Platformonomics TGIF #88: May 9, 2025

Amazon CAPEX – Q1 2025

$25.05B (including a mere $54M in finance leases) for the quarter, down 11% from Q4, up 68% from a year ago. Remember AWS spending is ~60% of this.

Commentary:

Plan to bring on an increasing amount of capacity in the back half of the year. Now turning to our cash CapEx which was $24,300,000,000 in Q1. The majority of this spend is to support the growing need for technology infrastructure. This primarily relates to AWS as we invest to support demand for our AI services and increasingly in custom silicon like Tranium as well as tech infrastructure to support our North America and international segments.

We’re also investing in our fulfillment and transportation network to support future growth and improve delivery speeds and our cost structure.

Not much CAPEX discussion or commentary, as the focus was on how tariffs affect the retail business. Some rumors of pullbacks on spending, similar to Microsoft.

Google CAPEX – Q1 2025

$17.2B for the quarter, up 20% from Q4, up 43% from a year ago.

Commentary:

Google’s network is robust and resilient, supported by over two-million miles of fiber and thirty-three subsea cables.

With respect to CapEx, our reported CapEx in the first quarter was $17.2 billion, primarily reflecting investment in our technical infrastructure, with the largest component being investment in servers, followed by data centers, to support the growth of our business across Google Services, Google Cloud, and Google DeepMind.

We still expect to invest approximately $75 billion in CapEx this year. The expected CapEx investment level may fluctuate from quarter to quarter, due to the impact of changes in the timing of deliveries and construction schedules.

Recall I’ve stated on the Q4 call that we exited the year in Cloud specifically, with more customer demand than we had capacity. And that was the case this quarter as well.

Much banging of the depreciation drum:

In terms of expenses, first, as I mentioned on our previous earnings call, the significant increase in our investments in CapEx over the past few years will continue to put pressure on the P&L, primarily in the form of higher depreciation.

In the first quarter, we saw 31% year-on-year growth in depreciation from the increase in technical infrastructure assets placed in service. Given the increase in CapEx investments over the past few years, we expect the growth rate in depreciation to accelerate throughout 2025.

Meta CAPEX – Q1 2025

$13.69B for the quarter, down 9% from Q4, up 104% from a year ago.

Meta bumps its 2025 CAPEX guidance from the beginning of the year’s mere $60-65B to $64-72B. Huzzah!

Guidance:

We anticipate our full year 2025 capital expenditures, including principal payments on finance leases, will be in the range of
$64-72 billion, increased from our prior outlook of $60-65 billion. This updated outlook reflects additional data center investments to support our artificial intelligence efforts as well as an increase in the expected cost of infrastructure hardware. The majority of our capital expenditures in 2025 will continue to be directed to our core business.

The first way is by significantly scaling up our infrastructure footprint. Our capex growth this year is going toward both generative AI and core business needs, with the majority of overall capex supporting the core. We expect the significant infrastructure footprint we’re building will not only help us meet the demands of our business in the near-term, but also provide us an advantage in the quality and scale of AI services we can deliver. We continue to build this capacity in a way that grants us maximum flexibility in how and when we deploy it to ensure we have the agility to react to how the technology and industry develop in the coming years.

Meta is starting to offer cloud services to third party developers (which I dub “LlamaCloud”), beginning with an API and inference for Llama. We’ve wondered previously how this doesn’t inexorably put them on path to becoming a more general purpose cloud provider and having to court enterprise customers. They got a question about this on the analyst call but gave a non-answer.

Question: Your CapEx spend is now close to some hyperscalers with very big client bases. Just help us conceptualize the kind of ecosystem you’re building with your CapEx. I know you gave a lot of help on the intro, but maybe the ROI works without direct enterprise spend to drive revenues. How you’re thinking about that?

Microsoft CAPEX – Q1 2025

$21.4B for the quarter, down 5% from Q4, up 53% from a year ago.

It looks like Satya will, in fact, “be good” for his $80B in FY25 CAPEX. With one quarter to go, they’ve spent $64B and averaged over $21B each quarter. Even with the spike in economic and policy uncertainty, that target seems safe as most current spend has been in motion for years. Future years may be a very different story.

Commentary:

We processed over 100 trillion tokens this quarter, up 5X year-over-year – including a record 50 trillion tokens last month alone.

Capital expenditures including finance leases were $21.4 billion, slightly lower than expected due to normal variability from the timing of delivery of data center leases. Cash paid for P, P, and E was $16.7 billion. Roughly half of our cloud and AI related spend was on long-lived assets that will support monetization over the next 15 years and beyond. The remaining cloud and AI spend was primarily for servers, both CPUs and GPUs, to serve customers based on demand signals including our customer contracted backlog of $315 billion.

We expect Q4 capital expenditures to increase on a sequential basis. H2 capex in total remains unchanged from our January H2 guidance.

…while we continue to bring datacenter capacity online as planned, demand is growing a bit faster. Therefore, we now expect to have some AI capacity constraints beyond June.

So, as a reminder, our earlier comments on FY26 capital expenditures remain unchanged. We expect capex to grow, it will grow at a lower rate than FY25 and will include a greater mix of short-lived assets which are more directly correlated to revenue than long-lived assets

Extensive answers to a question about data center commitment changes and cancellations that have been in the press in recent months (unclear if this is material or just greater scrutiny on typical portfolio churn):

The reality is, we’ve always been making adjustments to build, lease, what pace we build, all through the last 10-15 years. It’s just that you all pay a lot more attention to what we do quarter over quarter nowadays.

Having said that, the key thing for us is to have our builds and lease be positioned for what is the workload growth of the future. That’s what you have to go and seek to. There’s a demand part to it. There is the shape of

the workload part to it, and there is a location part to it. You don’t want to be upside-down on having one big data center in one region, when you have a global demand footprint. You don’t want to be upside-down when the shape of demand changes, because, after all, with essentially pre-training plus test time compute, that’s a big change in terms of how you think about even what is training. Forget inferencing.

Fundamentally, given all of that, and then every time there’s great Moore’s Law, but remember, this is a compounding S-curve, which is, there’s Moore’s Law, there’s system software, there’s model architecture changes, there’s the app server efficiency. Given all of that, we just want to make sure we’re building, accounting for the latest and greatest information we have on all of that.

And that’s what you see reflected, and I feel very, very good about the pace. In fact, Amy just mentioned, we will be short power. And so, therefore, but it’s not power. It’s not a blanket statement. I need power in specific places so that we can either lease or build at the pace at which we want. And so, that’s the plan that we’re executing.

Maybe, Amy, you can add to that.

AMY HOOD: Yeah, maybe just to add a little bit to Satya’s comments, just a reminder, these are very long lead time decisions. From land to build to buildouts can be lead times of five to seven years, two to three years. We’re constantly in a balancing position as we watch demand curves and many of things Satya watched.

And the second part is just to maybe remind you, when Satya talks about being short power, he’s really talking about data center space. And so, we’ve continued through the second half to put things in place. In fact, we talked a little bit about pulling even some of that space to be ready earlier and being able to deliver that earlier to customers this quarter, which is really good work by the teams, as we continue to get more and more efficient at that process. And I look forward to being able to continue to do that in the future.

I did talk about in my comments, we had hoped to be in balance by the end of Q4. We did see some increased demand, as you saw through the quarter. We are going to be a little short, still, a little tight as we exit the year, but are encouraged by that.

Cloud Growth Rates – Q1 2025

Never Take a Dependency on Elon Musk: Chapter 150

Our latest installment of Elon hosing developers comes in a missive from ReadWise:

I’m reaching out because I saw from our logs that you saved a tweet or thread to Readwise in the past 3 days, and likely hit an issue with the full post not being saved properly. I’m very sorry about the bug.

X (formerly Twitter) decided to change a set of endpoints over the weekend that our integration relied on, preventing us from accessing full threads or metadata (titles, author, etc).

Just wanted to let you know that over the course of Sunday we rewrote our integration and have fixed the issue!

Kudos to ReadWise (a great app) for cleaning up after Xitter breaking its sacred bond with developersAPI.

Filling Big Shoes: Warren Buffett Edition

I wrote my post on Warren Buffett’s retirement 12 years ago. Congratulations to him for taking his time getting around to it.

I’m long Berkshire, but need to consider my own advice here. The huge cash position makes me think they can still outperform for one more cycle (sold high, waiting to buy low).

The Opportunity Costs of 37Servers

I’ve stopped tracking the cloud repatriation “trend” because the only data point was 37Servers. But nice to see the tradeoffs coming into perspective.

BREAKING: Dunning-Kruger Renamed Dunning-Kruger-Trump