Because this was the end of Visa’s fiscal year, it offered brand new 2025 guidance for revenue and EPS. Both were in line with expectations. Considering the company represents trillions in annual spending, its take on the consumer matters a lot. Here’s how it feels about 2025:

“As we regularly say, we are not economic forecasters so we're assuming the macroeconomic environment stays generally where it is today. As such, we expect payment volume and processed transaction growth to remain strong and generally in line with full year 2024 levels.”

CFO Christopher Su

Visa trades for 26x forward earnings. Earnings are expected to grow by 13% this year and by 12% next year.

TMDX makes Organ Care Systems (OCS). With it, organs from donors can be preserved for longer periods of time and transported more easily to recipients. This is a new name in the coverage network that I plan to cover in more detail going forward. From my outsider perspective, this is the bear/bull debate:

Bears say margin pressure is structural, transplant availability creates a lack of visibility for forecasting and smaller competition can further erode pricing power and demand. It’s also expensive from a traditional valuation point of view (PEG ratio is more reasonable at roughly 1.3x).

Bulls say margin pressure is due to the build-out of aviation fleets to drive more competitive differentiation. That’s temporary. They point out that utilization rates of the fleet are still low and immature, which is leading to revenue headwinds. That’s also temporary. They’ll also point out that this is a share leader in a sector that should deliver strong growth for a long time. Finally, while margins are suffering today, they remain profitable and did maintain revenue guidance. I find the bull case more compelling, but I have a lot more work to do here.

"We maintain our conviction in our growth runway for 2025 and beyond. We remain well on track to reach our stated 2028 targets.”

Presser

TMDX reiterated $435 million revenue guidance, which missed by 2.3%.

TMDX trades for 70x earnings. EPS is explosively inflecting this year and expected to compound at a 50%+ clip over the following two years.

If there’s one thing the semiconductor industry loves, it’s constantly changing the names of products with an alphabet soup of acronyms for us to juggle. 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 MI325 HBM3E chip stands for. That’s how we’ll frame this coverage, with an emphasis on data center results.

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.” I

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. AMD’s new Ryzen AI PC has 20% more TOPs than Microsoft’s best unit. To AMD, TOPs superiority is imperative for running Copilots and GenAI apps on PCs with optimal latency, hallucination rates and performance. It’s how the firm claims to be a “leader in AI inference on the PC side.”

Next quarter revenue guidance missed by 0.7% while its 54.0% GPM estimate slightly missed 54.2% estimates. EBIT estimates for Q4 missed by 7%. AMD trades for 38x forward earnings. Earnings are expected to grow by 28% this year and by 60% next year.

The company remains very optimistic about 2025 data center growth. It also thinks the PC market in 2025 should materially brighten during the second half of the year. Channel inventory and comp dynamics should become more favorable by then.

EPYC is its lineup of data center CPUs for general compute. While GPUs are leaned on for complex inference, not all workloads here need this level of sophistication. CPUs work well for static, step-series instructions and tasks. They’re good building blocks to be accelerated by high performance GPUs. AMD took market share in the EPYC CPU market this quarter and enjoyed an accelerating pace of large enterprise wins. Its 5th generation of the hardware (called Turin) is now live and delivering significant total cost of ownership leads over alternatives like Intel. Today, these CPUs power Office 365, Facebook, Uber, Netflix and many, many other modern companies. I say this to hammer home the idea that CPU demand will not vanish because GPUs work so well for GenAI training and inference.

In other news, AMD, Intel and others formed an x86 (CPU architecture) “advisory group” to band together to accelerate CPU innovation. Advancement in performance and efficiency from CPUs has greatly slowed down recently. That’s why using them for GPU high performance compute use cases leads to “compute inflation.” That’s how Jensen puts it. These CPUs get overwhelmed by the amount of data needed to be processed and fail.

AMD’s new MI300x GPU ramp is currently happening. Meta is using it in its Llama frontier models while Oracle, Microsoft and other providers are offering public cloud instances to bolster virtual availability. It sees MI300x as having total cost of ownership (TCO) leadership across most models. At the same time, its newest MI325X chip offers a 20% performance boost vs. Nvidia’s Hopper 200 (H200) chip. That’s amazing, but Nvidia is already on its next Blackwell architecture, which remains best-in-class for performance. It’s just really hard to catch Jensen Huang here. Knowing this, AMD has successfully shrunk its platform release cadence from 2 years to 1 year to rival that fierce competitor.

“In the Data Center alone, we expect the AI accelerator TAM will grow at more than 60% annually to $500 billion in 2028. This is roughly equivalent to annual sales for the entire semiconductor industry in 2023.”

CEO Lisa Su

AMD now sees $5 billion in 2024 data center revenue vs. $4.5 billion previously. Margins on its newest MI chips will take time to reach previous model levels. That’s always the case.

“Once we continue to ramp up the revenue, we do think we'll have the opportunity to continue to improve gross margin.”

CEO Lisa Su

Nvidia Inference Microservices (NIMs) are essentially full-service packages of Nvidia software and hardware needed to run HPC models and apps. There has been significant industry criticism against Nvidia using its ecosystem to avoid allowing open integrations with AMD and others. This creates vendor lock, considering customers know Nvidia’s hardware is best. That makes them even tougher to compete with (in perhaps an unfair way according to some).

AMD isn’t playing victim here. It just launched its latest software stack with a “trust open alternative” to other products like Nvidia’s. This way, AMD can sell itself as the ecosystem partner that allows clients to use whatever products and vendors they want, rather than taking more of an all or nothing approach. Since debuting this software, AMD has used it to extract 240% performance gains from its MI300x chips. For other customers, it’s routinely delivering 30% performance boosts vs. other software products.

AMD’s supply chain is in good shape and poised to handle expected demand ramps for its new Ryzen processors and MI chips.

Client revenue rose 29% due to strong demand for its new desktop processors. Sales here rose by more than 10% Y/Y, with growth led by its new Ryzen 9000 series processors. These deliver leading productivity, gaming and content creation performance, per CEO Lisa Su. The newest Ryzen processors will launch next month. For mobile devices, Ryzen sales are enjoying large ramps from Asus, Lenovo and more. Per the team, HP and Lenovo will 3x usage of Ryzen AI programs this year. They’re expecting this to yield considerable market share gains.

When asked about Nvidia, leadership seemed to hint at thinking they’d be caught up to the company by the time they release MI 400 in 2026. The issue is that nobody knows what kind of performance dynamics Nvidia’s Rubin platform (will come after Blackwell) will yield. Still, it’s a massive market and while Nvidia has taken pretty much all of it to date, AMD is confident in taking a piece of this “multi-generational” opportunity down the road.

The quarter really wasn’t that bad… it just wasn’t Nvidia-level good. Investors have been yearning to see a direct uplift in revenue growth for other processor competition like Nvidia, Intel and many others. That has not come for anyone besides Nvidia. Yes, AMD did deliver triple-digit data center revenue, but overall growth continues to look very normal while Nvidia maintains triple-digit overall top line expansion. Furthermore, while a 53.6% EBIT margin for data center revenue is undeniably respectable, it doesn’t look as amazing next to Nvidia churning out a 70%+ GPM. Faster overall growth with better pricing power tells us that Nvidia remains well ahead in this high performance processor race today. That does not mean AMD can’t catch up with its MI series GPUs… it just means that it hasn't.

As a reminder, Google flags emails that use the name of its company too many times in a send. It assumes a sender is trying to impersonate them. For this reason, I’ll refer to the company as GOOG or “Search King” throughout the piece. 

Overall, revenue beat estimates by 2.4%. Cloud revenue beat estimates by 5.0%, YouTube slightly beat estimates and Search beat estimates by 0.9%. The firm’s 10.7% 3-year revenue compounded annual growth rate (CAGR) compares to 11.0% Q/Q & 13.3% 2 quarters ago.

YouTube ads & subscriptions crossed a $50 billion revenue run rate. Retail and insurance were the two strongest segments for advertising revenue this quarter (maybe insurers are finally getting needed rate hike approvals).

Cost of goods sold rose 12% Y/Y due to more content costs, depreciation from CapEx and moving its Pixel 9 device launches from Q4 to Q3. The profit beats would have been slightly, slightly larger without this headwind. OpEx rose 5% Y/Y due to real estate impairment charges, more depreciation. This was offset by lower legal fees. Specifically, R&D rose 11% Y/Y, sales & marketing (S&M) rose 5% Y/Y and G&A fell 10% Y/Y due to lower legal charges and previous headcount cuts. Finally, the Search Giant paid $3 billion in fees related to a 2017 European Commission fine. Excluding this, FCF margin would have been 23.4%. It also moved some tax payments from Q3 to Q4 last year, which makes the Y/Y FCF comp more difficult.

The company sees 0% Q/Q CapEx growth in Q4 vs. $13 billion during Q3. It sees CapEx growing in 2025, but at a slower pace than during 2024. The company also cited a Y/Y comp headwind from lapping seller strength in Asia Pacific last year. That will also mean tougher comps for Meta when it reports tomorrow too.

It trades for 23x forward GAAP earnings (probably closer to 21x after upward revisions). EPS is set to grow by 32% this year and by 14% next year.

CEO Sundar Pichai is adamant that its full stack approach to GenAI innovation is working. Its global data center infrastructure, tight processor/chip partnerships and its own tensor processing unit (TPU) innovation all ensure it has the performance and capacity needed to operate in a high performance compute world. Its “world class research” teams ensure its foundation Gemini models, through its own work and partnerships with Anthropic, stay cutting edge.

Next, it has 7 different products with over 2 billion monthly active users to season Gemini models with more relevant data than basically anyone else on the planet can. I’ve said it before and I will keep saying it: GenAI favors incumbents. If we consider two identical models in terms of parameter and technical construct, the better model will be whatever is trained on the most data.

Lastly, its BigQuery data warehouse paired with a managed cloud environment and Gemini means customers have an ability to tap into and further customize various models with whatever data they specifically need. It means developers can access insight to season models or work on apps across any cloud, with ultra-low latency and minimized cost.

So to summarize, its hardware and cloud products mean it can offer customers fully managed and maintained environments to run their GenAI workloads with optimal performance. Its gigantic consumer products mean clients can access more unstructured and structured data (and their own data) to get models to work best for their specific needs. Its BigQuery product means they always have easy access to all of this insight in a fully managed, secured environment. That is “full GenAI stack” in a nutshell.

In Search:

Search is where the value of hardware vertical integration is already quite concrete. In just 18 months, through equipment optimizations, it has cut AI overview query costs by 90%. This is despite the foundational Gemini model being used doubling in parameter size during the same period. The margin tied to GenAI queries vs. traditional search has been a key concern for this tech giant, and anecdotes like this work wonders to assuage that anxiety. There’s a lot more optimizing to do here to keep making this business more and more profitable for the Search King.

Shopping ads right above & below AI overviews are now live in the USA and the company is now adding placements directly within the AI-generated responses. This should mean more revenue for the same amount of digital real estate. All overview ad placements to date have performed very well for the company, and this should be no different. For now, AI Overview monetization is already at parity with the legacy search business, and it sounds like there’s more upside as it grows ad load.

This week, overviews debuted in 100 new countries representing 1 billion monthly users. Early learnings in the USA and these markets show “strong engagement” and incremental demand creation beyond traditional search. This is creating new use cases and more complex querying, which is becoming truer with time as users realize what questions they can now ask the software.

Lens is another GenAI tool that creates new ways to search. A user can access their photo library or camera to initiate a search with a picture and can enjoy real-time text translation too. Product search through Lens has become quite popular, which creates easy opportunities for more ad demand.

Circle to Search is yet another cutting edge tool that, as the name indicates, lets customers circle a good on their screens to initiate a query. This is most popular with younger users. Lastly, Astra is the company’s next planned AI assistant that supposedly offers a “leap forward” for AI assistants. We’ll hear a lot more about this next year.

Advertising – all products are review from last quarter:

It’s one thing to create products, but it’s better if those products are… well… used. Here’s some data:

Case Studies:

To catch up in GenAI innovation, the company had to make some difficult decisions. It laid off some people, combined several teams, overhauled organizational structure and set out to “reset its cost base.” It has done a decent job of that so far. Still, this company remains the mega-cap with the most headcount bloat and most cost to trim. That will unfortunately probably be a lever that it continues to pull in the coming years to keep bolstering margins. It just doesn’t need as many people as it hired throughout the pandemic. If Sundar is reading this, maybe time for a 5-day in-person work mandate like Amazon did to speed this process up. 25% of all source code at the company is now created by algorithms rather than people. That proportion will continue to rise and the newfound layers of automation will likely embolden it to keep trimming redundant costs.

Sometimes it’s easy to take elite numbers for granted. 35% Y/Y growth represents its fastest expansion in nearly two years. Expanding an EBIT margin from 3% to 17% at this size and with that type of growth is just not normal. This was phenomenal. I’d be curious to know how much of this success was tied to Nvidia GPU reselling, but we’ll never know. They won’t disclose it.

Vertex, which is its secure platform environment to enable faster GenAI model usage and app creation, is doing very well. API calls here were up 14x in 6 months.

The mega-cap added the ability for YouTube Shorts creators to upload up to 3 minute videos, which should boost the already 70% of channels posting Shorts. Monetization for YouTube Shorts vs. other ad formats “continues to close the gap.” The #1 streamer in the world (Netflix #2) is also testing a new user interface that groups content and channels more similarly to cable. This should help create consumer comfort and could speed up cord cutting. All in all, watch time growth was called “robust.”

Fantastic quarter. As I jokingly said on Twitter, these results look way too good for a company that is ~supposedly~ being disrupted every 5 seconds. In reality, the search business is performing very well, YouTube continues to briskly compound and cloud was even stronger. What else is there to say besides Search King is still King? I have no interest in trimming following this report. I’d likely add if Mr. Market throws another broad-based fit.