The $725 Billion Question: Inside Big Tech's AI Spending Boom

Quick Summary

The four largest US tech companies plan to spend roughly $725 billion on capital expenditure in 2026, up about 77% from $410 billion in 2025 — most of it on AI data centers.

This spending has propped up the whole market, but it is now colliding with a memory-chip shortage, a hawkish Fed, and rising "bubble" fears.

One key fact: the bottleneck has shifted from AI chips to the memory that feeds them — high-bandwidth memory is reported sold out through most of 2027.

Educational note: this is an explainer on how capex, chips, and share prices connect. It is not a recommendation to buy or sell anything.

Big Tech's AI capex is the single largest wave of corporate investment in modern history, and in 2026 it has become the hinge the entire stock market swings on. Four companies — Amazon, Alphabet, Meta and Microsoft — plan to spend around $725 billion this year building the infrastructure behind artificial intelligence, and whether that spending pays off is now the most important open question in markets. This piece walks through how the number got so large, what it is running into right now, and how share prices are actually reacting.

The past: from rounding error to record

A decade ago, capital expenditure at software companies was almost an afterthought. Building data centers was expensive, but nothing like today's scale.

AI changed the arithmetic. As of 2026, combined hyperscaler capex has roughly tripled the $226 billion spent in 2024, and last year's record $410 billion already looked enormous before this year's numbers landed. The driving logic is simple and slightly frightening: being short on computing power is the one mistake none of these companies believe they can afford. So they spend, because the competitor who under-invests risks losing the AI race permanently.

The present: one number holding up the market

By early 2026, planned capex had stopped being a line item and become a sentiment gauge for the entire market.

When Microsoft, Alphabet, Meta and Amazon reported earnings, investors were less focused on profits than on one figure: would the capex guidance keep rising? AI optimism had single-handedly kept markets afloat through the Middle East conflict, surging oil prices and stagflation fears, so a rising number meant the story lived on. It kept rising. Amazon guided to roughly $200 billion, Alphabet to $185 billion, Meta to around $125 billion, and Microsoft to about $120 billion — with Microsoft's finance chief attributing $25 billion of its budget to rising memory-chip and component costs alone.

Sources: Tom's Hardware and Statista, citing Q1 2026 earnings, April 2026.

The bottleneck nobody priced in: memory

Here is the twist that has reshaped the 2026 story. The scarce resource is no longer the AI chip itself — it is the memory that feeds it.

High-bandwidth memory, the specialised memory that advanced AI chips depend on, is reported sold out through most of 2027. Memory has gone from a boring, cyclical commodity to the single tightest link in the chain, and it now consumes a reported 30% of hyperscaler data-center spending, roughly four times its 2023 share. This is why memory maker Micron soared around 197% in 2026, and why, when Alphabet raised its capex guidance in July, investors sold Nvidia and bought memory names like Micron and SK Hynix instead — rotating toward the part of the chain that was actually scarce.

Sources: CNBC and ABC News, July 2026; Fortune, June 2026.

How share prices are actually reacting

The market's response in 2026 has been anything but a simple cheer. It has been sharp, two-sided, and increasingly nervous.

Spending more no longer guarantees a higher stock. Despite beating on revenue, Amazon's stock fell almost 6% on a day its capex plan spooked cash-flow-focused investors, and Microsoft spent much of the year as the weakest of the group. When capex rises, free cash flow falls — one manager put it bluntly: pour this much into AI and it reduces your free cash flow. In July, a semiconductor selloff wiped out more than a trillion dollars in market value, with the 25-largest-chipmaker index down roughly 9.5% in a month, as Wall Street openly questioned whether the returns would ever justify the outlay.

The bear and bull cases are worth stating fairly, because both are held by serious people:

• The bears point to dot-com-era valuations, a Bubble Risk Indicator from a BofA strategist hitting levels not seen since June 2000, and Moody's warning that the ultimate return on this spending "is unclear."

• The bulls counter that these companies are profitable, the demand is real, second-quarter 2026 chip earnings were set to grow around 131%, and Nvidia's forward valuation is actually well below its own five-year average.

This is exactly where a testable approach matters more than a loud opinion. "AI is a bubble" and "AI is the future" are both stories. How a specific stock or index has actually behaved around capex announcements is a checkable pattern. On a platform like HeyTheo, where the rule behind a signal is visible and can be tested against history, you can examine that behaviour yourself rather than pick a side of a shouting match.

The geopolitics: chips, rates, and a second AI ecosystem

Three outside forces now sit on top of the capex story, and each one can move it.

The first is interest rates. Under a more hawkish Federal Reserve, higher rates raise the discount applied to future earnings, which weighs hardest on exactly these long-duration, capital-hungry growth stocks. The second is the supply chain: the same AI data-center demand driving shortages has pushed up prices for ordinary chips in phones and laptops, a cost that ripples far beyond tech.

The third, and most structural, is China. Barred from buying the most advanced US chips, China has accelerated a domestic build-out, and it is working faster than many expected — one Morgan Stanley estimate put China's chip self-sufficiency at over 40% in 2026, up from around 20% in 2023, with a path toward roughly 85% by 2030. The blockbuster $8.6 billion IPO of Chinese memory maker CXMT, now the world's fourth-largest DRAM maker, showed how quickly a state-backed challenger can scale. The result is not one global AI market but two increasingly separate ecosystems — a dynamic every multinational and investor now has to price in.

Sources: Morgan Stanley via OfficeChai, May 2026; Business Day and Reuters, July 2026.

The future: three ways this resolves

Nobody knows how the capex wave ends, but the honest range of outcomes is narrower than the noise suggests, and it comes down to one thing: does the revenue show up?

If AI revenue grows into the spending, today's capex looks visionary and the infrastructure becomes as foundational as the early internet backbone. If revenue lags while capex keeps climbing — analysts already see the number topping $1 trillion in 2027 — the market's patience, which is conditional, breaks, and the multiple compression is brutal. The most likely path is messier than either: enormous long-term value, delivered unevenly, with sharp corrections along the way as expectations and reality repeatedly get out of sync. The build-out is real. The timing of its payoff is not knowable in advance, and anyone who claims otherwise is selling a forecast.

Key takeaway

Big Tech's $725 billion AI capex is both the engine holding the market up and its single biggest risk, because so much now depends on spending that has not yet proven its return. The sensible response is not to decide whether "AI is a bubble" in the abstract, but to watch the checkable signals: the direction of capex guidance, where the shortages actually are, how rates move, and how specific stocks react when the numbers land. Those are things you can track rather than guess at. HeyTheo lets you build a basket of these names and test how a rule would have behaved against them through history, so you can form your own view from evidence rather than headlines. As always, remember that past performance and historical patterns never guarantee future results.

Disclaimer: This article is published by HeyTheo Research. HeyTheo (app.heytheo.io) is a stock research and signal-generation platform. It is not a broker-dealer and not a registered investment adviser. This content is for informational and educational purposes only. It is not investment advice, not a recommendation to buy or sell any security, and not an offer or solicitation. Any companies named are referenced for illustration only and are not recommendations. Any technical conditions or historical patterns described are observations, not predictions. Past performance does not indicate future results. Investing involves risk, including possible loss of principal. HeyTheo does not execute trades; any transaction happens through your own broker. Consider your own circumstances and consult a qualified financial professional before making any investment decision.