Almost everything was sent during the week. In case you missed it:

I plan to publish a full Datadog review next week.

Demand:

Profits & Margins:

Balance Sheet:

Guidance & Valuation:

DDOG trades for 73x forward EPS. EPS is expected to grow by 2% this year, before resuming 23% compounding over the next two years. It also trades for 50x forward FCF. FCF is expected to grow by 7% this year before compounding at a 28% clip over the next two years. They’re in investment mode for 2025.

I also plan to publish an Airbnb review next week.

Demand:

Profits & Margins:

Balance Sheet:

Guidance & Valuation:

Q3 revenue guidance was slightly better than expected. They also reiterated 34.5%+ EBITDA margin guidance for the year and called for $2B+ in Q3 EBITDA. This was at least 6.9% better than expected, although estimates just dipped from $2B to $1.88B right before the report.

ABNB trades for 27x GAAP EPS and 15x EBITDA. GAAP EPS is expected to grow by 3% this year and 12% next year. EBITDA is expected to grow by 5% this year and by 11% next year.

Demand:

APP slightly beat revenue estimates & beat guidance by 4.6%.

Profits & Margins:

App beat EBITDA estimates by 2% & beat guidance by 3.9%. Free cash flow missed estimates by about 5%. APP sold a business division recently and did not reconcile continuing operations numbers for Q2 2023 FCF. The comp is not apples-to-apples.

Balance Sheet:

Guidance & Valuation:

Revenue guidance was 2.3% ahead of expectations, while EBITDA guidance was 2.9% ahead of expectations.

App trades for 37x forward EPS. EPS is expected to grow by 94% this year and by 42% next year.

Demand:

Axon beat revenue estimates by 4.5% and beat annual recurring revenue estimates by 3.6%.

Profits & Margins:

Axon beat gross margin estimates by a point, beat EBITDA estimates by 7.5% and beat $1.47 EPS estimates. This included a $75M tax benefit. Performance awards for the leadership teams are greatly impacting GAAP margins.

Balance Sheet:

Guidance & Valuation:

Axon raised annual revenue guidance by 1.5%, which beat estimates by 1.1%. It also raised annual EBITDA guidance by 1%, which slightly missed estimates.

Demand:

Rocket Labs beat revenue estimates by 6.7% & beat guidance by 7%.

Profits & Margins:

Balance Sheet:

Guidance & Valuation:

RBC Capital Markets analyst Matthew Swanson published takeaways from a follow-up call with Trade Desk leadership. They again highlighted their relative skew towards large enterprises, and especially automakers and CPG giants, where tariffs are hitting especially hard. The company also enjoyed some significant demand pullforward during Q1, which shifted some revenue out of this quarter and contributed to that large 7% beat. Next, leadership told RBC there was a “traditional amount of conservatism when bringing in a new CFO.” That, to me, means guidance was sandbagged and TTD should handsomely beat next quarter. And speaking of the new CFO, Trade Desk wanted to bring in more of a “strategic financial thinker” to usher in the next era of growth.

Finally, Swanson said they reiterated confidence in 2026, following Green casually guiding to an acceleration during the earnings call. Q3 and Q4 outperformance would likely put revenue growth around 19% for the year (somewhere around 21% ex-2024 election), with 20%+ growth then expected by management in 2026.

All of this makes me more confident in Friday’s earnings reaction being far overdone and this company being capable of growing right around 20% (with excellent margins) for a long time.

Aside from this RBC note (where they reiterated a buy rating), most sell-siders, like me, think this is large enterprise and tariff-driven. Wedbush and Bank of America were more concerned about structural growth in their downgrades, but candidly, I will go with the words of CEO Jeff Green over these guys. Listening to him has been uniformly the correct decision since I started following this name. Wall Street doubts in that man and this company have made me a lot of money over the last several years. I don’t think this time will be any different, but time will tell.

Fortinet commentary on the firewall refresh cycle led to the stock plummeting. Zscaler and many other cybersecurity names followed suit, with investors selling now and asking questions later. Zscaler does not have a legacy firewall business and is not reliant on product refresh cycles like this one is. It offers a software and cloud-native zero trust security platform that is meant to supplant these archaic firewalls. These truly zero trust platforms help customers do more with less, and are quickly gaining market share in the overall security market. Fortinet having issues with selling the product that is losing market share is not a reliable readthrough for Zscaler’s upcoming results. Cloudflare’s positive commentary on their own network security business is more relevant.

If there’s one thing the semiconductor industry loves, it’s constantly changing the names of products with a swarm of acronyms for us. Fun, fun, fun. Those acronyms all fall into neat categories: chips, networking and connectivity, and software. It’s these ideas and AMD’s positioning within them that matter to investors. Not that they’ve memorized what an MI350 HBM3E chip stands for. That’s how we’ll frame this coverage.

GPU: Graphics Processing Unit. This is an electronic circuit used to process information and data. The accelerated compute needed for GenAI apps and models pulls from next-gen GPUs. It thinks its “MI” series of GPUs (part of the “Instinct” product family) boasts best-in-class memory and bandwidth, which Nvidia would certainly disagree with. AMD also thinks its 2025 Instinct release will compete with Nvidia’s world-class Blackwell platform.

CPU: Central Processing Unit. This is a different type of electronic circuit that carries out assignments and data processing. CPUs fall in the general compute bucket. General compute CPUs are still optimal for static, step-series and instruction-based tasks. They’re also much cheaper than deploying next-gen GPUs when they can work for the specific use case. AMD’s new AI data center CPUs “extend leadership in performance per watt and dollar.”

NPU: Neural Processing Unit: Used for AI-enabled personal computers (PCs).

TOPs: Tera Operations Per Second. This measures NPU performance, with more TOPs being better. TOPs superiority is imperative for running Copilots and GenAI apps on PCs with optimal latency, hallucination rates and performance.

All of the sharp margin weakness seen below is related to Chinese export restrictions and the associated $800M inventory charge.

Guidance does not include any revenue from MI308 shipments to China, but leadership is optimistic that it will secure licenses to unlock that business. This could be a nice source of upside. It won’t be an immediate contributor, as AMD doesn’t really have much finished inventory on hand, but it would be material for sure. The annual run rate for the business was about $1.5B before rules changed. AMD also expects modest client and gaming growth during Q3 and for the embedded segment to return to Y/Y growth.

Data Center CPU Business:

AMD’s new EPYC 4005 Series CPU, per leadership, delivers convincing performance and total cost of ownership (TCO) leadership in the market. This spans cloud, enterprise and AI workloads. And that’s worth highlighting. While we all rightfully correlate massive CapEx guidance from mega-caps as very positive for GPU high-performance compute demand, a lot of that is still going to CPU generate compute. CPU leadership is a big part of its world-class supercomputers and how it powers more than 1/3 of the fastest in the world. Agentic AI and token generation require significant general compute capacity where CPUs are usable and more affordable. This is supporting rising demand for its new 5th generation EPYC products (called Turin), as well as durable demand for older EPYC models. 

All in all, data center CPU sales for public cloud and enterprise customers took more market share for the 33rd straight quarter, and doing that for a 34th is looking highly likely. In cloud specifically, hyperscale demand remained strong, as it added 100+ EPYC-based instances (cloud servers), including “multiple Turin instances with Google and Oracle Cloud.” There are now 1200 total AMD-based instances in public cloud environments. So far, this Turin-based compute is doubling performance vs. the previous generation (which was industry-leading). Nokia was named as a new Turin customer for its cloud platform. For Enterprise CPU’s it announced that KDDI will use its 4th generation EPYC CPUs for their virtual 5G network.

For on-premise data center CPUs, Dell is ramping up nicely, while Lenovo, HPE and Super Micro added 28 new Turin platforms with best-in-class total cost of ownership for enterprises.

Data Center GPUs:

It was an interesting quarter for AMD’s Instinct GPU platform. Export restrictions for its MI308 GPU sales to China drove a Y/Y decline in data center GPUs. I personally found this a little surprising. We’ve seen Nvidia report quarters where they are also prevented from selling their H20 chips to China, yet were still able to deliver explosive growth due to fantastic demand everywhere else.

Elsewhere, AMD’s MI300 and MI325 products secured more wins across a wide array of large customers, with 7 of the 10 top model builders now Instinct customers. Looking ahead to the rest of the year, its MI350 GPUs and specifically the MI355 are expected to be popular. Per leadership, they are “best-in-class in terms of performance, efficiency and scalability for GenAI and high-performance compute.” They think it “matches or exceeds Grace-Blackwell 200 (Nvidia’s current product) for training and inference and delivers 40% more output tokens per dollar, significantly lowering cost. Product began before expected in June and early adoption is exceeding expectations thus far. It’s worth noting that Nvidia will launch its Rubin platform later this year, which is expected to represent another large leap forward in performance and I think this is encouraging. At the very least, the gap is not widening between these two players, and AMD sees a “clear path to scaling the AI business to tens of billions in annual revenue.”

I wanted to go back to data center gross margin excluding the export restrictions for a moment. Nvidia’s gross margin excluding export restrictions in its most recent quarter was 71% and data center GPU revenue was above that. For AMD, the GPM is 54%, and GPUs continue to be a negative contributor. GPM is a great metric for gauging pricing power and technological edges, and it indicates that Nvidia still has a commanding lead in the world of GPUs. Both companies say they’re better than each other in terms of performance and total cost of ownership, but this data is notable and even AMD leadership talked about needing GPUs to be dilutive to GPMs to grow presence in the market. It’s a massive opportunity, there won’t be one winner, and AMD is the clear second best. It also doesn’t need a 71% gross margin to be very successful, and can keep undercutting Nvidia to win its fair share of the market for workloads not requiring the absolute best GPU.

Software:

Building a thriving software ecosystem to complement its chip business matters. It’s what enables seamless integrations with other vendors and drives performance gains for existing hardware. This quarter, it announced the newest iteration of its ROCm software platform (ROCm 7). This triples software-based training and inference performance contributions vs. ROCm 6. ROCm also added a new developer cloud to give builders easier access to its GPUs. That should be positive for overall work done to improve the performance of its hardware. Support for Llama, Gemma and DeepSeek models should be too.

GPU Roadmap:

Its MI40 GPU is on track to ramp throughout 2026. They think it offers 50% more memory bandwidth and scalability than Nvidia’s current offering. It’s also expected to 10X performance for model usage, which it “believes will be the highest-performance AI system in the world at launch.” Software integrations will be a big piece of this launch, as it gears up to deploy a full-stack AI rack scale product called Helios. It’s expected to be a material 2026 revenue contributor. Helios will be able to combine 72 GPUs, creating more powerful, dense and efficient compute capabilities (similarly to Nvidia’s NVLink). This level of GPU linking within the rack scale system is made possible by its recent purchase of ZT Systems. That team has already been integrated without drama and is supporting Helios’s demand and development. And while we’re talking about ZT Systems,

Client & Gaming Revenue:

Client revenue rose 67% Y/Y thanks to great Ryzen CPU demand – especially in desktop. It secured a long-term Microsoft deal to build custom chips for their PC and Xbox businesses and added several other Forbes 200 and large enterprise wins as well. That’s related to continued innovation, as it recently unveiled a new Ryzen product for “the most demanding workstation workloads” and 2x the performance of competing offerings for content creation and design use cases. For mobile client revenue, sellout rates rose by more than 10% Y/Y, with pricing power rising accordingly and market share gains continuing. They expect that to be an ongoing theme.

Embedded Segment:

This was a good quarter. They’re showing strong signs that new GPU products are popular and an ability to compete on price to plug potential performance gaps. The client and gaming segment performed extremely well and their leadership across major CPU buckets is loud and clear. I still think Nvidia is the clear GPU leader, as evidenced by their margin profile and growth trajectory. The multiples between the two companies are nearly identical, and Nvidia net profit growth is expected to lead AMD through calendar 2026. For this reason, I prefer Nvidia as an investment case, but don’t think there’s anything glaringly wrong with AMD. Lisa Su is a great CEO and the company should continue to be just fine. It can build a great GPU business as the second best vendor in this highly compelling space.

Wedbush and Bank of America both downgraded Trade Desk due to Amazon competitive concerns and structural growth deceleration. They clearly don’t believe management’s confidence in accelerating 2026 growth. I do.

Alphabet’s Gemini and other AI products from the tech giant were added to the Federal Government’s approved vendor list.

Amazon is now selling used cars on its marketplace.

Output Data:

Employment Data:

Inflation Data: