Nvidia's data-center revenue hit $75.2 billion last quarter, up 92% year over year, while Broadcom's AI semiconductor revenue reached $10.8 billion. Apollo Global Management now estimates total AI capital spending will exceed $2.7 trillion between 2025 and 2029, a figure that would push AI-related investment to nearly 3% of US GDP by the end of the decade.
The Hyperscaler Spending Spree
The big cloud providers are pouring money into AI infrastructure. Microsoft Azure grew 40% in the latest quarter, Amazon AWS grew 28%, and Alphabet's Google Cloud grew 48% with $240 billion in future orders. That's a lot of committed spending. Hyperscaler capex is expected to run at roughly 3% of GDP annually from 2027 to 2029, up from 0.3% in 2019 and 1.4% in 2025. The shift is dramatic.
Apollo's estimate of $2.7 trillion over five years isn't just a number. It represents a structural change in how the US economy allocates capital. AI spending as a share of GDP is projected to rise from 0.3% in 2019 to nearly 3% by the end of the decade. That's a tenfold increase in a decade, and it's happening faster than most forecasters predicted.
Stock Market Winners and Losers
Investors have rewarded the compute side of the trade. Compute stocks rose 64.7% over the full window but slipped 3.4% in the latest month. Apps and software stocks jumped 22.8%. Micron was the top performer in 2026, up more than 200%, driven by high-bandwidth memory demand exceeding supply. Broadcom rose about 20%, helped by custom AI chips for cloud clients. Nvidia gained only about 17% despite strong sales growth. The gap between the leaders and the laggards is telling.
Micron's surge reflects a specific bottleneck: high-bandwidth memory is in short supply, and the company is cashing in. Broadcom's custom chip business is growing as cloud giants look for alternatives to Nvidia's off-the-shelf GPUs. Nvidia's modest stock gain, relative to its revenue growth, suggests investors are already pricing in competition and margin pressure.
Power and Infrastructure Catch-Up
The power and infrastructure group is a different story. Vertiv raised its outlook after sales grew 24%. Eaton saw orders up 13% and backlog up 28%. Quanta connects data centers to the grid. Yet the group was up only 2% in the latest month, despite strong order books. The market seems to be waiting for the spending to translate into earnings.
These companies are the physical backbone of the AI buildout. Data centers need cooling, power distribution, and grid connections. The order books are full, but the stock performance lags. That could change if the spending accelerates, or it could mean the market thinks the infrastructure buildout is already priced in.
Chip Competition and Cash Flow
Some large tech firms are designing their own chips with Broadcom, eroding Nvidia's dominance. That's a real shift. Meanwhile, capex is climbing faster than the cash these companies generate, according to Apollo Global Management. That's a red flag for some investors. The question now is whether the cash flows can catch up to the spending, or whether the boom runs ahead of itself.
Apollo's warning isn't a prediction of a crash. It's a note of caution that the spending pace is unsustainable if revenue doesn't follow. The next few quarters will show whether the hyperscalers can turn their AI investments into profits, or whether the capex cycle peaks before the returns materialize.
