The number is almost too large to hold in your head.

The four biggest hyperscalers — Alphabet, Amazon, Meta, and Microsoft — have collectively committed more than $700 billion in capital expenditures for 2026, the majority of it flowing into AI infrastructure. That is up from roughly $400 billion in 2025, and Wall Street’s estimates were too low both of the prior two years running. If history repeats, the real number could be even higher by the time year-end rolls around.

And yet, the market is no longer simply cheering.

Why Every Investment Committee Is Talking About This

A year ago, announcing a large AI capex plan was essentially a catalyst. Stocks surged on the news. Investors rewarded ambition. The logic was simple: whoever builds the most compute today wins the AI era tomorrow.

That logic has quietly started to crack.

When Amazon revealed $200 billion in planned capex for 2026 — up 53% from the prior year and well above Wall Street’s estimate of $145 billion — the stock dropped roughly 8 to 10% on the announcement. CEO Andy Jassy argued that AWS capacity is being monetized as quickly as it is installed, pointing to a $142 billion annualized revenue run rate with growth accelerating to 24% year-over-year. Investors heard the defense and sold anyway.

What changed? Patience. Specifically, the dwindling of it.

Institutional investors are now drawing a hard line between two kinds of AI spenders. Companies that can show a clear, direct link between their capital expenditures and growing revenues are being rewarded. Companies that are spending heavily while operating earnings growth is under pressure — particularly those funding the buildout with debt rather than free cash flow — are facing a much more skeptical audience. Goldman Sachs Research noted that stock price correlation across the large AI hyperscalers has dropped from roughly 80% to around 20% since mid-2025. The group that once moved as one has fractured into winners and laggards.

The Bull Case

The structural argument for continued AI investment remains genuinely powerful, and dismissing it requires ignoring a lot of real evidence.

Goldman Sachs estimates that total AI-related infrastructure spending could reach $7.6 trillion through 2031. Morgan Stanley Research projects approximately $2.9 trillion in global data center construction through 2028 alone, with more than 80% of that spending still ahead. The International Energy Agency says data center electricity use rose 17% in 2025 — more than five times faster than global electricity demand overall — and projects global data center power consumption could double to 945 terawatt-hours by 2030.

The demand is not hypothetical. It is showing up in contracted revenue. Alphabet’s cloud backlog surged 55% sequentially to over $240 billion. AWS is growing at a three-year high. Hyperscalers are signing 10 to 20 year power purchase agreements with energy producers, giving the entire supply chain rare forward visibility. Analyst estimates for AI capex have been revised upward consistently for two years, and there is little in the current order pipeline that suggests a reversal.

The bull case, in short, is that this is one of the largest capital allocation cycles in corporate history — and it is still early.

The Bear Case

Here is where it gets genuinely uncomfortable for the companies doing the spending.

T. Rowe Price analysts point out that rising capital expenditure and input costs are increasingly testing hyperscalers’ ability to sustain aggressive spending as free cash flow comes under pressure. Only one of the four major hyperscalers — Microsoft — is expected to be able to cover its capex with cash generated from operations this year. The rest are turning to debt markets.

That changes the investment case meaningfully. These were, until recently, the most capital-light, free-cash-flow-generative businesses in the history of public markets. Now Big Tech is operating with higher capital intensity than telecom companies, whose investment ratios have been declining for years. US Big Tech capex intensity has climbed to roughly 23% of revenue — more than double the pre-ChatGPT level, according to Allianz Research.

And the returns question is not going away. Recent research suggests that businesses replacing workers with AI agents often fail to generate a measurable return on investment. Among S&P 500 companies, only 21% currently cite AI-related benefits to their operations, according to Morgan Stanley data. The monetization gap — between what is being spent and what is being earned — is the single largest source of investor anxiety heading into the second half of 2026.

What the Smartest Investors Are Actually Debating

Slight tangent, but it matters: the AI trade has already quietly moved in an important direction that most coverage has missed.

Goldman Sachs Prime Brokerage data showed hedge funds sold technology at a record pace heading into July, pushing Magnificent Seven positioning to its lowest point of the year. This was not a bearish macro call. It looked more like risk management — reducing concentrated single-stock exposure before a period of elevated volatility — rather than a conviction call that AI is over. Index and ETF exposure held steady while individual tech names were trimmed.

What this tells you is that the most sophisticated capital in the market has shifted from asking