Meta’s earnings offer an early test of whether AI spending still gets a pass
Big Tech investors are about to get a concentrated read on AI investment plans—and the market’s tolerance for them. Meta reports third-quarter results on Oct. 21, followed by Microsoft and Alphabet on Oct. 28, Amazon on Oct. 29, and Apple on Nov. 2. Then come the U.S. midterm elections, with data center development now a political issue. My colleague Eric Fry has marked Nov. 4 as a potential turning point for the AI trade.
That sequence matters because investors are starting to ask a harder question than how much companies will spend: can they show that spending is producing revenue and returns worth the cost? For three years, rising capital expenditure often signaled confidence and helped lift shares. That response is no longer automatic.
I still see AI infrastructure as a major investment opportunity, particularly for companies addressing power, cooling, wiring and connectivity constraints. But this earnings window is a useful stress test. Fry and I agree on the pressures facing the sector; we differ on where investors should look next.
More spending now comes with more scrutiny
Opposition to data center projects is growing. An estimated $130 billion in U.S. projects was delayed or blocked in the first quarter of 2026. New York enacted the country’s first statewide moratorium on new hyperscale facilities, while more than 100 local moratoriums are under consideration. The issue has drawn opposition from both parties. A House vote on who should pay for the power infrastructure data centers need passed 417–3, a strikingly lopsided result.
Credit investors are also judging the AI buildout more selectively. Debt linked to some of the sector’s most leveraged companies repriced sharply this month, and yields on some AI-related bonds moved well into double digits. That suggests lenders are weighing companies’ balance sheets even when stock investors haven’t made the same distinctions.
That leaves executives with a tougher earnings call. Spending less could look like retreating from AI. Spending more could raise concerns about debt, political resistance and new capacity chasing compute prices that are falling. The market’s response to capex guidance may tell us as much as the guidance itself.
Infrastructure costs rise as compute gets cheaper
Fry’s argument turns on a squeeze from both sides. Building AI infrastructure is getting more expensive as developers shoulder more power costs and face fewer easy sites with affordable land, available grid connections and supportive neighbors. Interconnection queues can stretch for a decade. A 23,000-GPU campus outside London stalled because electricity may not be available until the 2030s.
At the same time, the price of the computing infrastructure sells can fall. More efficient chips and leaner models can reduce the cost of running AI. In January 2025, DeepSeek showed how much AI could be done with cheaper technology; Nvidia lost $589 billion in market value in one session, while the Dow ended higher. The episode suggested that money could move between parts of the market rather than leave it altogether.
Rising build costs and falling compute prices make returns harder to protect, especially for companies carrying substantial debt. The data doesn’t establish which businesses will struggle most, but it does explain why investors are looking beyond headline spending totals.
Fry expects money to shift from builders to users
Fry calls his forecast “The Separation”: beginning Nov. 4, he expects investment to rotate away from some popular AI infrastructure stocks and toward companies using cheaper AI to improve their businesses. He says some infrastructure stocks could fall 50% or more. That’s a forceful prediction, not a settled outcome; the midterm result, by his account, won’t change his expected start date.
He points to an earlier call from the dot-com era. In December 2000, Fry warned investors away from Corning (GLW), then a prominent fiber-optic company, before it fell 87%. That same month, he recommended Humana (HUM), a health insurer using the internet to run its business more effectively. It later gained more than 4,000%. One historical example can’t prove that the same pattern will repeat, but it captures the distinction Fry is making: supplying a technology and putting it to work can produce very different investment results.
Why I remain constructive on infrastructure
I see a stronger demand commitment than in the dot-com buildout. Anthropic has reportedly committed to $518 billion in compute, with roughly 80% noncancelable. OpenAI’s revenue run rate is nearing $70 billion. Suppliers to the infrastructure buildout also report large order books: $53.4 billion at Quanta (PWR), $17.1 billion at EMCOR (EME), and $15 billion at Vertiv (VRT).
Switchgear maker Powell (POWL) has a record $2.4 billion backlog and says its fiscal 2027 is essentially booked. Modine (MOD) signed a $4 billion cooling supply agreement that included a $165 million upfront deposit. Those orders support my view that parts of the infrastructure supply chain still have substantial demand. They don’t, by themselves, resolve questions about project costs, financing or eventual returns.
Here’s the thing: Fry and I agree that AI infrastructure faces higher costs, political pressure and closer scrutiny. We disagree about what that means for the investment opportunity. The next two weeks should offer evidence, especially in how shares respond to capex plans.
What to watch in the reports
- Whether investors reward or penalize companies that raise capital spending.
- Whether spending is funded from operating cash flow or borrowed money, as credit markets are distinguishing between the two.
- How much announced capacity has power available. A project announcement and a powered facility are not the same thing.
By Nov. 4, investors will have heard from five major technology companies and voted in the midterms. Fry has outlined his case in a free presentation, which he says requires no signup. He discusses which companies he considers most vulnerable, where he expects capital to go, and identifies what he calls his top AI Applier pick by name and ticker.
Holding through this earnings run and the election is a position in itself. The useful question is whether the evidence supports your view of who benefits from AI spending—and what the market does when companies disclose the bill.
About the analyst
Luke Lango is editor of Hypergrowth Investing. His work focuses on finding early investment opportunities in fast-growing industries and major long-term trends. Learn more about Luke.

This page was printed from InvestorPlace Media: “Meta Is About to Show Whether Bigger AI Capex Still Wins”. ©2026 InvestorPlace Media, LLC.