Revenue: $5.915B – $5.959B (23 – 24% YoY growth); a slight raise.

• Free cash flow margin: At least 30% for the full year

Overall, these were incredibly solid numbers across the board, showing clear signs of top-line reacceleration and improving operational leverage as they shake off the final impacts of the July 2024 incident.

Now let’s move on to what CEO George Kurtz referred to as the “Mythos inflection moment”.

The Mythos Moment

First, a quick recap of what transpired over the past few months:

A shift began at the RSA Conference in late March 2026, slightly before Mythos was even announced. Founder and CEO George Kurtz mentions that every meeting he had that week boiled down to the same thing:

Every meeting was literally the same meeting all over. Help us protect these AI workloads that are running on the endpoints. All the developers are running it, marketing is running it, accounting is running it. We heard just crazy stories about AI run amok.

Organizations want to move faster in adopting AI without compromising security.

Then, in April 2026, more fuel was added to the fire with Anthropic releasing a new frontier model referred to as “Mythos,” which demonstrated an ability to identify software vulnerabilities and construct sophisticated cyberattacks at an unprecedented speed and scale.

Claude Mythos(Anthropic)

Source: Anthropic

Now, any reasonably technical person or AI agent can potentially conduct serious cyberattacks against enterprise infrastructure, and cybersecurity’s role in the enterprise has evolved overnight. Founder and CEO George Kurtz:

For the first time in my career, the market’s view of cybersecurity’s role has shifted from being viewed primarily through the lens of risk management, compliance, and protection to being recognized as a strategic accelerator and a critical enabler of AI adoption.

The funny thing is that if you have been following CrowdStrike for a few years, George Kurtz has said this forever. So, it was with pride that he said this. Cybersecurity is now a necessity for enterprises that want to leverage and accelerate AI quickly, as deploying AI across an organization without proper security controls is too risky. Secondly, on the defensive end, frontier AI models (e.g. Mythos) have increased the attack surface and lowered the barrier for adversaries, increasing the security gap enterprises were already facing.

In response to this potential threat, new programs were created: Anthropic’s Project Glasswing and OpenAI’s Trusted Access for Cyber (’TAC’).

CrowdStrike was the only cybersecurity company selected for both, underscoring its leadership in this space. These programs gave enterprise companies early access to frontier models to ensure that necessary cybersecurity guardrails were in place before the models were released to the public.

This quarter, CrowdStrike also took the initiative in launching Project QuiltWorks.

Source: CrowdStrike

This is an industry-wide coalition of the world’s top AI companies, consulting firms, and cyber insurance companies, all under one roof, with the goal of providing auditing services with the CrowdStrike platform to identify vulnerabilities within companies.

As you can imagine, all of this serves as a new marketing and sales pipeline for the business, positioning CrowdStrike as the thought leader and best-of-breed in this space.

It’s ironic how what initially started as a threat to the CrowdStrike business is now a sales generator. I mentioned that in last quarter’s analysis of the “AI Threat” and it looks like it’s playing out exactly as I saw it then:

To flip the script, rather than a threat, AI and the “agentic workforce” represent a massive new market for CrowdStrike. AI adoption means a new attack surface that must be secured. As it grows, CrowdStrike becomes an even more necessary layer of protection for the AI stack. We are already seeing this reflected in the growth of the modules (Identity, Endpoint, AI-DR, etc.) mentioned earlier in this article. CrowdStrike is set to benefit from this AI revolution, not be replaced by it.

This entire pivot was further proof to me of just how much the market is driven entirely by narratives. Especially in an environment with volatile price action, you can clearly see how negative narratives drive down prices, which then fuels an even more negative narrative. The cycle repeats endlessly until the story finally turns. Keep in mind, this whole roller coaster happened despite zero evidence of fundamental weakness.

So the narrative flipped, the fundamentals are stronger than ever, and the stock just doubled in four months. I bought back then, based on the valuation, which is in the paid part below.

Below: the customer behavior that proves how sticky this platform is, the AI giants that are now building on top of CrowdStrike, the exact price I’d buy more of this stock at, and what I’m doing with my position after a 7-bagger.

Don’t want to miss opportunities like this one again?

Subscribe now

Read more

Cloudflare (NET) reported its Q1 2026 results on May 7. I didn’t find the time to cover them right away and then I saw that it had its annual Investor Day on June 9. I waited a bit so I could cover both in one article. And here it is.

We’ve owned Cloudflare since I picked it at $39 in 2020, and that definitely helped me to unpack everything and put certain things in perspective.

Let’s walk through the numbers first.

The Numbers

• Revenue: $639.8M, +34% YoY. That’s the same growth as Q4. It beat the consensus of around $622M by about $17M or 2.7%.

• Non-GAAP EPS: $0.25

• Non-GAAP operating income: $73.1M, or 11.4% of revenue, up from $56.0M a year ago.

• Free cash flow: $84.1M, up from $52.9M in Q1 last year.

• Free cash flow margin: 13%

• RPO: $2.543B, +36% YoY. RPO is the remaining performance obligation. That’s revenue already under contract but not yet booked because the service still has to be delivered.

• DBNRR: 1118%. (dollar-based net retention, how much customers Cloudflare had a year ago are spending now, including the churn of customers that stopped).

• Large customers (over $100K a year): 4,416, +25% YoY. This is now 72% of revenue, up from 69% last year, and grew by 38%.

• Very large customers: deals over $1M grew 73% YoY. Cloudflare added as many $5M+ customers in this quarter as it did in all of last year.

Cloudflare has said on multiple occasions that investors should not focus on the DBNRR, but I still read a lot of takes saying it was bearish that DBNRR dropped from 120% in Q4 to 118%.

Cloudflare said that its sales teams focused on new businesses, and not as much on the expansion of existing customers. But still, retention was great.

Founder and CEO Matthew Prince on the call:

Our quarterly gross retention reached its highest level in four years, reinforcing that customers understand Cloudflare is a must-have rather than a nice-to-have.

Gross retention is different from dollar-based net retention. It looks at what you had in revenue from your customers last year and you withdraw all the churn of customers who left, but you don’t count what they spent more. So, the maximum is 100%. The highest gross retention in four years means very few customers are leaving.

The focus on new customers paid off. New business grew at its fastest rate since 2023.

Guidance went up as well. Management raised the full-year guidance to $2.805 billion to $2.813 billion, about 30% growth at the midpoint. For Q2, Cloudflare guided to $664 million to $665 million, also around 30%. I’m pretty confident the real result will be higher.

Also important: network capex is expected to be 14% to 15% of revenue for the year, which is still almost nothing compared to the tens of billions the hyperscalers are investing into data centers. And knowing this team, they’ll probably beat all of it.

The 20% Layoff

On the same day it reported this quarter, Cloudflare also announced it was letting go of about 1,100 people, about 20% of the company. That’s the news that made all the headlines, and I understand why it shocked people.

A company growing 34% laying off a fifth of its employees sounds bad. But when you look at what Cloudflare said about this, and at what it has been doing internally, it’s a very different story.

The cuts hit every function and every part of the world, with one exception: the salespeople who carry a quota and sit in front of customers. Those were barely touched, and management still expects its sales capacity to grow this year. The whole reorganization will cost $140 million to $150 million (about $40 million of that non-cash), mostly in the second quarter, and it should be done by the end of the third. The free cash flow targets for the year didn’t change.

So why do this now, when things seem to be going well? Because Cloudflare has ramped up the use of its own AI tools internally for the last six months. The results convinced management to act.

Cloudflare has been pretty cautious with AI internally, but that changed in November. Since then, people suddenly got much more productive, in some cases two times, ten times, even a hundred times as productive. Matthew Prince described it in his typical colorful language:

It was like going from a manual to an electric screwdriver.

The numbers the company shared were impressive. Cloudflare’s use of AI is up more than 600% in three months. In engineering, 97% of the team now uses AI coding tools, all of them running on Cloudflare’s Workers platform. People in finance, HR, and marketing all use AI intensely. That’s on an internal system the company built and calls Cloudflare OS. On the Investor Day, a few interesting examples were given.

The layoff is the consequence of that. Prince was clear about what it was and wasn’t:

This isn’t a cost-cutting exercise or an assessment of individuals’ performance.

To me, this is a clear signal that the second AI wave has begun. The first wave of the AI boom rewarded the companies that built the data centers, the chipmakers and memory suppliers. A lot of that (but maybe not everything) is now priced in. The next wave will reward the companies that take all that AI and turn it into real earnings by changing how they operate. Cloudflare is one of the first to do that systematically.

It also helps to compare this to the other layoffs we’ve seen this year. Plenty of big technology companies have cut staff, but most did so because they had to. Often, that was to compensate for their enormous data center investments. Cloudflare cut while growing 34% and raising its guidance. That’s a very different situation. And it’s actually a very good thing to do.

The other thing I appreciated was the honesty. Cloudflare paired the layoff with what it called industry-leading severance, and instead of hiding behind the usual corporate language, Prince called it a hard day and said the people leaving will go on to build great things. He also made clear this isn’t a company shrinking. He expects to have more employees in 2027 than at any point in 2026. When an analyst asked why he would do this right after such a strong quarter, he put it like this:

We’re the fittest we’ve ever been, but we’re gonna get even fitter to win the next chapter.

I respect this move a lot. I’m sure Cloudflare will take on many more employees once the new system is in place. Right now, there’s overcapacity. But the companies that will win over the long term are those that can leverage the efficiency gains at a larger scale and I think Matthew Prince knows this better than anyone else.

Sales productivity went up year-over-year for the ninth quarter in a row. Hiring of new salespeople grew at its fastest pace since 2023. We already saw that customer bookings grew the fastest since 2023. They also filled the pipeline at the fastest pace in five years and beat their internal plan for the quarter by more than they have in any first quarter since 2021, during that crazy COVID peak.

The customers keep getting bigger, too. Large customers now make up 72% of revenue, up from the high 50s a few years ago.

The Agentic Internet, Cloudflare’s Biggest Tailwind

Below: Cloudflare’s premium role in the agent traffic, why OpenAI and Anthropic both build on Cloudflare, the "Google Zero" problem of the internet and how Cloudflare seems to be the only one working on a solution, why the dropping gross margin is actually good news. And of course much more, like the valuation.

The value you get for your subscription is priceless!

Subscribe now

Read more

Kinsale (KNSL) reported its Q1 2026 results quite some weeks ago, but I had not covered them yet. Since the earnings, the stock is down about 13%.

If you zoom out a bit more, you see that Kinsale has the worst drop since it became a public company in July 2016. The stock is down almost 45%.

Last quarter, when premium growth slowed to 1.8%, I made the case that the slowdown was intentional and good. Kinsale was leaving underpriced business, to the competition.

So which is it this time? Is this a disciplined company riding out a soft market, or has the growth story simply ended? That’s an important question, as it means more or less the same as: Is the investment thesis still intact?

The Numbers

Gross Written Premiums: $482.0 million, down 0.5%. Yes, down. As far as I know, it’s the first decline. This is probably the main reason Kinsale’s stock dropped. Let’s part it for now, but obviously, we’ll get back to this.

Net Written Premiums: up 5.6%. Gross down, net up. We get back to that later as well, of course.

Net Earned Premiums: $406.9 million, up 11.2%. Earned premium is the part of written premium that gets “used up” as a policy runs off, and it is the part that shows up as revenue. It lags written premium.

Underwriting Income: $94.5 million, up 40%.

Combined Ratio: 77.4%, down from 82.1% a year ago. The combined ratio is the loss ratio (claims paid out) plus the expense ratio (the cost of running the business). Under 100% is an underwriting profit. So 77.4% means Kinsale kept about $22.60 of every $100 in premium as pure underwriting profit. That’s before a single cent of investment income. This remains very impressive.

Net Income: $112.6 million, or $4.88 per share, up 26.1%.

Net Operating Earnings: $117.8 million, or $5.11 per share, up 37.7%. Operating earnings leaves out the ups and downs of the stock portfolio, so it shows the insurance business more cleanly.

Operating Return on Equity: 24%. The average US insurer sits somewhere in the low teens, so this is still outstanding.

Book Value Per Share: $85.31, up 25.6% in a year. Great result.

Net Investment Income: $55.4 million, up 26.5%.

Float: now $3.3 billion.

Here’s an overview from the earnings slide deck.

So: EPS up almost 38%, a 24% ROE, book value growing at 25%, a combined ratio under 78%. Read these numbers and you wonder why the stock is anywhere near its lows. The top line growth and the quality of the earnings are the reason.

Below, for paid subs: why the biggest division is shrinking on purpose, the new competitors that worry me more than the giants, and what I'm doing with my position on Monday.

4000 words you can’t read as a free sub.

Subscribe now

Read more

Datadog (DDOG) reported its Q1 2026 results on my mother’s birthday (May 7), and I had not reviewed the earnings yet. Now that I have, I can say it was potentially the best quarter this company has ever delivered. And the market agreed. The stock is still up 58%, even after the drop in the last few days. It was up 90% just a few days ago. Crazy.

If you want the best stock screener, the best charts, the best analyst estimates, the best earnings calls, Morningstart research, and sooooo much more, join Fiscal and get 15% off.

I keep repeating this, but it’s good to see this live for long-time Multis. Stocks can do nothing for years and then suddenly shoot up like a rocket over a small period of time. As long as the company keeps performing and the fundamentals stay intact, this often happens. But it can take years of patience. Years of nothing and then suddenly everything. That’s the stock market for you.

Datadog’s outstanding quarter also crushed one of the most superficial bear cases I heard about Datadog: that AI is going to eat it. At the end of the article, you’ll understand why this is ridiculous.

The Numbers

Datadog’s Q1 revenue came in at $1.006 billion, up 32% year-over-year. The consensus was $960 million, so this was a $46 million beat, or 4.8%. It was the first time Datadog crossed a billion dollars in a single quarter. A nice milestone.

This was the fourth consecutive quarter of accelerating revenue growth, and not by a bit. Just look at this beautiful climbing line.

It’s impressive that Datadog can do this at this scale.

CFO David Obstler said on the call that ARR (annual recurring revenue) growth accelerated in each month of Q1, and management said the trend continued in April. I think that’s one of the reasons why the market reacted the way it did to these earnings.

Non-GAAP EPS was $0.60, beating the consensus of $0.51 by $0.09. Non-GAAP operating income was $223 million, good for a 22% operating margin. Free cash flow was $289 million with a 29% margin.

Datadog ended Q1 with about 33,200 total customers, up from about 30,500 a year ago.

When you plot this on a chart, which Fiscal can do, of course (get 15% off here), you see something interesting.

Customer growth is not responsible for the revenue acceleration we saw. So, what is? This.

If you wonder why the green lin, with customers with ARR of more than $1M stops a quarter earlier, it’s because Datadog only provides once a year. But you see the trend. It’s big customers that power the growth. There were about 4,550 customers with an ARR of $100,000 or more, up 20.7% from 3,770 a year ago. For the biggest customers, it’s probably even more. Don’t forget that Datadog has usage-based billing and with more AI, that means more usage, as I have always said, long before the market agreed.

Net retention has not moved in a long time. It’s always 120% or a bit higher.

The Core Business Keeps Accelerating

For most of 2025 and until a month ago in 2026, software stocks were beaten to a pulp on the fear that AI was going to disrupt or replace SaaS. The narrative was that LLMs and coding agents would make a lot of cloud software obsolete. Anything that wasn’t directly an AI play looked vulnerable.

And then Datadog reported. It was the turnaround for software, it seems. CEO Olivier Pomel on the call:

Non-AI customer revenue growth accelerated again this quarter to mid-20s percent year-over-year, up from 23% last quarter and 19% in the year ago quarter.

So, the non-AI part of Datadog’s business, the part that the market wrote off, just posted its third consecutive quarter of accelerating growth. From 19% to 23% to mid-20s. Or put differently, traditional SaaS customers accelerated their Datadog spend. SaaS dead? Hahaha.

This is the Anti-Non-AI Bubble I have talked about for months now. The SaaS market is much bigger than everyone seems to factor in now. Founder and CEO Olivier Pomel confirmed how much growth Datadog can still have just by taking market share alone already. From the J.P. Morgan conference two weeks after the earnings:

We’re the leader in observability, we’re #1 there, but we still have only 13.6% of the market according to Gartner. So this tells you how early it is in the market for us, how much opportunity there is ahead of us, even factoring out all of the new developments with AI.

The AI-native cohort, of course, is also growing rapidly. Datadog now has 22 customers spending more than $1 million annually on AI-related work, and 5 spending more than $10 million. And Pomel disclosed at the recent Investor Day in February that AI-native customers represent about 11% of Datadog’s total revenue today, up from 5% in 2024, and just 2% in 2023.

But many seem to forget that the other 89% is everything else, and that part is accelerating faster than it has in years. Datadog is in a premium position. it gets the AI tailwind on 11% of its revenue (and growing fast), and the SaaS tailwind on 89%. What a great double engine!

The Training Reversal

There was something else that caught my ear on the earnings call that I really liked. CEO Olivier Pomel on the call:

Interestingly, last year, when we reported earnings, we said we’re mostly interested in inference workloads and training is not really a market for us yet. Now we actually see training becoming a market.

That’s an important reversal. So, what changed? Pomel explained the mechanism in more detail at the J.P. Morgan conference:

Models went from being mostly pre-trained to being largely post-trained. The post-training was becoming increasingly specialized to different types of verticals. The stacks that are used for post-training also are becoming richer and richer. And we saw instead of having 5 to 10 companies doing that, now there were 50 to 100.

Twelve months ago, training wasn’t a market for Datadog but now it is. How can this change so fast? Olivier Pomel explained at the J.P. Morgan Conference that the kind of training had changed.

Pre-training is the huge, expensive first step in which a model learns from scratch on incredible amounts of internet data. Only 5 to 10 companies worldwide could afford that. Post-training is what comes after: based on the pre-training, you can teach the model to be good at something specific, like coding, or medical questions, or customer support. It’s cheaper and easier, and now 50 to 100 companies are doing it, but that could soon be 500 or 1,000. This is a very interesting development and another reason Datadog looks well-positioned to profit from AI.

In Q1, Datadog brought in two new deals with the AI research divisions of two of the world’s largest tech companies. One was a 7-figure deal, the other even an 8-figure deal. These are hyperscalers with top-notch engineering talent and a strong culture of building everything in-house, but they still came to Datadog. That says a lot.

Olivier Pomel on the earnings call:

Hyperscalers typically have a culture of building everything themselves, and they certainly have the balance sheet and the human capital to support some of that build-out. Like if there was ever a set of companies for whom it makes sense to do it themselves, that would be those companies. And yet, we see that they have the same issues.

Don’t forget that Anthropic also signed a bit contract. We discussed this last quarter.

Below: the rest of the AI story, why every huge AI company uses Datadog, and what I'm doing with the stock right now.

When you join Potential Multibaggers now, you get my entire investment system:

✅ Best Buys Now: every month, the 5 best stocks to buy

✅ My Proprietary Quality Score: rating companies on 17 quality metrics nobody else uses

✅ Deep Earnings Analysis: thorough breakdowns that reveal what really matters

✅ Valuations & Buy/Hold/Sell scales: cut through the noise and know exactly when to act

✅ 15,000+ Word Deep Dives: each pick gets 4-5 extensive articles so you understand exactly what you own.

✅ Private Community: direct access to me and 800+ serious long-term investors.

Subscribe now

Read more

Celsius Holdings (CELH) reported its Q1 2026 results on May 7, and the stock kept dropping. It’s now 12.4% lower than before the earnings.

Year-to-date, the stock is down 34% and from its 2024 all-time high, it’s down almost 70%.

The real question is, of course, whether the market is wrong and gives us an opportunity to buy, or if the weak stock price is a sign that there’s more.

Let’s look at the numbers first.

The Numbers

• Revenue: $782.6M, up 138% YoY, beating the consensus by $22M or 2.9%.

• Adjusted EPS: $0.41, up 128% YoY, beating the consensus by $0.12 or a whopping 41.4%.

• GAAP EPS: $0.33, up 120% YoY.

• North America revenue: $747.3M, up 144% YoY.

• International revenue: $35.3M, up 55% YoY.

• Brand split: Celsius $348M (+6% YoY), Alani Nu $368M (+60% pro forma YoY), Rockstar $67M.

• Gross margin: 48.3%, down from 52.3% last year in Q1, but better than the 47.4% in Q4 2025.

• GAAP net income: $110.1M, up 148% YoY.

• Adjusted EBITDA: $195.5M, up 181% YoY. EBITDA margin of 25.0%, up from 21.2% last year.

• Buybacks: 700K shares for $24.1M at an average price of $35.39. $236.1M still left of the $300M buyback program.

So on the surface, the results look really strong. Celsius posted substantial beats of the consensus, with better margins. And yet, the market didn’t seem to care. Let’s look into more detail.

The Celsius Brand Problem

I start here because I think the 6% growth for the Celsius brand is why the market doubts the company’s growth story. For the first time, Alani Nu was bigger than Celsius. Revenue for Alani was $368M, compared with $348M for Celsius.

A part of that is seasonality. Celsius is the fitness brand of the portfolio, and apparently, people drink fitness drinks more in spring and summer than in January and February. Alani Nu is more seen as an everyday brand. For full-year 2025, Celsius did $1.46 billion and Alani Nu (acquired in April 2025) did roughly $1 billion. So, on an annual basis, Celsius was still bigger last year. But Alani Nu posted 60% revenue growth versus 6% for the Celsius brand. Is it in trouble? Monster grew more than 20%. That 6% growth for the flagship brand looks very weak.

At the Deutsche Bank conference on June 2, John Fieldly explained what’s going on. Before PepsiCo, Celsius was distributed by more than 300 independents. This is important to know.

To get shelf space at certain retailers, Celsius often did custom flavors. Fieldly gave an example. A convenience chain of 300 stores didn’t believe Celsius would sell, so the company created a Cherry Cola exclusively for them. It worked, and therefore, they could also sell Celsius on the shelves. That flavor only existed for those 300 stores. And there were many flavors like that.

In the PepsiCo system, that approach doesn’t work. If you have national distribution, you need national consistency. So Celsius is removing all regional flavors and focusing on the national flavors. Fieldly:

We want to have a consistent portfolio around the country. As we say internally, we want to keep the fast cars on the track.

Each flavor removal immediately results in lost revenue. That’s a big part of why the Celsius brand grew only 6% this quarter.

The core Celsius 12-oz cans grew 11% in convenience stores. So, that proves Fieldly’s story. The assortment is shrinking, but what remains of the legacy brand is doing well.

Fieldly also said something at the conference you don’t often hear from a CEO:

Especially on Celsius, we should have had more innovation. That would have added more complexity with the integration of Alani and Rockstar, but we should have had more innovation, and we’re cycling over 6 innovations versus last year.”

So Celsius underinvested in flavor innovation on the flagship brand. The reason was the integration of Alani Nu and Rockstar. Understandable. Last year, there were 6 innovations, so that makes the comps tougher. Management will continue to focus on the integration now but go full-in on innovation in 2027.

Alani Nu: Firing On All Cylinders

If the Celsius brand looks weak but is better than it looks at first sight, Alani Nu is firing on all cylinders. I think we can already say now that this was a killer acquisition for Celsius. That it would overtake Celsius in sales so fast was really unexpected.

Revenue was $368M, up about 60% pro forma. Pro forma means this is against Alani Nu’s sales last year, before Celsius bought it. Scanner data, which measures what consumers bought, even showed the brand up 100% YoY.

So consumer demand grew faster than the reported revenue number suggests. That’s important, and it’s worth explaining why the two numbers are quite a bit apart.

The company gave us a breakdown:

Alani Nu sold $230.7M in Q1 2025, the quarter before Celsius bought the company. But not all of that was from the US energy drink market. Take out $32.6M from Canada and from non-energy products like protein bars, and you get $198.1M of U.S. energy drink sales. That’s the right number to compare this quarter with.

From there, scanner data shows U.S. energy drink sales grew 85% YoY, not the 100% I mentioned earlier. The difference is Cherry Bomb, a limited-time flavor that Celsius shipped to stores in Q4 2025. Consumers bought most of those cans in Q1 2026, but the company already booked revenue when the cans shipped, so the revenue was in the Q4 results. Without that distortion, growth would have been 85%.

Apply 85% growth to that $198.1M, and you get about $168M of additional revenue. That’s the white bar.

Then there’s a $ 26.5 M deduction. When Alani Nu was sold directly to retailers like Walmart, Celsius booked the full wholesale price. Now that Alani Nu is being sold through PepsiCo’s DSD (Direct Store Delivery) system (which means Pepsi trucks deliver cans straight to store shelves instead of to a retailer’s warehouse), Celsius sells the cans to Pepsi first, and Pepsi takes a cut before reselling them to retailers. So even if the same number of cans are sold to consumers, the revenue Celsius reports per can is lower.

So the takeaway is: consumer demand for Alani Nu grew by 85%. That’s phenomenal growth. Of course, even if Pepsi takes a part of that, volume is much bigger and on top of that, Celsius also had to pay its previous distributors, but in another way.

At the Deutsche Bank conference, Celsius shared that Alani Nu’s availability in convenience stores went from 65% in January to 91% in May. That’s a sea change.

Normally, when a brand expands distribution that fast, sales per store go down. The new stores are usually lower quality and it takes time for shoppers to find the product. With Alani Nu, sales per store actually went up over those five months.

The bear case on Alani Nu is that it’s a flavor brand, riding LTOs (Limited Time Offers), like Cherry Bomb and Lime Slush, that are fashionable for a few months and then disappear.

The fear is that once the LTO cycle slows, the brand fades. But the data shows the opposite. The brand is reaching more stores and the consumer is buying more per store, not less. That’s a really strong indicator that the brand has staying power.

There’s a related concern: that Alani Nu is just stealing sales from Celsius. Fieldly was asked this directly at the conference. His answer was that there’s some overlap, but Alani Nu is also reaching consumers that Celsius wasn’t. I think he’s right, and the proof is in the share number. If Alani Nu were just stealing Celsius’s sales, then the core Celsius brand wouldn’t have grown by double digits on a normalized basis (see above for that).

The market got the two big numbers wrong. The question is whether it’s also wrong about the rest, the margins, the Costco threat, and the current stock price. Below are my answers.

Subscribe now

When you join Potential Multibaggers now, you get my entire investment system:
✅ Best Buys Now: every month, the 5 best stocks to buy
✅ My Proprietary Quality Score: rating companies on 17 quality metrics nobody else uses
✅ Deep Earnings Analysis: thorough breakdowns that reveal what really matters
✅ Valuations & Buy/Hold/Sell scales: cut through the noise and know exactly when to act
✅ 15,000+ Word Deep Dives: each pick gets 4-5 extensive articles so you understand exactly what you own.
✅ Private Community: direct access to me and 800+ serious long-term investors.

Subscribe now

Read more

It’s not always easy in this overheated market to find opportunities, but I think this is one. Patient investors can now buy this stock at the cheapest valuation it has ever been, despite 33% revenue growth.

The company has no debt, a ton of cash and high free cash flow and it’s totally under the radar of most investors.

There are two headlines that may screen ugly if you don’t know the company’s context but that actually show strength.

And a negative catalyst on GAAP earnings just stopped and the company did an interesting acquisition. It also showed how AI already adds value. Not with big words, as many do, but with concrete numbers.

Curious?

Paid subscribers get this stock.

Wanna join? Don’t wait.

Subscribe now

Read more

The poll I launched on Slack showed that you were looking most forward to the Tempus and SEA earnings analysis and you got those two first. The number three was Hims & Hers. So, here’s that one.

I wrote in my previous earnings analysis that this is by far the most volatile stock I’ve ever covered, and nothing has changed. Right now, the stock is down 55% over the last year.

If you want a Fiscal subscription, you get 15% off with this link.

So, is this a moment to buy or should we be cautious? Let’s find out.

The Numbers

• Revenue: $608M, up 4% YoY.

• US revenue: $530M, down 8% YoY.

• International: $78M, up 969% YoY.

• Subscribers: 2.58M, +9% YoY. Personalized: ~1.7M, +20% YoY.

• Monthly ARPU: $80, down from $85 a year ago.

• Gross margin: 65%, down from 73%.

• Adjusted EBITDA: $44M, a 7% margin, down from $91M and 16% in Q1 2025.

• GAAP net income: −$92M, versus +$49M last year.

• Free cash flow: $53M.

• Guidance Q2: revenue $680–700M (+25 to 28% YoY), adjusted EBITDA $35–55M.

• Guidance FY2026: revenue raised to $2.8 to 3.0B, above the consensus of $2.71B.

• 2030 target: reaffirmed $6.5B+ revenue and $1.3B+ adjusted EBITDA.

Those numbers don’t look great, do they? So, I understand the questions and FUD (fear, uncertainty and doubt) around HIMS.

Until Q1 2025, everything seemed to go well. Revenue was up a lot every quarter and ARPU (average revenue per user) peaked in Q1 2025.

Right now, in Q1 2026, revenue was down QoQ and up only 4% YoY, margins were down, and it posted the first GAAP loss since Q4 2023. Add a miss of $8.75M for revenue and a miss of $0.03 for EPS and it’s not too crazy to think the company is falling apart before your eyes.

I’m going over a few things, though, that could put things in a different light.

The Revenue & EPS Miss

Let’s take the revenue first.

In March, the company stopped advertising compounded GLP-1s and shift to branded weight-loss products, mainly Novo Nordisk’s (NVO) Wegovy pen and Wegovy pill, although Eli Lilly’s Mounjaro is also possible.

I wrote about the mess that came before that decision: the $49 compounded pill that lasted just 48 hours before the FDA, the DOJ and the SEC came knocking and it was withdrawn. The branded pivot is the cleanup of that mess. It took the GLP-1 risk off the table.

But it changes how the money comes in. The compounded business sent out medication in longer cycles, while branded GLP-1s go out monthly. That means there is less upfront recognition. It’s a bit similar to a software company switching from upfront software sales to a SaaS model. So, it’s often the same dollar amount (or even more), but recognized over more periods instead of all at once. In accounting, that looks like a revenue decline.

CFO Yemi Okupe was very clear about this:

These dynamics affect only the timing of revenue recognition and not customer demand or engagement.

That timing change removed about $65 million from Q1 revenue. If you add it back, the quarter would have come in at $673 million, quite a bit higher than the $652 million consensus. So the revenue “miss” wasn’t really a miss. Of course, if revenue is lower, EPS will be lower as well. So, that “miss” had the same cause.

And don’t forget the comparison. Q1 of last year was the best quarter in the company’s history, as I already showed on the chart. GLP-1s had just started on the platform and sold like crazy. So, while the 4% YoY growth looks very weak, it’s not really how you should look at this quarter. On a comparable basis, revenue would have been up about 15%. That may not be a number that excites you, but it’s not as depressing as that 4% revenue growth, right?

US revenue fell about 8% to roughly $530 million, but it would have been up 4% YoY on a comparable basis. Again, not something that will make you jump around in joy, but still a lot better than what you think on first sight.

International revenue jumped almost 10x to $78 million. Of course, this is on a small basis, but still, it’s important. You can see the opposite movements in this chart.

The $92 Million GAAP Loss

Now the GAAP net income loss. I saw this scared quite some investors. After all, the company had been GAAP profitable in every quarter since Q4 2023.

The company gave us the full breakdown, so let’s use it. It’s in the appendix of the earnings call slides. Most will probably never look at that, but we do. I see many takes against adjusted numbers vs. GAAP and to a certain level, I understand that. But adjusted numbers can sometimes give you insights that GAAP earnings can’t. That was definitely the case this time.

After stock-based compensation, which is a quarterly item and hasn’t moved much, the biggest piece is restructuring: about $33 million, of which $28 million is in write-downs related to the compounded GLP-1 supply chain. If you build infrastructure for a business and then you decide to leave it, you write it off. That’s a one-off cost. Take it out, and the gross margin was about 70%, right in line with where Hims & Hers has always been.

And there are other one-offs and non-cash items. There’s a ‘loss’ of $17.6M that’s a provision because Hims & Hers expects that the acquisitions will have higher earn-outs. That’s exactly what you want. Maybe you know this from the contingent payments Topicus and Constellation have. There are certain milestones. You take a base case for your accounting, but if the companies do better than expected, the same percentage of earnings is a higher number. Or because you cross certain expectations, that percentage can even be higher. In other words, that’s something to celebrate, even if this is seen as a “loss” in accounting.

There was also $13.4 million in M&A costs. Hims & Hers has bought quite a few companies: Eucalyptus, YourBio, ZAVA and Livewell. The acquisition process produces costs (think of legal, for example) and the integration process as well.

Then there’s something we saw earlier this quarter from Shopify and Tempus, for example. There’s $9.7 million in losses because some of the investments Hims holds have dropped in value on paper, even though the company hasn’t realized them.

If you look for other numbers, you can look at $89.5 million operating cash flow and $59.5 million in free cash flow this quarter. I think that shows better how Hims & Hers did this quarter. Even if you want to deduct stock-based compensation, you get $22.7M in FCF-SBC. That shows that the GAAP loss is mostly accounting.

Novo Nordisk GLP-1

So why did Hims & Hers do this? The change to branded GLP-1 provides the company (and its investors!) with much greater regulatory certainty. There’s no more Novo Nordisk litigation, and the company goes out of the way of the FDA and the SEC.

And it’s working. Within six weeks of putting Novo’s products on the platform, Hims & Hers fulfilled more than 125,000 Wegovy shipments. Founder and CEO Andrew Dudum on the conference call:

We see it in the scale of new customers coming in the door today, which is larger than as Yemi said, any spikes we saw even during the most important seasonal campaigns such as New Year’s and Super Bowl campaigns.

The New Year and the Super Bowl campaigns are the two most important campaigns for the company. Now, the new ‘normal’ pace after the Novo Nordisk deal is bigger than those peaks. That’s not in the numbers yet, but it’s an important data point. Management says it’s on track to add more than 100,000 weight-loss subscribers per month.

If the company can keep up that pace, subscriber growth would probably land somewhere in the low 30s a year from now.

Subscriber growth looks slow right now, but that is mostly because these new subscribers are just starting to come in. We should see acceleration in the next quarters. The early engagement numbers look good as well. Nearly 90% of these new subscribers download the app, and, on average, they speak with a doctor or nurse from the Hims & Hers care team three times in the first month. That tells me these subscribers are engaged and that makes them more likely to stay.

On the economics, CFO Yemi Okupe said branded and compounded are “roughly comparable” on a dollar basis. The branded GLP-1 has a higher price but a lower margin, and the compounded was the other way around, so they more or less even out. So the company gave up a high-margin product that carried a lot of legal risk and replaced it with a fully legal one at a lower margin that brings in more people. I think that makes sense. It takes away a major risk while still driving growth.

I find it ironic. In the US, Hims & Hers and Novo Nordisk now sell Wegovy through a partnership. That is the company that tore up the first deal and accused them of “illegal sham compounding” last year. Providers on the platform can also send prescriptions to Eli Lilly’s LillyDirect. And then last week, Hims launched generic semaglutide in Canada at C$149 a month, undercutting everyone, on the same molecule Novo uses. It’s important to know that Novo Nordisk’s patent in Canada has expired.

There’s a second irony. While Hims & Hers reduced a major legal risk by partnering with Novo Nordisk, it also introduced a new risk. The partnership already broke before. But right now, Novo Nordisk’s Wegovy is a big revenue growth driver. If the relation gets troubled again, this might have serious consequences for Hims & Hers. But that risk is worth taking more than the legal troubles that the company might have faced otherwise.

Not A GLP-1 Company

I didn’t buy this stock for the GLP-1s. I’ve written that many times before. If this were a GLP-1 company, I would never have been interested. The GLP-1s may be 30% of revenue or less this quarter, given the accounting change we saw earlier. But people keep staring at that one part of the business and completely ignore the rest of the platform, which accounts for 70% or more of the revenue.

That platform is an attempt to bring personalized, proactive, data-driven healthcare. Peter Attia writes about this in his book Outlive, and calls it medicine 3.0. Hims & Hers not only wants personalized healthcare but is also on a mission to make it affordable for everyone. Right now, that kind of care costs the wealthy tens of thousands of dollars a year. Founder and CEO Andrew Dudum put it like this on the call:

Netflix and Spotify reshaped how people could access not only a broader range of content, but also the best the industry has to offer. Health care must evolve towards that same consumer-oriented distribution model.

This quarter, some things pointed to this again. Take testosterone, for example. Tens of thousands of men now already take it, and the early results are strong. Andrew Dudum:

More than 95% of individuals utilizing a testosterone support offering experienced an increase in testosterone levels within the first 2 months of treatment, with an average increase of over 80%.

Most men never even know they have low testosterone levels. They just feel tired and blame it on their age, stress or something else. With Hims, they can take an at-home blood test, get treated, and then retest to see if it worked. That is a Medicine 3.0 example, because you find the problem before the patient even knows to look for it.

Then there is Labs. You take a blood test at home, and the app reads out more than 130 biomarkers for you, with AI helping to explain them. Andrew Dudum on the call:

On the Labs side, also, it’s just providing people data that, frankly, used to cost somewhere between $5,000 or $10,000... And then from there, 70% of those people are identifying an area of concern and an area of clinical risk that the platform can actually help treat.

So seven out of ten people who run Labs tests find something they should treat and they can find the treatment on the same platform that just identified the problem. That is great for cross-selling, of course. It’s the kind of preventative care I want to see. And it makes Labs a great entry point to everything Hims & Hers does and sells.

Sexual health and hair loss each took three to four years to cross $100 million in annual revenue. Weight loss did it in seven months.

Andrew Dudum thinks testosterone, menopause and Labs can each get there soon as well.

The GLP-1 noise is the easy part. The hard question is whether there's a real moat underneath, and that's exactly what comes next, along with my verdict on the stock.

If you are a paid subscriber already, you can just scroll past this. If you are not, please read this.

When you join Potential Multibaggers now, you get my entire investment system:

✅ Best Buys Now: every month, the 5 best stocks to buy

✅ My Proprietary Quality Score: rating companies on 17 quality metrics nobody else uses

✅ Deep Earnings Analysis: thorough breakdowns that reveal what really matters

✅ Valuations & Buy/Hold/Sell scales: cut through the noise and know exactly when to act

✅ 15,000+ Word Deep Dives: each pick gets 4-5 extensive articles so you understand exactly what you own.

✅ Private Community: direct access to me and 800+ serious long-term investors.

Subscribe now

Read more

Two weeks ago, Tempus AI (TEM) reported its Q1 2026 results. The stock fell about 12% in the days that followed, and it’s now down about 55% from its October high of around $103.

Tempus was the most requested earnings analysis in our community. Your wish is my command!

I want to start with a contradiction. CEO Eric Lefkofsky opened the call by saying, “We had a great quarter,” and the press release talks about accelerating demand. But what if you read this? Revenue down sequentially, EBITDA negative, hereditary growth collapsing, profitability out the window. That sounds pretty bad, doesn’t it?

The difficulty is that both things are true and that can be confusing. So, this is a quarter where you have to know the context before you judge and that’s what I will try to provide in this article.

But first, a quick reminder of what this company is, as it’s not one that many really understand.

What Tempus Does

Many see Tempus as a diagnostics company and they are. But the tests are just the start of the story. The real business is using those tests to collect enormous amounts of patient data, DNA, RNA, imaging, and clinical records, and then structuring that messy data with AI and licensing it to pharmaceutical companies that can’t get it anywhere else. The confusing part is that this business currently accounts for only about 25% of the company’s revenue. 75% is still the tests.

Of course, the data business has much higher margins. And there’s a flywheel. The more data comes in through the tests, the better the tests become, which sells more tests, which generates more data and so on. Almost no one else in the industry has both parts of the ecosystem.

The whole idea behind the company is on one slide from the earnings call deck.

The next slide completes the story.

The Numbers

Revenue: $348.1M, +36.1% YoY, beating the consensus of about $345.4M.

Gross profit: $222.0M, +43.1% YoY, growing faster than revenue.

Non-GAAP EPS: -$0.13, a $0.07 beat, improved from -$0.24 a year ago.

GAAP net loss: -$125.9M, versus -$68.0M a year ago and -$54.2M last quarter.

Adjusted EBITDA: -$2.8M, versus -$16.2M a year ago, down from +$12.9M in Q4.

Diagnostics revenue: $261.1M, +34.7% YoY, on oncology volume growth of 28%.

Data and Applications revenue: $87.0M, +40.5% YoY, with Insights (data licensing and modeling) up 44.1%.

Guidance: raised to $1.59–1.60B for the year (25% growth), adjusted EBITDA guidance kept at about $65M.

Some numbers evolve in opposite directions, and that’s puzzling. Gross profit grew 43% and revenue grew 36%, so the company became more profitable at the gross level. But the net losses almost doubled. Both are true. I’ll get to that later in the article.

What The Market Didn’t Like

I have recently started looking at this before my analysis. The reason is that it forces me to look into the negative story first and if I feel that it sounds convincing, it can influence the rest of my article and even the long-term thesis evaluation. If you first look at the information and then only at the end to the bearish story, it’s hard to rewrite the full article from a new perspective if you think bears have a valid point. That’s the great thing about investing and the investing process: you can always keep learning.

As far as I can see, there were four things the market didn’t like.

• The sequential revenue decline, from roughly $367M in Q4 to $348M in Q1.

• The negative EBITDA. A year ago, management made a real moment of crossing into positive adjusted EBITDA, and Q1 went back into the red.

• Hereditary growth: the headline says 54%, but Tempus bought Ambry in February 2025, so most of that growth comes from Ambry. Real growth was just 7%.

• What I would call ‘the gap.‘ I mean that there’s a big gap between gross profits up so much and GAAP losses down so much.

Let’s go into them one by one.

Below, for paying Multis: the full breakdown of all four worries, the data business and the upside that nobody's pricing in, my five new Selling Rules, the Quality Score, the valuation, and my Buy-Hold-Sell call.

Subscribe now

Read more

A few hours ago, I launched a poll in the community. The question was which earnings analysis I should do next.

SEA Ltd (SE) was number two. That was great, as Zack had promised the SEA analysis for Friday, but he was early and already sent it this morning. Awesome, Zack!

So, even if Sea was only number two in the poll (you can still vote if you are in the community for the next one you want me to tackle!), the SEA analysis is the first to be published. But, I’ll clear the floor for Zack. I’ll see you back for the Selling Rules, the Quality Score update and the valuation. At the end of the article, you should know if SEA is a BUY.

But to know that, we need top-notch information on the earnings. Take it away, Zack!

Hi Multis

It’s been a tough earnings quarter for multiple Potential Multibaggers stocks in terms of post-earnings price reactions. In the private chat group, I made a tongue-in-cheek comment the day before Sea Limited’s earnings:

I was generally right with the metrics, except for the price reaction:

A 10%+ earnings reaction made me think the sell-off streak might be broken, but a few days later it was reversed, and we’re pretty much back where we started.

It’s another testament to how random stock price reactions to earnings can be. There’s a lot of noise and just a bit of signal. To separate both better, we have to analyze the earnings. Let’s jump into the results to find out.

Sea Limited Group Results

First, we look at the results of the whole group.

Source: Sea Limited

Revenue came in at $7.1B, growing 47% year-over-year, beating estimates by a whopping 11%. I already want to stress that right now, at the stage Sea is at, this is the most important metric to look at. The company is investing heavily and this 47% revenue growth shows that the investments are working. The majority of revenue growth came from Shopee and Monee; we’ll dive into each segment in detail later, of course.

Gross profit was $3.1B, growing 40.7% year-over-year.

Gross margins were 44% versus 46% the previous year. Investments in Shopee continued to pressure margins.

Operating Income was $593M, growing 30% year-over-year.

Adjusted EBITDA increased to $1B, growing 9.3% year-over-year.

For Sea Limited, we primarily use Adjusted EBITDA as the core profitability metric to remove accounting distortions that don’t reflect the true growth of the business, specifically Bookings (more on that in the Garena section).

Single-digit EBITDA growth doesn’t sound like much, but the fact that the company grows its profitability despite the big investments is great.

The largest contributor to Adjusted EBITDA was Garena, which had a standout quarter. You can see that in the graphic below with my annotated year-over-year growth:

Source: Sea Limited

Net Income was $438M, growing 6.7% year-over-year which led to an EPS of $0.70. Again, good to see that despite the big investments, Sea remains solidly profitable.

Additionally, Sea repurchased 1.8M shares for $168.4M this quarter under its $1B share repurchase program.

The limited FY 2026 Guidance was reiterated: GMV growth of 25% year-over-year and adjusted EBITDA no lower than 2025 in absolute dollar terms. Sea has a history of often general, loose guidance, so this is no surprise.

Overall, the top-level business metrics looked great, but the individual segment metrics tell the real story.

Shopee

Gross Merchandise Volume (GMV) was $37B, 30.2% growth year-over-year.

Orders came in at 4B, 29.3% growth year-over-year.

Gross Merchandise Volume (GMV) was $37B, up 30.2% year-over-year, while orders came in at 4B, representing 29.3% year-over-year growth.

Source: Sea Limited

Investment continues to drive higher volume and more purchases:

• Monthly active buyers increased 16% year-over-year

• Monthly average purchase frequency grew 12% year-over-year.

• Fulfillment order volumes grew 25% quarter-over-quarter

Shopee VIP membership continues to boom, reaching over 10M members (up 40% quarter-over-quarter) with program retention averaging 80%. Keep in mind, it had only 1M members just one year ago. That’s 10x growth in twelve months!

VIP is Shopee’s higher-tier membership subscription (think Amazon Prime or Coupang WOW) offering perks such as free shipping with no minimum spend.

Source: Shopee Indonesia

Unsurprisingly, VIP members spend more, with a double-digit uplift of 30% to 40% in some markets. VIP members currently contribute around 20% of GMV across Asia, and the program just expanded to Brazil in April.

On the AI front, innovation continues to drive efficiency and growth for Shopee:

• AI-enhanced search/recommendation algorithms for buyers and AI-generated content tools for sellers led to a 14% improvement in purchase conversion rates year-over-year.

• 80% of customer queries are now handled by AI chatbots, helping reduce cost-per-contact by 30%.

• AI-driven personalization and targeting contributed to the strong ad growth

Moving on, Shopee Revenue grew to $5.1B, 45% year-over year:

Source: Sea Limited

Breaking out the Shopee revenue segments further:

Core Marketplace Revenue grew 60% year-over-year to $3.8B. Marketplace revenue consists primarily of transaction fees and advertising revenue.

Product Revenue grew 50% year-over-year to $612M. These are first-party products that Sea sells itself rather than relying on third-party merchants. It comes at much lower single-digit gross margins (4% this quarter), but management believes this is worth the cost to improve the buyer experience, expand selection, and ensure prices remain competitive.

Value-added Services Revenue was down 5.9% year-over-year to $692M. This drop was due to shipping subsidies as they build out better delivery and fulfillment services. Despite that, we’re seeing those investments pay off. From CEO Forrest Li:

With greater economies of scale, we are seeing lower delivery costs per order for these faster services compared to last year. For example, in Indonesia, our instant delivery service can deliver orders in as little as two hours in urban areas. Order volumes for this service grew over 35% in the first quarter, with cost per order reducing by around 20% year-over-year.

The overall Take Rate increased sequentially to 13.7%. That’s largely due to ads, which grew 80% year-over-year, and to an increase in ad take-rate of more than 90 bps year-over-year.

Source: Sea Limited

Despite that, Adjusted EBITDA was still down 15% year-over year at $223.2M

Shopee’s Adjusted EBITDA margin continues to track lower than in previous years, but as management has reiterated multiple times, the company is in an investment cycle. Management’s medium- to long-term target is to achieve an Adjusted EBITDA margin of 2% to 3%, up from currently 0.6%, as you can see. That said, the longer this goes on, the more critics will point to it as a persistent, structural defensive posture due to rising competition. The rising take rate is a counterargument.

Fred Liu of Hayden Capital had an insightful comment related to this in his most recent investor letter:

For example, China is the most competitive ecommerce market in the world with four large ecommerce companies all competing heavily. Despite this, all the companies are making ~2% EBITDA / GMV – 3x higher than Shopee’s current 0.7%. I think China clearly illustrates that relative competition and absolute market share isn’t necessarily correlated with lower terminal margins.

Source

It seems more than plausible for Shopee to increase its EBITDA margin to its target goal once this investment cycle scales out, despite competitive forces.

Competition

Some new data came out this past quarter on the e-commerce landscape in Southeast Asia. Below is a chart I made aggregating data from a report:

Source: Momentum Works

TikTok Shop is indeed gaining GMV share, some at the expense of Shopee, but also other players, which aren’t depicted here. More importantly, Shopee still gained market share in the biggest market (Indonesia) and retains a dominant share over the region overall:

Source: Momentum Works

Some of this data may elicit some fear regarding competition. Even I had questions, but I also found alternative third-party data depicting TikTok Shop’s GMV decelerating, not accelerating:

Source: X

So what data should we trust? Below is a recent interaction I had with Fred Liu on the topic:

Source: X

That looks better for Shopee, doesn’t it? It was a good reminder of the wide range of conclusions one can draw depending on where data is sourced. It’s on us as investors to aggregate information and make our best educated guess. Third-party data is merely one data point among direct earnings reports. One piece of data should not be the end-all, be-all.

Personally, I see rising Shopee GMV, orders, revenue, and take-rates over the past few quarters. Management also commented on the competitive environment in Q4 2025:

Regarding your question on the competitive landscape, I think what we observed is relatively stable competitive landscapes across most of the market. Yeah, I think that we didn’t observe anything very different from what we see from last quarters.

TikTok Shop has grown in the region but so far has not adversely affected Shopee. In addition, there is no proof that they can’t co-exist. TikTok Shop targets a spontaneous, discovery-led “shoppertainment” audience buying low-cost items like fashion and cosmetics.

Shopee, on the other hand, captures high-intent, utility commerce where buyers proactively search for goods, including consumer electronics and everyday household products. More importantly, they have a comprehensive platform to nurture existing customers and drive repeat purchases, rather than relying on one-time, impulse-driven relationships.

If anything, Shopee is encroaching on TikTok’s territory with its expanding partnerships with YouTube and Meta as confirmed by CEO Forrest Li:

Our content ecosystem continues to grow healthily. In the first quarter, orders from live streaming and short-form video grew more than 50% year-on-year.

These orders accounted for more than 25% of total physical goods orders in Southeast Asia. To further strengthen our content ecosystem, we continue to deepen our content partnerships.

Orders driven by YouTube more than doubled year-on-year. Our collaboration with Meta is scaling well with over 4.5 million affiliates across our market, up nearly 30% quarter-on-quarter. In Indonesia, we have extended our Meta collaboration to enable seamless product promotion and checkout, not just on Facebook, but also on Instagram.

That was Shopee, next we move to the fintech segment of Sea Limited.

Monee

Revenue was $1.2B, up 58% year-over-year.

Adjusted EBITDA was $275M, up 14% year-over-year.

Source: Sea Limited

While topline revenue growth has held strong, Adjusted EBITDA growth continues to slow due to two reasons:

1. Provisions for credit losses are increasing due to expansion of the credit book.

2. Sales and Marketing expenses continue to grow significantly (up 142% year-over-year this quarter).

Similar to MercadoLibre and Nu Holdings, Sea Limited is capitalizing on the opportunity to expand its lending services aggressively to capture market share. Just last quarter, they opened up their SPayLater (buy now, pay later) application to all users, and that momentum is continuing:

Source: Sea Limited

Loans Outstanding grew 71% year-over year to $9.9B

First-time borrowers this quarter were 4.9 million with 38 million active credit users, up more than 35% year-over-year.

Despite continued credit expansion, non-performing loans (NPLs) remained steady at 1.1%. As Kris mentioned in previous articles, NPLs can be artificially held down when a loan book grows this fast; there are simply more “good” new loans, outnumbering the bad, before the most recent cohorts have had time to season or default. In the future, it is completely normal to see NPLs rise slightly.

Sea has also been actively expanding credit to Brazil:

• The Brazil loan book crossed $1B, growing more than 250% year-over-year.

• SPayLater GMV penetration on Shopee Brazil is at 10%, well below mature Asian markets, indicating a massive runway.

• They obtained the SCFI license, which will allow Monee to broaden the scope of financial services offered.

Sea also continues to expand credit use cases to offline settings (Off-Shopee SPayLater). This past quarter, Off-Shopee SPayLater loans in Thailand and Indonesia exceeded 20% of the total portfolio. On-Shopee lending now accounts for less than half of their total fintech business. CFO Tony Hou noted:

This proves that we’re not only able to drive our SPayLater or in general, lending in the Shopee ecosystem, but also we successfully drive this in the off-Shopee ecosystem.

Critics argue that the fintech arm was only successful because it was piggybacking on Shopee’s massive user base. With Off-Shopee expanding past On-Shopee lending, that argument is convincingly refuted. Monee has evolved into a standalone financial solution.

The last point to highlight is the growth of on-book versus off-book loans:

Source: Sea Limited

Taking another look at this graphic, we see that off-book loans have stagnated for a while. In the past two quarters, however, we have seen some growth again, albeit just 10% sequentially. Off-book loans are loans funded without the use of Sea’s balance sheet, allowing it to obtain revenue from financial fees while offloading the risk to other institutions. This is a part of a bigger thesis I have regarding Monee, but more on that at the end of this article.

Moving on to the final segment of Sea Limited, its gaming division, Garena.

Garena

Q1 2026 came in stronger than expected, marking the best-performing quarter since the pandemic-fueled peak of 2021:

Bookings for the quarter came in at $931M, 20% growth year-over-year.

As a reminder, we choose to emphasize bookings over GAAP revenue due to revenue recognition rules. When Garena users buy in-game credits, it isn’t recognized as revenue until it is actually spent in-game.

While a 20% growth number doesn’t seem mind-blowing on the surface, zooming out reveals it was built on a very high base. Q1 2025 had a massive 51% year-over-year growth thanks to the Naruto Shippuden campaign. Stacking 20% growth on top of that is incredibly impressive at the maturity Garena is.

This quarter’s boost can be attributed to Free Fire’s collaboration with Jujutsu Kaisen, which generated over 700 million official content views.

Source: Free Fire

Unsurprisingly, the bookings boost lifted profitability, monetization, and engagement across the board:

Source: Sea Limited

Adjusted EBITDA came in at $574M, 25% growth year-over-year with a strong 61.6% margin.

Source: Sea Limited

Quarterly Active Users were 666 million.

Paying Active Users were 73 million.

Average Revenue (Booking) Per Paying User was $12.8

Quarterly Paying User Ratio increased significantly to 10.9% which means a higher percentage of active players are converting to paying users.

To close out this section, let’s take a look at the quarterly bookings chart again:

With this link, you get a 15% discount for Fiscal AI, so you can make these great charts yourself. Or use their top-notch screener. Or listen to the earnings calls, or… So much more.

After a slow period in 2023 and 2024, the quarters from 2025 onward prove that Garena can generate bookings acceleration almost on demand through high-impact partnerships with global IPs (Naruto, Squid Game, Jujutsu Kaisen). How sustainable this is in the long term remains to be seen, but I am very optimistic after seeing how significant these partnerships have been in driving bookings lately.

The Future: What Sea Limited is Building

This quarter was fantastic, with all business segments showing strong momentum. Let’s zoom out for a moment to gain perspective on what management is building for the long term.

As we saw in the Momentum Works report, Shopee is the undisputed #1 e-commerce player by GMV in Southeast Asia. Instead of just chasing raw user growth, they are deepening existing relationships through logistics investments and the VIP program. 10M VIP members against a backdrop of roughly 400M active buyers means penetration is only about 2.5%. Compare that to Amazon Prime or Coupang WOW, which exceed 50% penetration, and you see the massive runway ahead.

Additionally, continual improvement of customer experience leads to an increase in mindshare among buyers; they no longer consider comparing prices with other competing platforms, even if there may potentially be lower prices elsewhere.

CFO Tony Hou from the 2025 Q3 earnings call:

And Shopee remains consistently regarded as the e-commerce platform offering the most price-competitive products in both our Asian markets and Brazil, based on Qualtrics surveys.

I live in the US and I can resonate with this feeling as a frequent shopper of both Amazon and Costco.

When I go to purchase an item at either place, I can rest assured that I will most likely have paid close to the best possible price. Even if it isn’t, it’s not worth the effort to look elsewhere to save a bit more; the customer experience for members (e.g. perks, shipping, returns) more than makes up for it.

The large, high-engaged Shopee base serves as a zero-acquisition-cost funnel for Monee. When a Shopee buyer checks out, they are offered SPayLater:

Source: YouTube

Sea underwrites these loans using proprietary internal transaction histories, a massive data advantage over traditional banks in underbanked regions. Once a user builds a clean repayment history, Monee cross-sells them higher-value, off-Shopee financial products.

Combining these two businesses together creates the flywheel effect:

Source: Google Gemini

In the Monee section, I also mentioned off-book loans offering additional optionality. The shift toward an off-book model is the ultimate scaling mechanism. Actively utilizing partner capital allows Monee to act as a lending marketplace, taking high-margin origination fees while avoiding balance-sheet risk. This can unlock immense operating leverage.

What’s even better is that this model can be replicated across other developing countries that are underpenetrated in both e-commerce and financial services. We see that happening within Latin America (Brazil):

Brazil was the fastest-growing market in the past two quarters, and the investment is just getting started. In Q1 2026, SEA opened up three new fulfillment centers, bringing the total to five. The VIP program just launched this past April, and the loan book grew 250% year-over-year.

If you can’t tell already, Sea is successfully running the same playbook from Asia now in Brazil, and there’s nothing stopping them from applying it to other developing countries.

Here is the playbook summarized below.

Phase 1: Launch Shopee in the target country with aggressive, cash-burning market-share play, e.g., free shipping vouchers, discounts, low take rates. Subsidize with cash from other countries and Garena.

Phase 2: Once an active user base is established, transition from third-party to building out a proprietary logistics network (SPX Express) to drive down per-order delivery costs, speed up fulfillment, and establish a logistics moat.

Phase 3: Leverage Shopee customers to cross-sell low-risk fintech services such as a digital wallet for payments, but once credit history is established, start opening applications for SPayLater loans.

Phase 4: Lock-in engaged customers with higher-value services such as Shopee VIP, Off-Shopee loans, insurance, and wealth management services. Profit from long-term lifetime value from customers.

Taking a step back, we can clearly see Sea Limited executing this playbook masterfully. To realize it, you just have to look past the short-term price action and see the forest (or I guess Forrest?) from the trees. Shopee is solidifying its moat through a high-barrier logistics network and a sticky membership ecosystem; Monee is aggressively expanding its user base into a legitimate regional financial powerhouse; and Garena continues to serve as the ultimate cash cow, efficiently funding the entire enterprise.

So was the muted earnings reaction deserved? I think after reading this, you probably know the answer. But I’ll let Kris officially answer that with his upcoming parts of this article.

If you enjoyed this analysis and want to read more of what I have to say, you can find me on X @ZackStacksKap. Thanks for reading! Back to Kris.

Thanks, Zack. Fantastic work as always.

So that’s the quarter, and it was a strong one across all three segments. The question Zack kept pointing at is the one that matters: was the muted price reaction deserved, and is SEA a buy here?

That’s my part, and it’s for paying Multis.

Below the paywall:

The Selling Rules for SEA, the lines that would make me trim or sell.

The full Quality Score update.

My full valuation and where SEA lands in my QPI (Quality-Price Index), and what I’m doing with my own position in the Forever portfolio.

Subscribe now

Read more

Last Wednesday after the close, Nu Holdings (NU) reported its Q1 2026 results. The stock dropped more than 6% the next day but it had already dropped before and that means it’s now down 35% since its high in January of this year, just a few months ago.

Within minutes, the bear narrative was circling on X: provisions are surging, NIMs (Net interest margins) are compressing, asset quality is deteriorating, and the growth model hit its limits. Wow, that was fast! But was it correct?

If you are a shareholder, no matter how well you know the company, this will always sound convincing to some extent, even if it’s pure noise. But what is signal and what is noise? The company should help us, shareholders, to make that distinction and fortunately, Nu Holdings is generous in that regard.

This was one of the most informative earnings releases I’ve seen in a long time. Management analyzed almost every line item that the bears pointed out. I want to thank Nu’s management for making my job a ton easier.

It’s funny to see how two Latin American companies (NU & MELI) beat a large majority of American companies in providing detailed information to their shareholders.

Let me walk through what the market hated, what the data actually shows, and then everything else that matters for the next ten years of this company.

The Numbers

Let’s first look at the numbers. (FXN= FX-neutral, the business growth without currency impact.)

• Revenue: $5.32B, +53% YoY, beating the consensus of $5.06B by $260M.

• EPS (GAAP): $0.18, missing consensus by $0.01.

• Net income: $871M, +41% YoY FXN, down 5% QoQ from $895M.

• Adjusted net income: $937M, beating the consensus of $928M.

• Gross profit: $1.88B, +27% YoY FXN, down 7% QoQ, mostly because of elevated CLA (credit loss allowance), more on that later.

• Gross profit margin: 43.8%, missed by 8.5 pp

• Customers: 135M, +14% YoY. 4 million net adds in the quarter.

• ROE (return on equity): 29%. Adjusted 31%.

• Total credit portfolio: $37.2B, +40% YoY FXN, +7% QoQ.

• Deposits: $42.4B, +22% YoY FXN.

• NIM: 21.1% (from 19.5% in Q4 2025).

• Risk-Adjusted NIM (net interest margin): 9.5% (from 10.5% in Q4 2025).

First, a quick note on the EPS miss. Nu introduced a new P&L (profit & loss) framework in Q4 2025. Many have not yet used this new model. On top of that, the consensus is a mix of IFRS and adjusted numbers and you will sometimes even see a miss on revenue because of that confusion. So, I wouldn’t worry about the EPS “miss” too much. IR officer Guilherme Souto opened the call by pointing this out immediately:

We are aware that consensus estimates across the sell side reflect the mix of IFRS and managerial frameworks, and we encourage everyone to use the reconciliation report as a reference point for aligning models going forward.

I love it when a CFO points analysts to this directly. I think the estimates will be more accurate going forward.

What The Market Didn’t Like

Let’s start here, as I always try to understand the negativity to assess whether it’s warranted or not. I think three numbers and one extra element drove the (limited) selloff after the earnings and the bigger 35% drop since the all-time high.

The first is the 15-90 day NPL (Non-Performing Loan) ratio, which jumped from 4.11% to 5.00% QoQ. An 89-basis-point move in just a quarter is the largest in over a year. Read in isolation, that’s an ugly headline.

The second is the CLA (Credit Loss Allowance, the money Nu sets aside on its balance sheet for expected losses on its credit book), which jumped to $1.79B, +33% QoQ. That doesn’t look great either, does it?

The third one is risk-adjusted NIM. NIM, or net interest margin, is what a bank earns on its lending after subtracting its funding costs. Risk-adjusted NIM goes one step further by subtracting credit losses. That makes it a better measure of the credit book’s actual profitability. It dropped from 10.5% in Q4 2025 to 9.5% in Q1 2026.

Some, obviously people who want to feel sophisticated, point out the macro economy in Brazil and point out that 44% of Brazilians defaulted on their credit, so the loan market should absolutely be avoided at all cost.

Out of those four factors, you can build any bear case you want and that’s what I see happening quite widely.

And probably, if you are a shareholder, at least one of them, probably multiple, or all of them sound convincing. But none of them mean what you think. That’s quite a statement, I know. But if you have read this article, please let me know if your worries have decreased or increased.

Let me walk through the first three factors one by one. The macro thing comes later.

The NPL Bridge

For the first time, as far as I know, Nu broke down the 89 bp NPL jump into its drivers.

So, about two-thirds of the deterioration is simple seasonality. Every Q1 in Brazil looks like this. Just compare the last three Q1 numbers: 5.1% in 2024, 4.8% in 2025, 5.0% in 2026. You see the same pattern every year.

The 90+ NPL ratio, which is what turns into write-offs and losses, came down slowly from 6.6% in Q4 2025 to 6.5% in Q1 2026. That’s well below the 7.0% peak from Q3 2024.

So the headline that scared everyone is doing what it always does in Q1. But this time, because of the headline consensus misses (which came from quite some analysts using an old model in their estimates). That’s why the NPLs were singled out more, ignoring the seasonality.

CFO Guilherme Marques do Lago put it well. But before I quote him, I want to explain what NPVs are. It stands for Net Present Value, the value of all future profits from the loan book after costs.

The goal of our credit operations, it’s not to minimize NPLs at a point in time. Instead, it is to optimize for resilient NPVs. NPLs only capture the cost side of the equation. They say nothing about the revenues we generated from the customers who perform.

This matters and it’s also how I think. Banks that obsess over keeping NPLs low only lend to the safest customers, at the lowest spreads, with mediocre returns as a result. Nu price risk into the spread and accept that some customers won’t pay, and that makes it grow into much wider customer bases at much higher returns. The first is the incumbent model. Nu does the second.

The CLA Surge

The same logic can be applied to provisions. Just as a reminder. Credit Loss Allowance is the money on the balance sheet for expected losses on the credit book. It went from $5.30B at the end of 2025 to $6.10B at the end of Q1, an $800M increase. Management gave us the explanation again.

Growth and seasonality together explain $690M of the $800M, or 86%. Credit quality contributed almost nothing.

There’s an accounting rule that I also mentioned in the context of Mercado Libre. When a bank originates a new loan, it has to set aside the entire expected lifetime loss on that loan immediately, before earning a single dollar of interest. So if you’re growing fast, you front-load losses into today’s earnings while the offsetting interest income shows up over the coming quarters. That makes things look much worse than they actually are. That’s the ‘growth’ component, by far the biggest contributor.

Nu’s total exposure grew 44% YoY FXN to $70.7B. Total exposure is the credit card limits Nu has given its customers plus the money actually borrowed, so it’s much bigger than the drawn credit portfolio. Every dollar of that exposure gets a provisioning charge upfront, whether the customer ever uses the limit or not.

Look at the mix on the right side. In Q1’26, credit cards were 73% of the new exposure, and unsecured lending was 25%. That’s 98% combined. A year ago, credit cards were 60% and unsecured 28%, or 88% combined. Secured lending declined from 12% to just 2% over the last year, due to regulatory changes to the FGTS product at the end of 2025. FGTS (Fundo de Garantia do Tempo de Serviço) is a Brazilian employment-savings fund used to secure consumer loans.

Cards and unsecured loans have a higher expected loss than secured loans. They also yield much more, but the provision rises as more credit cards and unsecured loans are issued.

In short, the same dollar in loans now has a higher yield but also higher provisions than six months ago. That’s not because customers are of lower quality but because the yield is higher.

The NIM Drop

Now the risk-adjusted NIM. Yes, 9.5% is lower than 10.5%. But the overall NIM moved the other way. NIM, that’s before credit losses, went from 19.5% in Q4 to 21.1% in Q1, up 160 bps. As the risk-adjusted NIM went -100 bps, as we just saw, that means that the gap widened by 260 bps in one quarter. That sounds ugly, right?

But it’s simple. That gap is the same accounting requirement at work. The yield on the new originations shows up in NIM but the expected losses on those same originations show up in CLA (Credit Loss Allowance) right away. This is a way you can look at: the real losses will arrive over the coming quarters, but they are already in the books. At the same time, interest income continues coming in every quarter. By Q3-Q4 of this year, the two lines should meet back up. CFO Guilherme Marques do Lago:

We expect risk-adjusted NIM to move back towards the level we operated at during the second half of 2025 as the dynamics of first quarter normalize over the coming quarters.

So the compression is probably just a quarter of seasonality, not a structural break.

The Cohort Economics

Even if seasonality explains the Q1 rise, the loss rates Nu shows for some cohorts appear high in absolute terms. 22% in losses on unsecured lending and a whopping 33% in losses on Mexican credit cards. Aren’t those scary numbers?

Here, Nu’s most interesting new disclosure comes in. They gave us the full unit economics of the three biggest unsecured franchises, with every component broken out.

• Brazil unsecured lending: 62.2% in interest revenues, -12.2% in funding costs, -22% in losses = 28% net lending margin. That leads to a cohort ROA (Return on Assets) of 15.8% for a loan book with a 7-month duration. Do you already see what is happening here?

• Brazil credit card: 28.3% net interest revenue + 8.5% non-interest revenue, -0.3% funding, -16.6% losses = 19.9% net margin. Cohort ROA of 8.9% on a 3-month duration.

• Mexico credit card: 46.1% interest revenue + 21.5% non-interest revenue, -6.3% funding, -33.3% losses = 28% net margin. Cohort ROA of 14.1% on a 2-month duration.

For context: global bank credit card portfolios typically earn ROAs of 3-5%. Of course, that’s a different context. The best Brazilian incumbents earn 8-14% on cards, with much lower loss rates because they only lend to the top tier of customers. Nu is in a completely different league, and three things make that possible:

• Brazil’s policy rate (the Selic) is at 14.5%, which means lending spreads are structurally enormous.

• Nu’s cost to serve is $1.00 per active customer per month, an order of magnitude below that of the incumbent banks.

• A short-duration credit book that lets management react fast when something changes.

Just look at those durations again. Brazil credit card: 3 months. Brazil unsecured: 7 months. Mexico credit card: 2 months. Traditional Latin American banks operate with durations measured in years. Mortgage-heavy banks are in the high single digits in duration.

CFO Guilherme Marques:

We are not a loan book lender waiting quarters and quarters to see the impact of a credit policy change. We see it in days, and we act on it immediately.

Nu uses duration as a risk-management instrument. That’s why they can underwrite cohorts with high loss rates and still earn returns that don’t exist anywhere else in banking. You saw that in the numbers. In other words, that’s not risk-taking, it’s pricing risk at the correct premium.

Float

The credit story took all the attention this quarter, but something else caught my eye.

Look at this graph.

As you can see, float made up 41%. Float is the interest Nu earns on customer deposits that aren’t lent out as loans. With $42.4B in deposits and only $22.8B in net credit portfolio, that’s a loan-to-deposit ratio of 53.8%. So, for every $1 of deposits Nu has, only about 54 cents is currently lent out. Or put differently: Nu has a big pile of unloaned deposits sitting on the balance sheet, earning the local risk-free rate. Deposits compound faster than loans, which means that float income grows with them.

In Q1, the contribution of credit also dropped because CLA was higher, as we saw earlier. That part is seasonal and will largely reverse in the rest of the year. If you take a step back, you see that gross profit is becoming more diversified across credit, fees, and float. And that’s a good thing. It’s making Nu stronger.

This is the final proof. If you thought that Nu was a one-trick credit card pony, this quarter shows you are wrong. Fees and float together now make up two-thirds of gross profit.

Two Tailwinds In Brazil

In the Q&A, CFO Guilherme Marques do Lago mentioned two structural Brazilian tailwinds.

The first is the new Brazilian income tax exemption for monthly earnings up to BRL 5,000 (about $1,000). There are even partial reductions up to BRL 7,400 (about $1,500). That means that most ordinary people with a job will profit and those people are exactly More income means more loaning capacity and lower expected loss rates on the cohorts already on the book.

The second is about what I mentioned as a part of the bear case: macro. Yes, credit default numbers in Brazil are simply insane if you are used to the West. That 44% default number is correct.

But the problem is that bears make it sound as if this came out of the blue and now it’s an incredible, unprecedented time in Brazil, which is simply not the case. The default rate has been around 40% for a very long time already and yes, it went up slightly. Well, slightly, from give or take 40% to 44% is still a lot, as it’s up 10% but it’s not as if this is a new situation in Brazil. Don’t forget that this is the environment has operated in for years and it has thrived by pricing loans very correctly. We’ll see later in this article why they have even become better at this. Hint: AI.

On top of that, the government also knows this and it launched something they hope can help.

But back to the government initiative. Desenrola 2.0, the second round of a Brazilian federal debt renegotiation program. The first round (in 2023-2024) helped about 15 million Brazilians with defaulted debts renegotiate them at steep discounts, clear their names from the credit bureau blacklist, and rejoin the credit system.

Desenrola 2.0 is expected to launch in Q2 and Q3 2026 and should bring millions more newly bankable consumers into the system. Nu, as the dominant brand in Brazil, is well positioned to capture them as new customers. CFO Guilherme Marques do Lago called the program “either neutral or positive” for Nu.

Efficiency

Now to the efficiency ratio. The efficiency ratio measures how much a bank spends in operating costs to earn one dollar of revenue. The lower, the better. You may know this from your favorite insurance company.

A traditional Brazilian incumbent typically runs at 40-50%, meaning they spend 40-50 cents to earn one dollar. Nu came in at 17.6% this quarter. That’s about 18 cents to earn one dollar, which is less than half of an incumbent. Pretty impressive.

But it’s even better. Nu makes investments in three things that other banks are not doing: bringing employees back to the office, expanding internationally (to the US), and building out its AI infrastructure. If you strip those out, the “core” efficiency ratio is 16.6%. With just $16.6, the company can earn $100. That’s great. A year ago, the efficiency ratio was 21.4%, so this is a big improvement. The operating leverage is doing its work.

NU CFO Guilherme Lago told us not to extrapolate, though:

The first quarter came in better than expected for 2 reasons working together.

Number one, revenues accelerated faster than we anticipated, driven by both ARPAC outperformance and continued portfolio growth.

Second, OpEx came in below plan...

Roughly 1/3 reflects structural efficiency gains that are durable and compounding, mainly AI-driven improvements in operations and collections, software platform consolidation and hiring discipline. Now the remaining 2/3 reflect timing items that will normalize in the next quarters.

He gave us the full-year guidance: roughly 20% efficiency for 2026, in line with where 2025 ended, but including all the strategic investments. The core efficiency ratio (the trend excluding those investments) continues to go down, which is great to see.

ARPAC & Cohorts: No Plateau In Sight

Monthly ARPAC (Average Revenue Per Active Customer, remember that, as I will use it multiple times) for Q1 2026 was $15.9, +23% YoY FXN, up from $15.0 in Q4.

That’s the average across all customers. Mature customers, those who’ve been with Nu for 96 months (8 years), generate $30.8 in monthly ARPAC, nearly double the company average. And it keeps going up year after year.

Look at the cohort progression. It looks linear and so far, there is no plateau in sight.

Now think of the customer base. Nu has added more than 65 million customers over the last three to four years.

Just insane. But that entire cohort is still in the stage of the low ARPAC in the graph. If they do ever remotely do as well as the older cohorts, the revenue growth from existing customers alone will be huge. And that’s before Nu adds new customers, launches new products, or starts in another country.

Monthly costs to serve customers are around $1.00 per active customer. The ARPAC of $15.9 against $1.00 of cost, that’s phenomenal operational leverage.

The cost to serve is up 19% YoY, which is faster than historically, but still extremely low. ARPAC is up 23%, so the gap is still broadening. I expect the cost to serve to increase if the company focuses more on small- and medium-sized businesses and the wealthier segment of the population. But of course, in that case, the ARPAC should also rise.

Mexico: The First Quarter Of Profitability

It was great to hear that Mexico has already seen its first profit in this quarter. And not on some adjusted or managerial measure, but on an IFRS basis (International Financial Reporting Standards, comparable to GAAP).

That was even faster than management had foreseen. CEO David Velez:

In Mexico, we have moved from a $30 million quarterly loss to our first quarter of IFRS profitability, a milestone that arrived ahead of our own internal plan.

And again, management gave us an interesting slide full of information.

The number of customers rose from 2 million to 15 million over four years, while ARPAC nearly doubled from $6.5 to $12.4. The efficiency ratio dove from 120% to 42%. That’s incredibly strong and another testament to how well the company is doing. And as we already saw, the cohort ROA on credit cards in Mexico is 14.1% on a duration of just 2 months.

Mexico is no longer the experiment. Nu is already the third-largest financial institution in Mexico by customer base. The playbook that management applied in Brazil is working in Mexico.

The Mexican population is around 130 million, so there is still a lot of room to grow from the 15 million customers Nu has now. If Mexico grew to the same size as Brazil, you could expect more than 80 million customers. I’m not saying they will do that but that’s the potential. And management is not shy to show the opportunity.

Nu’s gross profit market share in Mexico is below 1%, in a $43B annual pool that’s expected to grow by 15% per year. Compare that to Brazil, where Nu has a 7% share of a $100B pool growing at 8%. The Brazilian pool is bigger and Nu’s share is much bigger, but the Mexican pool is growing almost twice as fast and Nu’s share is tiny. Over the next decade, Mexico could very well compound faster in absolute terms than Brazil.

Colombia is the third country where Nu is scaling. The customer base is approaching 5 million now and the country is still in investment mode, just like Mexico was from 2020 to 2025. Nu is now intentionally operating at a loss there to build out the brand. It’s that same playbook again

The US Opportunity

In previous articles, I already mentioned multiple times that Nu received conditional approval from the OCC (the Office of the Comptroller of the Currency, the US national bank regulator) for a national bank charter in January. Management has confirmed the seriousness of the move but has not shared the go-to-market strategy or any other details so far. But co-founder Cristina Junqueira, Chief Growth Officer, has relocated to the US and is now the CEO of NU there.

Again, management wanted to be very clear about this expansion. For competitive reasons, it doesn’t want to share its approach, which I think is fair, but it was clear to shareholders what this would cost.

CEO David Velez:

To be precise, the maximum OpEx headwind we expect from US investment in each of 2026 and 2027 is less than 100 basis points on our consolidated efficiency ratio. And this is inside the 20% efficiency ratio level Lago mentioned before. For a company at our scale, that is quite affordable.

Beyond that, any additional investment is explicitly contingent on clear evidence of product market fit and a credible path to profitable scalability.

This is how companies should take bets. The downside is capped and restricted in time. The upside is asymmetrically high if Nu finds product-market fit. If they don’t, the worst case is less than 100 bps of drag on the efficiency ratio for two years, and the core business doesn’t feel it. But if they succeed, Nu enters a whole new era in the world’s largest retail banking market.

Velez has been calling these expansions “call options” since Mexico was the new market. The Mexico call option has already turned out to be worth several billion dollars. Colombia is still vesting but looking promising. The US is a much harder market with stronger digital competition, but with that asymmetry comes huge potential upside.

There was an interesting question on the earnings call in this context. The question was why Brazilian investors are more constructive on the Nu story, including the US thesis, while foreign investors are very skeptical. It also hit home because when I talk about this potential expansion with Americans, they are either very enthusiastic (a minority) or very skeptical. CEO David Velez:

Every time we’ve launched a new country, the locals have been skeptical. When we launched in Brazil, the locals were very skeptical. When we launched in Mexico, the locals were very skeptical... if you’re a local by definition, it’s very hard for you to reimagine how things can happen differently. You are too consumed by the status quo.

That’s a very interesting answer on several levels. As NU is an innovator, the typical consensus is that they do not stand a chance. “Not here, not in this market.” That’s what most locals think. If you are American, maybe that’s what you think right now. And I understand that for sure. If Nu were to expand to Belgium, I might have the same reaction. It’s also why incumbents in any market consistently underestimate what a new player, with a new model, can do. It’s basically a summary of the Nu story for thirteen years already.

Small And Medium-Sized Enterprises

One analyst asked about SMEs. David Vélez:

I actually think this is probably one of the most underappreciated opportunities we have at Nu. We should be speaking more about it, but we’re not. The reality is we’ve kind of silently have built the largest SME base in Brazil with our 5 million SME customers effectively built with 0 customer acquisition cost.

5 million SME (Small and Medium Enterprise) customers. At zero customer acquisition cost. That’s crazy. Nu built this by cross-selling its small-business products to its existing personal-banking customers, since a large share of Brazil’s working population also runs a small business. Today, there are over 2 million Nubank SME credit cards in circulation, plus a growing range of secured and unsecured credit lines (some of them using Brazilian government guarantee programs).

In a country where small businesses employ around 70% of the workforce, this is a very interesting option.

AI: From Talking To Doing

Every company is talking about AI right now. A lot of it is theater, buzzword bingo. But Nu’s AI commentary was very specific, and we can check the numbers.

The first phase, AI Assistance, is largely complete now. AI tools have been rolled out to roughly 100% of employees. Engineering throughput is up 50% YoY, weekly token consumption is up around 10x since January and testing cycles are 90% faster. I love that level of detail!

The second phase, Workflow Reinvention, is in progress. AI executes the routine work and humans hold the judgment.

The most concrete example is Pix with AI, which already has more than 10 million MAU (Monthly Active Users). Pix is Brazil’s instant payment system, launched by the central bank in 2020. It is now used by virtually every Brazilian, for everything from paying for a coffee to paying rent. Nu’s AI layer on top of Pix lets customers do things like pay a friend by simply describing in plain language what they want to do, instead of typing in account numbers. That may sound small, but features like this can encourage people to use their Nubank app more.

The third phase is the AI-Native Bank, which is still in its early stages. The AI Private Banker gives customers financial insights, payment guidance, credit advice, and debt resolution, and it’s already at 15M+ MAUs.

Nu’s proprietary foundation models, branded nuFormer, are live for credit card decisioning in Brazil and Mexico, and for unsecured lending in Brazil. CEO David Velez:

We’re now able to use real-time AI valuation for every personal loan request, priced and approved individually based on its predictive net present value in under 1 second.

Let’s go back to where we started: why the stock dropped. The unsecured and credit card lending book is growing 53% YoY, with high ROAs, despite the big losses some love to accentuate.

And now we know how Nu does this. For every loan application, the AI runs the math on what the loan should yield, what the customer is likely to lose, and whether the loan’s lifetime profit is positive. In less than a second. So when you hear bearish takes that Nu takes on riskier customers to grow the book, it’s because Nu is pricing risk more precisely than anyone else in the industry.

The AI productivity is also showing up on the cost side. Remember that CFO Guilherme Marques do Lago said that about a third of Q1’s efficiency outperformance came directly from AI: improvements in customer operations, debt collection, software consolidation, and more disciplined hiring.

In other words, NU uses AI in its core operations, which is extremely important for its positioning.

Below the paywall:

My five new Selling Rules for Nu. I didn’t have any yet, so I built them from scratch. One of them puts a hard number on management’s promise that risk-adjusted NIM goes back up by Q4. If they miss it, I want to know.

The full Quality Score update. Including the one metric where Nu just knocked Adyen off the top spot after years. I didn’t expect that.

My valuation. You can’t value Nu like a bank and you can’t value it like a normal growth stock either. So I built something else for it

Read more