Ben’s Portfolio update end of February 2026
Returns and portfolio holdings:
Portfolio Notes 2022 -15.6%* *Jul-Dec, since I started posting my portfolio on Saul’s and fully adopting my version of Saul’s investing approach. 2023 77.8% 2024 31.7% 2025 24.9% 2026 YTD Month Jan -8.2% -8.2% Feb -15.9% -8.3%
These are my current positions:
Feb 2026 Jan 2026 First buy* Nvidia 21.8% 22.7% 5/13/2020 Cloudflare 21.2% 20.8% 11/2/2020 Datadog 12.5% 13.6% 5/13/2020 Snowflake 11.2% 12.1% 2/8/2021 Crowdstrike 9.8% 10.9% 5/13/2020 Axon 9.6% 6.7% 4/2/2024 AppLovin 5.9% 3.2% 11/18/2025 Samsara 3.4% 3.2% 1/8/2024 Zscaler 2.3% 3.0% 3/4/2021 Astera Labs 2.3% 1.9% 11/18/2025 Monday 0.0% 2.0% 9/13/2021
*held through today
Company comments
Cloudflare:
Key insights:
- Blowout quarter, accelerating YoY revenue growth for the third quarter in a row: Q1: 26.5%, Q2: 27.8%, Q3: 30.7%, Q4: 33.6%. New Q1 guide points to ~34% YoY.
- Very strong new FY26 guide: 28.9% YoY. For comparison, their initial FY25 guide was for 25.4% YoY and they ended up growing 29.8% YoY in FY25.
- RPO growth of 48% YoY and cRPO growth acceleration to 33.2% (from 30% last Q and last year), NRR growing to 120% (from 111% last year) and customer growth exploding (total customer count up 40% YoY), the company is primed to continue to grow revenues north of 30% YoY.
- Matthew Prince outlined a fundamental re-platforming of the Internet: Because AI agents generate exponentially more outbound web requests than human users, Cloudflare’s network is capturing a massive tailwind in traffic and compute demand. Management noted that top AI companies are increasingly choosing Cloudflare’s unified stack over traditional hyperscalers because of its neutrality, lack of egress fees, and speed to market.
- Gross margin stayed pressured, but close to their 75% goal, and again for the good reason that they get more and more paid traffic through their network, which carries slightly lower margins.
- Even with that, they demonstrated strong operational leverage as operating expenses increased 28% YoY, compared to revenue growth of 33.6% YoY.
- While they’ll continue to ramp Capex at 12% to 15% of revenue, that didn’t stop them from generating record free cash flow, and for the first time since they started to invest heavily in AI in 2024, I see an inflection where free cash flow is again starting to grow much faster than Capex, showing that they can monetize these AI inference investments.
Cloudflare reported Fiscal Q4 2025 on 02/10/26 and their results were nothing short of a(nother) blowout, taking my expectations and systematically exceeding them across nearly every metric. To put the magnitude of this 4.3% guidance beat (the largest in 16 Q’s) and the revenue acceleration to 33.6% YoY into perspective, analysts at Jefferies described the quarter as “gold-medal-worthy”. On the earnings call, CEO Matthew Prince, called it a “terrific quarter” and highlighted that new Annual Contract Value (ACV) bookings grew nearly 50% YoY. This marks the fastest ACV growth rate the company has delivered since 2022, definitively signaling that Cloudflare has re-entered a hyper-growth trajectory.
The total customer number growth rate was absolutely staggering, and the context provided on the call makes it even more impressive. Adding nearly 37,000 net new customers in a single quarter - driving total customer growth up a record 40% YoY - is a testament to their self-serve funnel and expanding product suite. However, the real narrative update came from the enterprise segment. While adding 289 large customers ($100k+ ARR) was perfectly in line with my 7% QoQ expectation, the massive 55.5% YoY growth in $1M+ customers (now at 269) was the hidden gem. Management revealed they closed the largest annual contract in company history during Q4, a mega-deal averaging $42.5 million per year. This proves that Cloudflare is no longer just a vendor for the enterprise; it is becoming foundational infrastructure for the world’s largest companies. The jump in NRR from 114% in Q2 to 119% in Q3 was not a blip as NRR continued to go up to 120% this Q4, confirming that the land and expand motion - especially with Zero Trust and Workers - is sticking.
The central narrative theme of the Q4 call was the rise of AI and the "Agentic Web”. Matthew Prince outlined a fundamental re-platforming of the Internet, arguing that the industry is transitioning away from traditional, seat-based software licenses toward compute, connectivity, and guardrails for autonomous digital workers:
”Now, the agentic internet is emerging and we can already see its trends. If humans looked at five sites when they were making a decision, agents might look at 5,000. If humans had to fall back on generalized software and interfaces, agents allow for infinite customizability of every software application for every need. If humans follow a common circadian rhythm to work, agents never need to sleep. Agents, in other words, are the ultimate infrastructure multiplier. In turn, they are reshaping the very economics of software. The industry is transitioning from a business model defined by seat licenses to one where the winners are those providing the compute, connectivity and rails and guardrails for these new digital workers at scale. Cloudflare was built for this moment. We are uniquely architected to capture value on both sides of the agentic interactions. That means we win when AI applications are built on Cloudflare Workers, but we also win just from the increased usage of all of our products and agentic internet drives.”
Because AI agents generate exponentially more outbound web requests than human users, Cloudflare’s network is capturing a massive tailwind in traffic and compute demand. Management noted that top AI companies are increasingly choosing Cloudflare’s unified stack over traditional hyperscalers because of its neutrality, lack of egress fees, and speed to market.
From a Go-To-Market (GTM) perspective, the earnings call served as a victory lap for the sales reorganization the company endured over the last couple of years. Global sales productivity increased YoY for the eighth consecutive quarter, surpassing the all-time high set back in 2021, and the sales team achieved its highest quota attainment in four years. Interestingly, CFO Thomas Seifert, noted that “pool of funds” contracts accounted for roughly 20% of Q4 ACV bookings. While this usage-based model aligns perfectly with surging AI traffic, he cautioned that it introduces a new seasonality dynamic - driving higher Q4s and relatively softer Q1s due to variable consumption patterns. This explains why the Q1 guide of $620.5M might look like a modest QoQ bump (1.0%), even as underlying momentum implies the mid-30s YoY growth I modeled out. Looking out further, their initial FY guide for FY26 was for 28.9% YoY revenue growth, which is up from their last year’s initial FY guide for FY25 which was for 25.5% YoY growth, suggesting that the current strength in revenue growth will continue through FY26.
Finally, margins and cash flows highlight the beauty of Cloudflare’s underlying financial architecture. Operating expenses increased 28% YoY, compared to revenue growth of 33.6% YoY, demonstrating strong operational leverage. However operating income margin actually stayed at 14.6%, just as in 4Q24, as a result of continued gross margin pressure, where gross profit only grew 29% YoY. This again, just like last quarter, was the result of paid versus free customer traffic again increasing. So Gross margin dipping slightly to 74.9% (just below their 75-77% target) was directly attributed to the volume of AI and paid traffic, leading to the highest allocation of network expenses to COGS in the company’s history (we talked about this in my last recap of Cloudflare). Yet, this didn’t hurt their bottom line. With an operating margin holding steady at 14.6% from last Q4 and a record Free Cash Flow margin of 16.2% ($99.4M in Q4). My long-term thesis regarding TTM FCF (currently $260M) catching up to (and eventually overtaking) TTM Capex (currently $316M) is well-supported by management’s guidance that 2026 Network CapEx will remain highly disciplined at 12% to 15% of revenue. Just look how one of my favorite profitability plots for Cloudflare is starting to show another turn-around:
Combined with RPO growing a massive 48% YoY, and the fastest sequential RPO growth (16.5%) since the 2020 saga, Cloudflare has proven it can simultaneously scale an AI-native global network, land gigantic $40M+ enterprise deals, AND print record cash flow.
Overview of how Cloudflare performed versus my prior expectations:
- Revenue expectation: $604M (7.4% QoQ, 31.3% YoY), implying a 2.5% beat.
→ $614.5M (9.3% QoQ, 33.6% YoY), a 4.3% beat and the strongest beat in 16(!) consecutive quarters with an average beat of 2.0% in those four years. - Q1 new revenue guide: $619M (2.5% QoQ, 29% YoY) which I would interpret as $631M (4.5% QoQ, 31.7% YoY) expecting YoY growth rate to slightly accelerate.
→ $620.5M (1.0% QoQ, 29.5% YoY), which I now interpret as $642M (4.5% QoQ, 34.0% YoY), implying a 3.5% beat. - I would like to see NRR at 119%.
→ NRR grew to 120%, demonstrating that the Q2 (114%) jump to Q3 (119%) was not a blip. - I would like to see total customer growth around 7% QoQ (~20700 net adds, I know, kind of insane).
→ total customers grew an incredible 12.5% QoQ, with 36914 net adds. These are five(!) times as many adds as they typically had before 2Q24. It’s hard to parse that total customer count grew a record 40% YoY! That is the highest customer growth number I have on record since 2018, and that at a large base of now 332,466 customers. - I would like to see large customer growth around 7% QoQ (~281 net adds).
→ large customers (100k+ ARR) grew 7.2% QoQ, adding 289 and now reaching 4298 large customers. We also got an update on $1M+ customers, of which they now have 269, growing 55.5% YoY. - I would like to see RPO grow around 12.5% QoQ to $2.41b (43% YoY).
→ RPO grew 16.5% QoQ to $2.50b (48% YoY, which is up from 42.6% last quarter). - I would like to see cRPO grow around 12.5% QoQ to $1.54b (30.7% YoY).
→ cRPO grew 14.6% QoQ to $1.57b (33.2% YoY, which is up from 30.3% last quarter). - I would like to see Gross Margin greater or equal to 75%.
→ Gross Margin was 74.9%. “paid versus free customer traffic again increased significantly both year-over-year and quarter-to-quarter, resulting in the highest allocation of network expenses to cost of goods sold versus sales and marketing ever. The underlying economics of our network, driven by its inherent scalability and efficiency remained unchanged.” - I would like to see operating income around $95M (15.7% margin).
→ Operating income was $89.6M, a 14.6% margin just as last Q4. - I would like to see a FCF margin around 13%.
→ FCF margin reached a company record of 16.2%, up from 10.4% last Q4 and another datapoint that shows TTM FCF again starts going up while TTM Capex stays almost constant. If this trends stays the next milestone will be that TTM FCF (currently $260M) will start to exceed TTM Capex (currently $316M). - Thoughts from previous quarter: Cloudflare’s 3Q25 earnings recap.
Datadog:
Key insights:
- Again, blowout quarter, accelerating YoY revenue growth for the third quarter in a row: Q1: 24.6%, Q2: 28.1%, Q3: 28.4%, Q4: 29.2%. New Q1 guide points to ~30% YoY.
- Strong new FY26 guide: 19.6% YoY. For comparison, their initial FY25 guide was for 19.0% YoY and they ended up growing 27.7% YoY in FY25. The new guide also excludes ANY revenue from their largest AI native customer OpenAI, which is thought to spend around $150M-$300M annually with Datadog. This shows that even if some of the AI native customers will churn in the coming year, Datadog should be able to maintain close to 30% YoY revenue growth.
- Revenue growth acceleration due to their non-AI-native customers was even stronger, accelerating to 23% YoY and up from 20% YoY in Q3 proving that “cloud cost optimization” is in the past and multi-year enterprise platform commitments have returned.
- Large customer growth was very impressive: The $100k+ cohort accelerated to 19.4% YoY, up from 16.3% in Q3 and 13.6% in Q2. The $1M+ cohort accelerated to 30.5% YoY, up from 16.7% last Q4. The corresponding multi-product users also expanded meaningfully, even by 2% in the 6+ and 8+ cohorts.
- As a result, NRR stayed at 120%, RPO growth stayed above 50% YoY, with cRPO growth at about 40% YoY, meaningfully above current YoY revenue growth and Billings accelerated to 33.3% YoY, from 29.6% in Q3.
- With their Bits AI agent (observability AI agent) reaching general availability in this quarter, Datadog is capturing the tailwind of customers’ shift in artificial intelligence from the experimental phase to full production, successfully monetizing the complexity that generative AI introduces into application development.
- After four quarters of front-loaded investments, Datadog has now achieved the operational crossover from where revenues again grow faster than operational expenses and they delivered strong profitability margins in Q4: FCF margin was 30.5%, operating margin was 24.1% and net margin was 28.2%.
- Finally, whether Datadog will be disrupted by AI or not remains to be seen, but all forward looking numbers point to the exact opposite: That AI is currently a huge tailwind for them and while there is no doubt that some customers will churn (management called this volatility in the AI native cohort), the forward looking numbers they reported suggest that the underlying tailwind will be more than compensating for that.
Datadog reported Fiscal Q4 2025 on 02/10/26 and again delivered a blowout quarter that didn’t just meet my expectations but meaningfully exceeded them on several fronts. Delivering $953.2M in revenue (a 29.2% YoY increase), the company added $67.5M in raw sequential revenue - exceeding my expected $63M. More importantly, the narrative on the earnings call confirmed that this reacceleration is durable. Management highlighted an “inflection” in their broad-based business outside of the AI-native group, noting that core business revenue growth accelerated to 23% YoY (up from 20% in Q3). Paired with Q1 2026 guidance of $956M - which, I interpret with a typical beat, sets up a target of $994M and slightly above 30% YoY growth - Datadog has demonstrated that the era of “cloud cost optimization” is in the past and multi-year enterprise platform commitments have returned.
Perhaps the most positive data point is the significant outperformance in the large customer cohorts. I was looking for ~120 new $100k+ customers and ~508 total $1M+ customers. Instead, Datadog added 250 new $100k+ customers and printed a staggering 603 total customers in the $1M+ cohort (up 30.5% YoY, after being up 114% in FY21, 47% in FY22, 25% in FY23 and 17% in FY24). This enterprise momentum was at the center of management’s commentary, where CEO Olivier Pomel announced a record $1.63 billion in bookings (up 37% YoY). Datadog signed 18 deals over $10 million in Total Contract Value (TCV), including two deals over $100 million and an eight-figure land with a leading AI model company (would a leading AI model company sign a new 8-figure deal if they could “vibe code”, or if it would make sense for them to put significant resources into trying to replicate what Datadog does any time soon?) This enterprise traction directly fueled RPO ($3.46B, +52.4% YoY) and Billings ($1.21B, +33.3% YoY), proving that mission-critical platform consolidation is happening all the way to Fortune 500 companies (where Datadog now has 48% penetration).
The cohort expansion metrics, particularly the amazing YoY growth acceleration in the 6+ (38.3% YoY growth), 8+ (63.5% YoY growth), and newly disclosed 10+ (96.2% YoY growth) product categories, perfectly align with the primary go-to-market narrative from the earnings call: vendor consolidation. Customers are standardizing on Datadog as a unified platform rather than cobbling together open-source or point solutions. Pomel cited several specific examples, such as a European data company consolidating seven different tools to adopt nine Datadog products in an almost seven-figure deal. This platform stickiness (evidenced by the NRR stabilizing at ~120%) has accelerated Datadog’s core products: Infrastructure Monitoring exceeded $1.6 billion in ARR, Log Management crossed $1 billion, and the APM (application performance monitoring) and DEM (digital experience monitoring) suite also crossed $1 billion in ARR, with core APM accelerating into the mid-30% range as the company’s fastest-growing core pillar.
A major theme from the analyst Q&A was the shift in artificial intelligence from the experimental phase to full production, which demands observability. Datadog is capturing this tailwind from both AI-native and traditional enterprises. The company now has about 650 AI-native customers, with 19 of them already spending over $1 million annually and including 14 of the top 20 AI companies. Furthermore, over 5,500 customers are now utilizing Datadog’s AI integrations. With the Bits AI SRE agent reaching General Availability and already driving significant time savings in root-cause analysis, Datadog is succesfully monetizing the complexity that generative AI introduces into application development.
Finally, OpEx, now for the first time in 5 quarters growing slower than revenue (28.9% vs. 29.2%) is the final piece of the puzzle. After four quarters of front-loaded investments, Datadog has achieved the operational crossover I was looking for. The company generated $291 million in free cash flow (a strong 30.5% margin), together with the 24.1% operating margin and 28.2% net margin. CFO David Obstler maintained Datadog’s typical conservative posture for the FY2026 outlook (their initial FY25 guide was for 19% YoY growth while the just released initial FY26 guide was for 19.6% YoY growth), especially applying a “very conservative assumption” to their largest customer’s consumption modeling, but the underlying math is clear: With revenue growth accelerating, enterprise mega-deals signing at a record clip, and OpEx growth naturally decelerating past its Q2 2025 peak, Datadog is fundamentally primed to deliver significant and durable operational leverage and cash generation as it scales past a $4 billion run rate in 2026.
I want to finish this earnings recap of Datadog with a detailed analysis of the current “AI is eating software” theme and how it might affect Datadog specifically.
This narrative (“AI is eating software”) has triggered a strong reassessment of SaaS valuations. The fear that autonomous agents and AI-assisted vibe coding will commoditize enterprise platforms is the hottest debate in tech right now.
To give this the analysis it deserves, let’s first put on the bear suit and tear down Datadog’s future, and then look at the reality of why their moat is likely far deeper than the skeptics realize.
If we assume the most aggressive timeline for Agentic AI, Datadog will quickly transition from a mission-critical platform to an overpriced legacy UI. Here is the bear thesis:
Datadog’s historic value proposition is providing a "single pane of glass”, as they like to say, for human engineers to make sense of complex systems. But AI agents don’t need data visualizations, heat maps, or dashboards. They just need raw data via an API. If autonomous SRE (Site Reliability Engineering) agents can ingest open-source logs, identify a memory leak, and autonomously fix it, paying a premium for Datadog’s human-centric UI becomes completely obsolete.
Historically, Datadog’s moat was the exceptional engineering effort required to build a unified observability suite. Today, the barrier to writing code has basically collapsed. A small team of developers using tools like Cursor or Claude Code can vibe code an observability layer using free OpenTelemetry (for collecting, processing, and exporting telemetry data) standards and cheap cloud storage. They can offer a majority of Datadog’s functionality at a fraction of the price.
Even if Datadog survives, the open-source and AI-native alternatives will hurt their pricing power. As companies realize they can create custom, AI-monitored infrastructure without paying much, Datadog will be forced to reduce prices to maintain market share. And that will hurt their current ~80% gross margins.
The bears would argue this isn’t in the far future. We are already seeing AI agents resolve simple IT tickets today. This suggests that, as agentic frameworks from Anthropic, OpenAI, and open-source models mature, enterprise willingness to sign multi-million dollar observability contracts will fall off a cliff.
Ok, are you scared yet? Well, not so fast, let’s turn this around and explore why Datadog is the ultimate AI "Pick and Shovel” play:
While the bearish narrative sounds intimidating, it fundamentally misunderstands what Datadog actually does at an enterprise scale, and how AI actually behaves in production. Here is the defense of Datadog’s strong moat:
AI Explodes Complexity (and Datadog Taxes Complexity): Agentic AI doesn’t simplify software; it makes it exponentially more complex, distributed, and non-deterministic(!). When an autonomous AI agent starts making decisions like calling APIs, writing to databases, or putting prompts together, it creates a big trail of telemetry. If an agent starts hallucinating or gets stuck in a loop, the businesses running those agents need to know what happened. Datadog has positioned itself as the deterministic control layer that monitors these probablilistic AI models (specifically through their LLM Observability suite).
The SaaS companies that start dying offer things like seat-based workflow tools. If AI replaces human workers, those software seats no longer exist. Datadog, however, is usage-based. They charge based on compute, data ingestion, logs, and traces. Because AI agents work 24/7 and generate significantly more machine-to-machine data than humans, AI usage directly drives Datadog’s revenue up, not down.
Yes, you can absolutely vibe-code a nice-looking dashboard over the weekend. What you cannot vibe-code is a backend capable of ingesting trillions of data points and exabytes of logs securely, in real-time, across a hybrid-cloud environment for a Fortune 500 company. Datadog’s true moat isn’t its code; it is its data gravity, its integration into over 1,000 enterprise systems and its compliance certifications (SOC2, FedRAMP). Rip-and-replacing this embedded infrastructure with a weekend AI project is a catastrophic security and compliance risk that no enterprise CIO will take.
Last, and maybe most importantly: Datadog is not passively waiting to be disrupted. They are actively cannibalizing the space with their own AI. Their Bits AI SRE and Dev agents are already conducting hundreds of thousands of investigations, identifying root causes, and performing fixes directly within the platform. They have the proprietary, unified enterprise data that foundation models desperately need to be useful in a corporate environment. This gives them a very unique moat: By collecting and training on traces, logs, metrics, and security signals from many different enterprises, Datadog creates a data environment that makes its AI agents very effective. While general foundation models typically don’t have access to real-world production systems (other than their own), Datadog’s models are trained on trillions of data points from many different customers, making their insights difficult to replicate for any siloed tool. And the more customers Datadog gets, who share their unique and differentiated data with Datadog, the more valuable Datadog’s AI agents become to existing and new users as they get better when they are trained on more and more diversified data. This is the very definition of a Network Effects moat!
To summarize, AI will hurt lightweight software that merely acts as a wrapper around human tasks. But for infrastructure layers like Datadog, AI is a massive tailwind; it’s a picks and shovels play of the new and coming AI gold rush.
Overview of how Datadog performed versus my prior expectations:
- Revenue expectation: $949M (7.1% QoQ, 28.6% YoY), implying a 3.7% beat.
→ $953.2M (7.6% QoQ, 29.2% YoY), a 4.2% beat. - Q1 new revenue guide: $954M (0.5% QoQ, 25% YoY) which I would interpret as $987M (4% QoQ, 29.6% YoY) expecting YoY growth will accelerate above 29%.
→ $956M (0.3% QoQ, 25.5% YoY), which I now interpret as $994M (4.3% QoQ, 30.6% YoY), implying a 4% beat. - My Q4 revenue expectation implies about $63M raw sequential revenue increase (up from $48M last Q4).
→ $67.5M. - I would like to see RPO at around $3.49b (25% QoQ, 53.6% YoY growth).
→ RPO was $3.46b (24.0% QoQ, 52.4% YoY) with cRPO growing 40% YoY (significantly faster than revenue). - I would like to see Billings at around $1.16b (30% QoQ, 27.9% YoY growth).
→ Billings was $1.21b (35.5% QoQ, 33.3% YoY). - I would like to see QoQ customer growth around 2.2% (~700 new), for the $100k+ cohort, around 3% QoQ (~120 new) and for the $1M+ cohort, around 10% YoY (~508 total).
→ Total customers grew 2.2% (700 new), $100k+ customers grew 6.2% QoQ (250 new), $1M+ customers grew 30.5% YoY!! (603 total). - I would like to see continued multi-product adoption progress with 2+, 4+, 6+ and 8+ products cohort percentages to stay stable at 84%, 54%, 31% and 16%.
→ percentages were 84%, 55%, 33% and 18%. They also added a new 10+ products customer cohort which grew to 9%, from 8%. It is worth pointing out that the 6+, 8+ and 10+ cohorts accelerated YoY growth 30.6% → 38.3%, 46.0% → 63.5% and 75.2% → 96.2%, respectively. - I would like to see NRR around 120%.
→ NRR was around 120%. - I would like to see OM ~24%, NM ~29%, FCFM ~26%.
→ Operating margin was 24.1%, net margin was 28.2% and FCF margin was 30.5%. This was also the first quarter this FY where operating expenses (28.9% YoY) grew again less than revenue (29.2% YoY), after growing faster than revenue in the previous four quarters. OpEx growth peaked in 2Q25 and started decelerating after, while revenue growth started to reaccelerate in 2Q25 and continues to do so through Q4. Good setup for operational leverage. - Thoughts from previous quarter: Datadog’s 3Q25 earnings recap.
Monday:
Monday reported Fiscal Q4 2025 on 02/09/26 and missed almost all of my KPI targets for this quarter. Net new revenue was again much lower than I expected confirming that they are stuck at ~17M net new revenue which is likely only up from their previous ~13M ceiling because of a prior price increase. Monday has shifted their entire focus on larger enterprise customers. As a result we saw total customer NRR dropping this quarter, yet I was looking for stability here. In fact, NRR dropped by 1% in all customer cohorts. Combine that with customer adds on the weak side, at least for the less valuable cohorts, and we get a new narrative that their land and expand has significantly deteriorated (while customer adds and NRRs were still growing last quarter). The only two good numbers I saw this quarter were RPO which continued to accelerate in YoY growth and new product traction with ARR from those now at 11% of total ARR. Also their $500k+ customer cohort continues doing well. Next quarter’s guidance points to a significant additional slowdown for revenue growth to about 22% YoY and the new FY guide suggests this slowdown will continue through the FY. So, after three reports in a row that were on the weak side, it was time for me to pull the plug (which I did in the morning they released their results). Monday might do well going forward with their push upmarket, but with revenue growth continuing to decelerate, NRRs dropping, customer growth rates below my expectations in all cohorts AND profitability margins dropping at the same time, the numbers tell me that even a successful push upmarket might likely not be enough to stop the deceleration in their financials.
Overview of how Monday performed versus my prior expectations:
- Revenue expectation: $337.5M (6.5% QoQ, 26% YoY), implying a 2.6% beat.
→ $333.9M (5.4% QoQ, 24.6% YoY), a 1.5% beat. - Q1 new revenue guide: $347M (2.7% QoQ, 23% YoY) which I would interpret as $356M (5.2% QoQ, 25.8% YoY).
→ $339M (1.5% QoQ, 20.1% YoY), which I now interpret as $344M (3.1% QoQ, 21.9% YoY), implying a 1.5% beat. - I would like to see raw sequential revenue increase around $20.7M.
→ $17.0M. - I would like to see around 264600 total customers (19600 net adds, similar to last year), around 63951 customers with 10+ users (~876 net adds), around 4352 customers in the $50k+ cohort (395 net adds), around 1763 customers in the $100k+ cohort (160 net adds) and around 88 customers in the $500k+ cohort (10 net adds).
→ no total customer number given, 10+ was 63914 (839 net adds), $50k+ was 4281 (288 net adds), $100k+ was 1756 (153 net adds), $500k+ was 87 (9 net adds). - I would like to get an update on CRM, Dev, Service and Campaigns products with new products ARR greater or equal to 11%.
→ new products ARR was 10.7%, up from 10.2% last quarter. - I would like to see that NRR greater or equal to 112%, NRR10+ >=115%, NRR50k >=117% and NRR100k >=117%.
→ total NRR was 110%, NRR10+ =114%, NRR50k >=116% and NRR100k >=116%. NRR dropped by 1% in all cohorts. - I would like to see RPO around $836M (36.3% YoY, 12% QoQ).
→ RPO was $839M (36.6% YoY, 12.3% QoQ) with cRPO growing 31% YoY. - I would like to see operating margin around 14.6% (~$49.4M operating income), net margin around 20% and FCF margin around 24.6% with FCF of around $83M.
→ Operating Margin was 12.6%, Net Margin was 16.5% and FCF margin was 17%, all down significantly from both last quarter a the corresponding quarter a year ago - Thoughts from previous quarter: Monday’s 3Q25 earnings recap.
AppLovin:
Key insights:
- They met my Q4 revenue expectation and new Q1 guide looks somewhat weak, but is affected by seasonality after the strong holiday Q4 quarter, where advertising tends to be strong.
- Despite market fears, increased competition from players like Meta partly strengthens AppLovin’s position by driving up auction values and increasing fee revenue without compromising their most profitable ad placements.
- AppLovin’s e-commerce expansion is gaining serious momentum as successful pilot scaling and new generative AI tools help remove the creative barriers that previously hindered non-gaming advertisers.
- Overall, I thought this was a very strong quarter, mainly because achieving 66% revenue growth alongside expanding 84% EBITDA margins at a multi-billion dollar scale is almost unheard of in public markets.
AppLovin reported fiscal Q4 2025 on 02/11/26. Revenue was $1658M (18.0% QoQ, 65.9% YoY) versus my expected $1656M (17.9% QoQ, 65.7% YoY), and a 4.6% beat. QoQ growth was fantastic, accelerating right through typical holiday seasonality. Their new Q1 revenue guide of $1760M (6.2% QoQ, 51.0% YoY) is probably set low to make it easier for management to play the beat and raise game, with the potential upside that their e-commerce platform (currently scheduled to reach general availability in H2) might become generally available early in H2 rather than late. With that, I am not too concerned that revenue growth will rapidly slow, especially considering that “and despite typical seasonality where Q1 should be softer than Q4, we are guiding to meaningful sequential growth.” So, I interpret this baseline to mean Q1 will eventually land closer to $1857M (12% QoQ, 60% YoY), implying a 5.5% beat.
The highlight of the report was the bottom line, which I wanted to see hold steady at a high level. Well, adjusted EBITDA margin came in at a staggering 84.4% (beating my >=80% expectation) and free cash flow margin hit 79% (crushing my 70-75% target) - wow! The rest of the profitability margins were also great and beat my expectations throughout, as you can see in the bullet points below which summarize and compare to my prior expectations.
So let’s unpack this: Does this mean they are underinvesting in growth to juice their margins?
No, not really. I think what this tells us is how absurdly strong the operating leverage of their software model is. You can see this by looking at their QoQ flow-through to adjusted EBITDA, which management noted was approximately 95%. In other words, almost every incremental dollar of revenue is dropping straight to the bottom line because their AI models scale without requiring a significant proportional increase in headcount. In any case, this shows the strength of AppLovin’s unit economics.
A few weeks ago I saw a lot of noise trying to correlate the rise of large AI models and Meta’s entry into in-game advertising with a decline in AppLovin’s future moat. I saw some people on X (twitter) implying that increased competition in the MAX auction would compress margins and slow growth, but I don’t think this is valid. If one would look closely at how the MAX ecosystem functions, one should really realize that increased bid density actually expands the overall advertising pie. In other words, when for example Meta bids more aggressively, it increases the total demand in the auction, driving up the overall value of ad inventory. Because AppLovin takes a 5% fee on competitor bids, increased competition from Meta actually grows AppLovin’s revenue. Also, as CEO Adam Foroughi noted on the call, when competitors win an impression, it is highly likely to be an impression that AppLovin’s own AXON models valued less. So all this competition narrative can tell us is that the ecosystem is getting healthier, while AppLovin continues capturing the most lucrative economics - making the recent market panic seem quite overblown. Of course, it’ll be interesting to see any future pressure on margins that could signal competition is successfully deteriorating Applovin’s moat. For now, looking at margins, one could argue the opposite is true.
Initially, I thought the e-commerce rollout might be a mixed bag because they are still restricting self-service to referral-only rather than full general availability. But their pilot metrics are scaling beautifully. Management highlighted that e-commerce business (which has been live for about a year and a half) customers who have lapped the platform are increasing spend materially. It was great to hear the anecdote about an Israeli cookware company scaling from $4 million to $16 million in revenue and reaching profitability via AppLovin. What I initially thought could be a bottleneck - the lack of native video creatives from e-commerce brands (who are usually more familiar with advertising using things like banner ads) causing onboarding breakage - is actively being solved. They are currently piloting generative AI creative tools with over 100 customers to automate video generation. If they can eliminate the cost of video creation, the barrier to entry for non-gaming advertisers vanishes. It’ll be interesting to see how successful this will be.
Overall, I thought this was a very strong quarter, mainly because achieving 66% revenue growth alongside 84% EBITDA margins at a multi-billion dollar scale is almost unheard of in public markets. While I think the broader market narrative around AI disruption brings additional stock volatility, we’ll get more clarity - especially regarding their e-commerce general availability outlook - with their next quarterly report.
Overview of how AppLovin performed versus my prior expectations:
- Revenue expectation: $1656M (17.9% QoQ, 65.7% YoY), implying a 4.5% beat.
→ $1658M (18.0% QoQ, 65.9% YoY), a 4.6% beat. - Q1 new revenue guide: $1846M (11.5% QoQ, 59% YoY) which I would interpret as $1921M (16% QoQ, 66% YoY).
→ $1760M (6.2% QoQ, 51.0% YoY), which I now interpret as $1857M (12% QoQ, 60% YoY), implying a 5.5% beat. - I would like to see a GAAP Gross Margin of 87-88%.
→ Their GAAP Gross Margin was 88.9%. - I would like to see a FCF margin of 70-75%.
→ Their FCF Margin was 79%. - I would like to see a GAAP Operating margin greater or equal to 75%.
→ Their GAAP Operating Margin was 76.9%. - I would like to see a Net margin greater or equal to 55%.
→ Their Net margin (from continuing operations) was 66.5%. - I would like to see an adjusted EBITDA margin greater or equal to 80%.
→ Their adjusted EBITDA Margin was 84.4%.
Astera Labs:
Key insights:
- Q4 revenue growth was slightly below my expectations, but strong Q1 guide makes more than up for this and points to reaccelerated YoY revenue growth from Q4 to Q1.
- They signed a large new warrant agreement with Amazon making them a long-term partner. I view this agreement as a win-win-win for Astera (who potentially will 7x its current FY25 revenue until 2033 with Amazon alone), Amazon (who might effectively get some discount on Astera’s products, but only if the share price goes up) and the investors (who will profit from significant revenue growth and increased revenue visibility; although not guaranteed).
- Their workhorse product (Aries retimers) is slowly gaining multi-customer traction, reducing customer concentration risk.
- Their new products, Scorpio (switching) and Taurus (Ethernet/smart cable modules) are quickly gaining revenue share by growing much faster than Aries, further diversifying Astera’s revenue portfolio mix.
- One concern weighing on Astera is that SBC-subtracted operating expenses are currently growing much faster than revenue. I believe this is short-lived though, because operating expenses are currently increased because of acquisition costs for XScale, a new Israeli design center and because revenue from 800G Taurus and Scorpio X won’t fully materialize until late 2026/2027.
- Also, gross margins are currently pressured because of a mix of a new warrant agreement with Amazon and a higher fraction of module-based revenue (Taurus/Scorpio, growing faster than Aries and carry lower margins than pure silicon). Yet, the core business is so profitable that they can absorb a 20% quarterly OpEx hike and still deliver 40% operating margins.
Astera Labs reported Fiscal Q4 2025 on 02/10/26 and delivered a quarter that reinforces its position as a critical infrastructure play in the AI era. While the top-line beat was more modest than my projection (delivered $270.6M (17.4% QoQ, 91.8% YoY), versus my expectation of $276M (19.9% QoQ, 95.6% YoY)), their Q1 guide made more than up for this with $292M at the midpoint vs. my prior expectation of $282, which was based on the assumption that they’ll continue their trend of slightly decelerating YoY growth. Also, from a narrative point of view, the company’s strategic positioning has arguably strengthened - specifically regarding the diversification of its revenue streams and the cementing of its relationship with its largest customer via a significant new warrant agreement.
Astera signed another commercial agreement with Amazon (not the first time!) that includes issuing warrants equivalent of 3.26M shares (versus currently 184M shares outstanding, or 1.8%) vesting based on $6.5 billion in cumulative product purchases over the next several years. For context, that’s over 7x Astra’s entire FY25 revenue and valid through February 5, 2033. This effectively locks in AWS as a major, long-term partner, reducing competitive risk, but will create a 2% non-cash headwind to Gross Margins starting 2Q26 as these warrants are amortized: “To account for the warrant, you do take a non-cash charge for the value of what vests, and that goes directly against revenue and effectively directly against gross margins as well. So as the warrants are achieved, we are kind of modeling a non-cash hit to gross margins of about two points a quarter starting kind of in the Q2 time frame.” I think it is important to fully understand this: This is not a trading “product for shares” agreement. On the contrary, Amazon still pays (at least initially) full cash price for every product they get. However, as they hit certain spending milestones, these 3.26M warrants will vest, giving Amazon the opportunity (but not the obligation) to buy newly created shares from Astera at a price of $142.82 per share. By the way, I don’t think it was a coincidence that the stock moved closer to a share price close to $143 after this announcement, which effectively priced the company at the corresponding enterprise value. So on the positive side for Astera, they could get money twice from Amazon, once when they buy their products and a second time when (and if) they execute the warrants. Of course Amazon would only do this if the current share price at the time will be higher than the warrant price. On the negative side, this will either dilute shareholders or cost Astera the difference in share price if they decide to buy back the stock in order to avoid dilution. I view the latter scenario as basically giving Amazon a cash discount on their products. Given that second scenario, it is not surprising that SEC rules dictate that they cannot fully recognize the full amount of revenue from those product sales, even though they’ll initially get the full amount of cash. So they’ll effectively have to reduce their revenue from Amazon by the estimated discount value, depending on Astera’s current share price value. And of course the way this will show up in their financials is also as a gross margin hit, when their apparent costs of goods sold goes up by that amount. Whether, in the end, they’ll go the discount route by buying back shares or the dilution route is up to Astera. I think for us investors, the dilution or equivalent discount amount is less of a problem, because it is tied to how much the stock price will go up from here. In other words: If Astera stock goes to $300 per share (as an example), will any investor who doubled their money be sad about having been diluted by 1.8% or that Amazon got a 7.7% discount on buying products worth $6.5B (~$150*3.26M shares = $500M)? I doubt it … Anyways, that’s at least how I think about this warrant agreement and if you think I got this or part of it wrong, please let us know!
Next, Scorpio (switching) exceeded 15% of total revenue in FY25 (beating the 10% target, “so we originally set out for a 10% bogey. We did cross above 15% for 2025”). The Scorpio P-Series is shipping in volume, and the Scorpio X-Series (targeted at backend GPU-to-GPU clustering) has started initial shipments, with a volume ramp expected in 2H26. This is a direct play on the “scale-up” architecture (like NVIDIA’s NVLink/Infiniband domain) but using PCIe/CXL.
Taurus (Ethernet/smart cable modules) was the strongest performing product family in Q4, growing 4x YoY. This confirms Astera is winning in the front-end network (GPU-to-Switch) transition to 400G and 800G Active Electrical Cables (AECs).
Aries (Retimers) remains the workhorse and grew 70% YoY, with PCIe Gen 6 becoming the industry standard. Management cited multi-customer traction implying reliance on a single hyperscaler (or two, if we count Microsoft as well as Amazon) is decreasing.
One concern is rising operating expenses. Q1 guidance is for a 20% increase over Q4, after being up 22% QoQ from Q3 to Q4. So the concern is that SBC-subtracted operating expenses grow faster than revenue and if that would continue the company’s bottom-line will be in trouble. But based on the earnings commentary, this increase appears to be a calculated, temporary investment spike rather than a loss of discipline. Here is why a “Profitability Cliff” scenario is unlikely in the near term: A significant portion of the OpEx increase is tied to the acquisition of XScale. These are inorganic additions to R&D headcount to accelerate the Scorpio X-Series roadmap. This is a land grab move to ensure they have the IP to compete with Broadcom and Marvell in the switch market. They also formally announced a new design center in Israel. This is a one-time step-up in fixed costs to access a new talent pool, not a variable cost that will scale up forever. Finally, management explicitly stated that the OpEx ramp is to support the $25B TAM opportunity, a 10x increase over the previous TAM, that 1they see over the next 5 years. The revenue from these R&D dollars (specifically for 800G Taurus and Scorpio X) won’t fully materialize until late 2026/2027.
So while the times of 77%+ Gross Margins may be paused for FY26 due to a mix of the previously mentioned Amazon warrants and a higher mix of module-based revenue (Taurus/Scorpio which carry lower margins than pure silicon), the good news is that Operating Margin is holding at 40%. This suggests that while Gross margins are compressing slightly, the core business is so profitable that they can absorb a 20% OpEx hike and still deliver 40% operating margins.
Overview of how Astera Labs performed versus my prior expectations:
- Revenue expectation: $276M (19.9% QoQ, 95.6% YoY), implying an 11% beat.
→ $270.6M (17.4% QoQ, 91.8% YoY), a 8.7% beat. - Q1 new revenue guide: $282M (2% QoQ, 77% YoY) which I would interpret as $302M (9.6% QoQ, 89% YoY).
→ $292M (7.7% QoQ, 83% YoY), which I now interpret as a range of potential outcomes with $303M (12% QoQ, 90% YoY) at the lower end, implying a 4% beat and $317M (17% QoQ, 99% YoY) at the higher end, implying a 8.6%. The reason for the wide range is their limited financial history and whether Q1 is seasonally weak or not. - I would like to see a Gross Margin of 76-77%.
→ Gross Margin was 75.7%, a slight QoQ compression due to a higher mix of hardware modules (Taurus/Scorpio) which carry lower margins than pure silicon. - I would like to see an Operating margin around 40%.
→ Operating Margin was 40.2%. - I would like to see a Net margin around 38%.
→ Net Margin was 38.7%.
Axon:
Key insights:
- After a mixed Q3, this Q4 was clearly a blowout with revenue growth well ahead of my expectation and accelerating YoY to 38.5%, up from 30.6% in Q3.
- ARR, just like in Q3, missed my expectation, but again, the underlying bookings were surprisingly strong, resulting in RPO growth acceleration to 42.6% YoY, from 39.0% in Q3 and demonstrating the lumpiness in ARR recognition is more due to deal-to-revenue conversion timing, and not a weakening in demand.
- Very strong new FY26 guide: 30% YoY (topline). For comparison, their initial FY25 topline guide was for 27.2% YoY and they ended up growing 33.5% YoY in FY25.
- I didn’t like that they gave an FY28 target (3 years out), because I don’t think they needed to to that and because I doubt that they have visibility that far out. However, if true revenue should continue to grow north of a 29% CAGR for the next 3 years, which would be fantastic.
- Adjusted gross margin dropped to 61.1%, missing my >62.7% expectation, which was based on the assumption of a revenue mix-shift towards higher margin software. However, this quarter, hardware sales were exceptionally strong. And since those hardware sales are seeding the market with future higher margin software upsells, I am not too concerned about this temporary gross margin drop.
- Despite the gross margin drop, net margin (22.3%), EBITDA margin (25.9%) and FCF margin (19.7%) were strong this quarter.
Axon reported Fiscal Q4 2025 on 02/24/26. If the third quarter presented a complex narrative around near-term profitability AND revenue recognition missing my expectations, Q4 delivered a resounding validation of the company’s long-term strategic momentum. The business is firing on all cylinders, widening its moat through accelerated adoption of its AI-driven ecosystem and expanding deeper into enterprise and international markets. Unlike Q3, where both revenue and adjusted EBITDA fell short of my targets, Q4 saw significant outperformance on both the top and bottom lines. Revenue of $796.7 million was way above my $767 million expectation, delivering 12.1% sequential and 38.5% YoY growth. Even more strikingly, adjusted EBITDA came in at $206 million, crushing my $173 million target and easing the profitability concerns that surfaced last Q. With top-line reacceleration combined with operating leverage, this quarter perfectly encapsulated the overall thesis: Axon is successfully stacking multiple S-curves - from TASER 10 to its expanding AI software suite - that are translating into immense financial scale.
The intricate gross margin story, however, requires another nuanced look. While adjusted EBITDA margins expanded beautifully, adjusted gross margin dropped to 61.1%, missing my >62.7% expectation. Management clearly attributed this compression to two factors, which were basically the same as last quarter’s reason for GM pressures: ongoing global tariffs and, more importantly, an increased Platform Solutions product mix within Connected Devices. Essentially, hardware-related growth was phenomenally strong this quarter, pushing overall gross margin down even as Software & Services margin remained structurally stable. I view this as a high-quality problem. As Connected Devices revenue surged 38% year-over-year to $454 million, Axon was effectively seeding the market. Every new TASER 10 or Axon Body 4 sold acts effectively as a hardware trojan horse for future high-margin software subscriptions. This dynamic is heavily reinforced by Software & Services revenue continuing its rapid rise, growing 40% year-over-year to $343 million.
Looking at customer adoption metrics, the $95 million in net new ARR technically missed my aggressive $125 million goal, bringing total ARR to $1.35 billion. However, just like in Q3, any initial optical softness in recognized ARR was completely overshadowed by an absolute explosion in underlying bookings. Future contracted bookings (RPO) increased to a staggering $14.4 billion, blowing past my $13.7 billion estimate and growing 43% year-over-year. Management highlighted on the call that full-year bookings surpassed $7.0 billion, with Q4 bookings up more than 50%, representing a major acceleration. The narrative around the “AI Era Plan” is moving in the right direction, representing about $750 million or approximately 10% of total bookings. Coupled with an increase in Net Revenue Retention (NRR) from 124% in Q3 to 125% in Q4, alongside big narrative wins like the largest single-customer booking in company history within Corrections and international bookings crossing $1 billion for the first time, the lumpiness in ARR recognition is clearly just a matter of deal-to-revenue conversion timing, not a weakening in demand.
Regarding guidance, I don’t think the massive Q4 outperformance was a result of forward pulled demand; rather, it established a higher baseline: They initiated a full-year 2026 outlook of 27% to 30% revenue growth, marking a confident start to the year. Finally, I found it interesting that they introduced new 2028 financial targets: achieving approximately $6.0 billion in annual revenue which corresponds to a 29% CAGR [(6000/2779)^(1/3)-1] and a 28% adjusted EBITDA margin (and lowering their cap on dilution from 3.0% per year to 2.5% per year). While I think it would be amazing if they could maintain 30% YoY revenue growth for the next 3 years, I find it kind of unnecessary for a rapidly growing company to provide guides like this. While this long-term guide implies a high degree of confidence from leadership that recent strategic acquisitions like Fusus, Prepared, and Carbyne are rapidly expanding their TAM, I kind of wish they hadn’t given this guide because I doubt that they have the visibility three years out. Maybe they felt it was necessary to show the current market that they won’t be disrupted by AI. Whatever it maybe, the core thesis, however, remains fully intact: Axon is selling more software to more users at higher price points, and the margin profile will follow the software mix-shift over time.
Overview of how Axon performed versus my prior expectations:
- Revenue expectation: $767M (8% QoQ, 33.5% YoY), implying a 1.3% beat and implying a 0.4% beat of their FY guide given in Q3 (compared to past Q3 FY guide beats: FY21: 1.6%, FY22: 2.6%, FY23: 0.7%, FY24: 0.6%).
→ $796.7M (12.1% QoQ, 38.5% YoY), a 5.2% beat (and beating the Q3’s FY guide by 1.4%). - Q1 revenue expectation: $805M (5% QoQ, 33.5% YoY), assuming similar YoY growth as in Q4.
→ $836M (6.2% QoQ, 38.6% YoY), assuming similar QoQ and YoY growth as in Q4. - I would like to see around $125M net new ARR (total ARR to ~$1.38b).
→ $95M net new ARR (total ARR was $1.35b). - I would like to see RPO around $13.7b.
→ $ RPO (future contracted bookings) was $14.4b. - I would like to see NRR around 124%.
→ NRR was 125%. - I would like to see gross margin greater than 62.7%.
→ adjusted gross margin dropped to 61.1%. The decrease in total company gross margin and adjusted gross margin is primarily due to global tariffs and increased Platform Solutions product mix within Connected Devices, partially offset by growth in Software & Services; essentially: Hardware related growth was very strong this Q, pushing GM down. Software & Services GM stayed constant. - I would like to see adjusted EBITDA around $173M.
→ adjusted EBITDA was $206M. - Thoughts from previous quarter: Axon’s 3Q25 earnings recap.
Wrap up
Other than Monday, which I sold out of promptly after reviewing their earnings, this earnings season has been exceptionally strong for our portfolio companies that reported so far. Cloudflare, Datadog and Axon were the highlights with clear blowout numbers and strong revenue growth acceleration. I also think AppLovin and Astera Labs posted strong numbers and from what I have seen so far, Nvidia posted another blowout, while Snowflake and Zscaler reported very solid numbers, into which I’ll still have to dive a bit deeper. Looking forward to seeing what Crowdstrike and Samsara will post when they report next week.
With that, I am wishing you all a great March!
Ben
Past recaps
2022: Jul 2022 | Aug 2022 | Sep 2022 | Oct 2022 | Nov 2022 | Dec 2022
2023: Jan 2023 | Feb 2023 | Mar 2023 | Apr 2023 | May 2023 | Jun 2023 | Jul 2023 | Aug 2023 | Sep 2023 | Oct 2023 | Nov 2023 | Dec 2023
2024: Jan 2024 | Feb 2024 | Mar 2024 | Apr 2024 | May 2024 | Jun 2024 | Jul 2024 | Aug 2024 | Sep 2024 | Oct 2024 | Nov 2024 | Dec 2024
2025: Jan 2025 | Feb 2025 | Mar 2025 | Apr 2025 | May 2025 | Jun 2025 | Jul 2025 | Aug 2025 | Sep 2025 | Oct 2025 | Nov 2025 | Dec 2025
2026: Jan 2026