Platformonomics TGIF #95: August 1, 2025

News

The four hyperCAPEX companies spent $96.78 billion on CAPEX, up 25% from Q1 and up 65% from a year ago. Nearly 12 digits. In a quarter.

The usual caveat: this is overwhelmingly cloud/AI spend except for Amazon which also spends on warehouses, trucks, satellite constellations, etc. AWS was 61% of Amazon’s CAPEX spend in calendar 2024 (they don’t break out quarterly), so the AWS Q2 spend was probably on par with Google and Microsoft ($20B+).

UPDATE: Amazon is now breaking out CAPEX by segment on a quarterly basis. AWS spent $16.043B in Q2, so only 48% of company CAPEX spend. That is down 22% from Q1, up 25% from a year ago. 🚨🚨🚨

The companies remain supply-constrained, and are starting to highlight ballooning depreciation costs on the horizon, which will impact earnings. And the useful life of cutting edge GPUs remains unknown.

Amazon-wide CAPEX was $33.32 billion, up 32% for the quarter, up 85% for the year. Amazon gave no specific breakdown of their CAPEX other than saying it was “primarily” for AWS, so we’ll assume AWS CAPEX was ~60% of the total.

UPDATE: Amazon is now breaking out CAPEX by segment on a quarterly basis. AWS spent $16.043B in Q2, so only 48% of company CAPEX spend. That is down 22% from Q1, up 25% from a year ago. 🚨🚨🚨

Despite beating on both revenue and earnings, the market didn’t like Amazon’s guidance and/or AI positioning. “It is early days…” blah blah blah “…it is early days” didn’t compel. This the first time I can recall that the FinTwit mob going after CEO and AWS AI product manager Andy Jassy. Time for a Polymarket bet on whether Bezos returns as CEO?

CAPEX:

Now turning to our cash CapEx, which was $31,400,000,000 in q two.

AWS continues to be the primary driver 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. Additionally, we continue to invest in our fulfillment and transportation network to support growth of the business, improve delivery speeds, and lower our cost to serve by investing in same day delivery facilities as well as robotics and automation.

Supply constraints:

You know, as I mentioned, well, we have more demand than we have capacity at this point. And I think that and and you see, you know, some of the constraints, and they kind of exist in multiple places. The single biggest constraint is power. But I you know, you also see constraints off and on with chips and then some of the components that, you know, once you have the chips to actually make the servers, you know, you there are you know, sometimes you have new generations of chips that are a little bit later than they’re supposed to be and sometimes you get the chips and, you know, the yield you get in making servers isn’t what you what you expect when you get to ramp.

Guidance:

We expect Q2 CapEx to be reasonably representative of our quarterly capital investment rate for the back half of this year.

Depreciation:

AWS margins also saw headwinds from higher depreciation expense as well as unfavorable impacts from year over year fluctuations in foreign exchange rates. The depreciation expense is a result of our growing investments in capital expenditures in AWS. As we’ve said in the past, we expect AWS operating margins to fluctuate over time, driven in part by the level of investments we are making at any point in time. We will continue to invest more capital in chips, data centers, and power to pursue this unusually large opportunity that we have in generative AI.

“Early Days”:

Question: So there is a Wall Street Finance person narrative right now that AWS is falling behind in GenAI with concerns about share loss to peers, etcetera. Can you just sort of address that? What what is your rebuttal to that?

So on on the first one around AI, the first thing I would say is that I think it is so early right now in AI.

<six paragraphs of platitudes>

I still think it’s very early days in AI and in terms of adoption.

Such a missed opportunity to let us know exactly how many steps we now are into the AI race.

$22.45 billion, up 31% from Q1, up 70% from a year ago.

Google is ramping CAPEX nicely, but it has taken awhile (and Google had skimped on CAPEX for a few years before the AI boom).

Data centers:

We operate the leading global network of AI-optimized data centers and cloud regions.

CAPEX:

With respect to CapEx in the second quarter, our CapEx was $22.4 billion. The vast majority of our CapEx was invested in technical infrastructure, with approximately two-thirds of investments in servers and one-third in data centers and networking equipment.

Supply constraints:

While we have been working hard to increase capacity and have improved the pace of server deployments, we expect to remain in a tight demand-supply environment going into 2026.

As Sundar mentioned earlier, this is not the type of investment that’s a light switch. It takes time to make this investment. So what you’re seeing now is investments we made some time ago that’s now translating to additional capacity coming online, but more of that towards the back end of the year.

Guidance:

Moving to investments, given the strong demand for our Cloud products and services, we now expect to invest approximately $85 billion in CapEx in 2025, up from a previous estimate of $75 billion.

Our updated outlook reflects additional investment in servers, the timing of delivery of servers and an acceleration in the pace of data center construction, primarily to meet Cloud customer demand.

Looking out to 2026, we expect a further increase in CapEx due to the demand we’re seeing from customers as well as growth opportunities across the company.

Depreciation:

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 second quarter, depreciation increased $1.3 billion year-over-year to $5 billion, reflecting a growth rate of 35%. Given the recent increase in CapEx investments, we expect the growth rate in depreciation to accelerate further in Q3.

$17.01 billion, up 25% from Q1, up 100% from a year ago.

Despite being the runt of the hyperCAPEX litter, Meta is currently by far the most entertaining of the bunch to watch, as we try to match their ambitions to their capabilities.

No Big Tech company oscillates more in perception between future dominance and doomed roadkill than Meta. They’re already into their second AI cycle, as they take a mulligan and do a complete restart.

So many interesting paradoxes. If you throw enough money around on a restart, maybe no one will ask why the last effort went awry? Why does the company need to create a new organization “unencumbered by the bureaucracy of the $1.8tn company“? Personal superintelligence sounds great, but why would anyone trust them to run it? (Apple so much better positioned, but, alas, execution issues). Putting datacenters in tents? In hurricane country? As the hyperCAPEX runt, how exactly does Meta plan to “have the largest compute fleet of any company“? (don’t make me publish the cumulative CAPEX chart yet again). And where will they get the power?

CAPEX:

Capital expenditures, including principal payments on finance leases, were $17.0 billion, driven by investments in servers, data centers and network infrastructure.

Guidance:

Turning now to the capex outlook. We currently expect 2025 capital expenditures, including principal payments on finance leases, to be in the range of $66-72 billion, narrowed from our prior outlook of $64-72 billion and up approximately $30 billion year-over-year at the mid-point. While the infrastructure planning process remains highly dynamic, we currently expect another year of similarly significant capex dollar growth in 2026 as we continue aggressively pursuing opportunities to bring additional capacity online to meet the needs of our AI efforts and business operations.

On the CapEx side, the big driver of our increased CapEx in ‘26 will be scaling GenAI  capacity as we build out training capacity that’s going to drive higher spend across servers, networking, data centers next year.

So at present, we’re not really thinking about external use cases on the infrastructure, but I’d say it’s a good question.

Depreciation (may be optimistic on deprec schedules – CHAOS)

That’s driven primarily by a sharp acceleration in depreciation expense growth in 2026, largely driven by recognizing incremental depreciation from assets that we purchased and placed in service in ‘26 as well as from infrastructure deployed through 2025 that we’ll recognize a full year of depreciation next year.

While we’re still very early in planning for next year, there are a few factors we expect will provide meaningful upward pressure on our 2026 total expense growth rate. The largest single driver of growth will be infrastructure costs, driven by a sharp acceleration in depreciation expense growth and higher operating costs as we continue to scale up our infrastructure fleet

We also expect a greater mix of our CapEx to be in shorter-lived assets in 2025 and ‘26 than it has been in prior years.

Financing CAPEX:

…on your second question about how we expect to finance the growing CapEx next year. We certainly expect that we will finance some large share of that ourselves, but we’re also exploring ways to work with financial partners to codevelop data centers.

We don’t have any finalized transactions to announce, but we generally believe that there will be models here that will attract significant external financing to support large-scale data center projects that are developed using our ability to build world-class infrastructure while providing us with flexibility should our infrastructure requirements change over time. So we are exploring many different paths

$24.2 billion for the quarter, up 13% from Q1, up 27% from a year ago. Satya “I’m good for my $80 billion” Nadella in fact spent $88.2 billion in the promised fiscal year.

Data centers:

We opened new DCs across six continents, and now have over 400 datacenters across 70 regions, more than any other cloud provider.

We stood up more than two gigawatts of new capacity over the past 12 months alone.

And we continue to scale our owned datacenter capacity faster than any other competitor.

Demand:

And, while we brought additional datacenter capacity online this quarter, demand remains higher than supply.

CAPEX:

Capital expenditures were $24.2 billion, including $6.5 billion of finance leases where we recognize the full value at the time of lease commencement. Cash paid for P, P, and E, was $17.1 billion. The difference is primarily due to finance leases. More than half of our spend was on long-lived assets that will support monetization over the next 15 years and beyond. The remaining spend was primarily for servers, both CPUs and GPUs, and driven by strong demand signals.

Guidance:

Capital expenditure growth, as we shared last quarter, will moderate compared to FY25 with a greater mix of short-lived assets. Due to the timing of delivery of additional capacity in H1, including large finance lease sites, we expect growth rates in H1 will be higher than in H2.

We expect Q1 capital expenditures to be over $30 billion driven by the continued strong demand signals we see. As a reminder, there can be quarterly spend variability from cloud infrastructure buildouts and the timing of delivery of finance leases.

In Azure, we expect Q1 revenue growth of approximately 37% in constant currency driven by strong demand for our portfolio of services on a significant base. Even as we continue bringing more datacenter capacity online, we currently expect to remain capacity constrained through the first half of our fiscal year.

Google and Microsoft continue to take share from AWS, but #3 Google isn’t catching up to #2 Microsoft. At what point does Google kick Workplace out of the Google Cloud bucket? (it only grows with price increases).

Who could possibly have foreseen this?

Stargate is turning into a party round of both cash and electricity. Raising in increments of billions and megawatts is the long road to trillions and gigawatts.

OpenAI has raised $8.3 billion at a $300 billion valuation, months ahead of schedule, as part of its plan to secure $40 billion in funding this year

The site, which will be jointly owned by the two European companies, has the initial goal of delivering 230 megawatts of capacity with plans to add additional 290 megawatts.

I will yet get around to writing up Oracle’s looming cash crunch. Not only do they struggle to build data centers, they can’t afford to build data centers.

Historically an adept fundraiser, thin-skinned fractional CEO and diner operator Elon Musk is having to raise debt and grab money from SpaceX and maybe Tesla just to sustain a billion dollar (!) a month spending rate at xAI. Never mind his scale, distribution, and credibility issues.

Allow me to fix this:

Millions of Americans across the eastern U.S. are seeing their monthly electric bills spike, and many of them have no idea why. But there’s a culprit: idiot policymakers who have ignored and mismanaged our energy system for decades. Closing perfectly good nuclear plants, refusing to build transmission lines, etc.

Antitrust Incoherence: Hipster Antitrust Incoherence

This is niche, but hilarious. The “hipster antitrust” movement has been hugely influential in recent years, despite lacking any kind of an intellectual model for antitrust beyond shouting “break them up”. You can argue these hipsters helped drive Silicon Valley into Trump’s arms, amongst other consequences.

Lately the hipsters have been vociferous in their opposition to the Abundance book and movement, which takes the radical position that governments should try to make their policies actually work.

Abundance author Derek Thompson decided to dig into the hipsters’ criticisms:

The sharpest criticisms of the book Abundance have sometimes come from the antitrust movement. This group, mostly on the left, insists that the biggest problems in America typically come from monopolies and the corruption of big business.

I’m not an economist or a lawyer. I’m just a journalist. To the extent that I’m good at anything, it’s calling people on the phone and writing down what they say. So I reached out to the primary sources that Musharbash quotes throughout the piece.

What I found was astonishing. The economist Musharbash cites told me that his theories had been misapplied. The housing analysts quoted in the piece told me Musharbash distorted their points and reached dubious, or even flatly wrong, conclusions. The leading monopoly researcher I spoke to, whose work has been celebrated by the antitrust left, told me that the entire thrust of the article—and, by extension, much of the antitrust-housing philosophy—defied sophisticated antitrust analysis.

The essay you’re reading is very long. But I can sum it up in one sentence: The Musharbash essay on Dallas—like too much of the antitrust left’s work on housing—is filled with out-of-context quotes, overconfident assertions lacking evidence, and generally misguided claims.

The subsequent online discourse has been hilarious. And the hipsters have responded with insults and unsubstantiated claims, but no effort to refute the specifics.

On the other hand, given what they published this week, maybe we do want them to keep providing cutting edge advice to the world’s authoritarian regimes?

McKinsey discovering data centers is probably another sign of a top…