The great cash-returning machines of the digital age have quietly transformed into something else entirely — infrastructure builders staking their futures on artificial intelligence. In less than three years, the hyperscalers that once rewarded shareholders with buybacks and dividends have redirected nearly all of their operational surplus toward AI compute, data centers, and power capacity, with 2026 capital expenditures projected at $667 billion — a proportion of cash flow that surpasses even the feverish spending of the Dot-Com era. What is unfolding is not merely a shift in accounting prio
Big Tech's AI Bet: Capex Spending Hits Dot-Com Boom Levels as Free Cash Flow Dries Up
Nearly every dollar of internally generated cash is being reinvested back into infrastructure.
Mark
Why is this happening now? Why did hyperscalers suddenly abandon the buyback model?
Mimi
Because the competitive threat is existential. If one company pulls back on AI infrastructure spending to return cash to shareholders, a rival will outspend them and own the market. It's a prisoner's dilemma, except the stakes are the future of computing itself.
Mark
But $667 billion is an enormous number. How do we know this spending actually produces returns?
Mimi
We don't. That's the whole tension. These companies are making a strategic bet that AI will generate revenue and margins large enough to justify the investment. But they're doing it without proof yet.
Mark
The article mentions that buybacks fell from 43% to 16% of cash flows. That's a dramatic reversal. Are shareholders upset?
Mimi
Some are. But many institutional investors understand the logic. If you're a long-term holder, you'd rather own a company that dominates AI in five years than one that returned cash today and lost market position.
Mark
What happens if capex growth actually does decelerate in the second half of 2026?
Mimi
That's when the market gets to reassess. Right now, valuations are built on reinvestment narratives—the story that spending is necessary. Once spending slows, investors can finally see whether the revenue and earnings justify the capital intensity.
Mark
Is this sustainable? Can hyperscalers keep spending at this rate indefinitely?
Mimi
No. That's why Goldman expects deceleration. At some point, the infrastructure is built out enough. The question is whether that happens before or after the AI market actually materializes at scale.
Der Puls
- Big Tech's capital expenditure is surging toward $667 billion in 2026 — a 62% year-over-year leap that has outpaced analyst estimates by $127 billion in just a matter of months.
- The hyperscalers are now consuming 92% of their operating cash flows through reinvestment, a concentration of spending that exceeds even the reckless optimism of the Dot-Com Boom.
- Shareholder returns have been quietly gutted — buybacks have collapsed from 43% of cash flows in early 2023 to just 16% today, with some companies like Oracle and Alphabet turning to debt markets to fund the expansion.
- The disruption is rippling through investor frameworks, as traditional valuation models built on free cash flow and earnings visibility become temporarily unusable.
- A potential deceleration in capex growth during the second half of 2026 offers a narrow window of relief — a moment when analysts may finally be able to price these companies on earnings rather than on the promise of AI dominance.
The great cash-returning machines of the digital age have quietly transformed into something else entirely — infrastructure builders staking their futures on artificial intelligence. In less than three years, the hyperscalers that once rewarded shareholders with buybacks and dividends have redirected nearly all of their operational surplus toward AI compute, data centers, and power capacity, with 2026 capital expenditures projected at $667 billion — a proportion of cash flow that surpasses even the feverish spending of the Dot-Com era. What is unfolding is not merely a shift in accounting priorities, but a civilizational wager: that the companies which secure AI dominance today will define the economic landscape of tomorrow.
The era of Big Tech as a reliable cash-returning machine has ended. For years, Amazon, Microsoft, Google, Meta, and Apple operated as financial engines — generating enormous free cash flow, keeping capital spending lean, and returning the surplus to shareholders. That model has been dismantled in under three years.
Wall Street now expects these companies to spend $667 billion on capital expenditures in 2026, up $127 billion from estimates made just months ago. That represents 62% year-over-year growth. More telling than the number itself is what it reveals: nearly 92% of operating cash flows are being plowed back into AI-related infrastructure — compute, data centers, networking, and power capacity. Goldman Sachs analysts note this proportion exceeds what hyperscalers spent even during the Dot-Com Boom.
The reallocation is stark. In early 2023, roughly 43% of cash flows went toward stock buybacks. Today that figure has collapsed to 16%, with gross buybacks declining 15% year-over-year in 2025 alone. Some companies have even turned to debt markets to fund expansion. The message from management is unambiguous: securing AI dominance through scale has replaced optimizing returns to shareholders as the governing priority.
Goldman Sachs sees capex potentially climbing to $700–$725 billion under optimistic scenarios, though the growth rate is expected to begin decelerating in the second half of 2026. That slowdown, paradoxically, could benefit investors — offering visibility into where free cash flow bottoms out and allowing valuations to rest on earnings again rather than on reinvestment narratives.
At its core, this is a civilizational bet. Every dollar redirected from shareholder returns into AI infrastructure is a wager that artificial intelligence will eventually generate the revenue and margins to justify the extraordinary outlay. The companies that once epitomized capital-light efficiency are now staking their near-term returns on the promise of long-term dominance — and whether that trade-off ultimately pays off remains the defining question hanging over Big Tech.
The era of Big Tech as a cash machine is over. For years, the hyperscalers—Amazon, Microsoft, Google, Meta, Apple—operated like financial engines: they generated enormous free cash flow, kept their capital spending lean, and returned the surplus to shareholders through buybacks and dividends. That model has been dismantled in less than three years.
Wall Street now expects these companies to spend $667 billion on capital expenditures in 2026, up $127 billion from estimates made just a few months ago during fourth-quarter earnings season. That represents 62% year-over-year growth. But the number itself is less important than what it reveals about where the money is actually going. Nearly 92% of the cash these companies generate from their operations will be plowed back into infrastructure—primarily AI-related compute, data centers, networking, and power capacity. That proportion exceeds what hyperscalers spent during the Dot-Com Boom, according to Goldman Sachs analysts Ben Snider and Ryan Hammond.
The shift in capital allocation is stark and deliberate. In early 2023, hyperscalers directed roughly 43% of their cash flows toward stock buybacks. Today that figure has collapsed to 16%. Gross buybacks declined 15% year-over-year in 2025 alone. Some companies, including Oracle and Alphabet, have even tapped debt markets to help fund their expansion, a sign of how aggressively they are pursuing growth. The message from management is unambiguous: the priority is no longer optimizing returns to shareholders. It is securing dominance in artificial intelligence through sheer scale.
Goldman Sachs expects capex estimates to edge higher still, potentially reaching $700 billion to $725 billion under optimistic scenarios. But the growth rate itself is expected to begin decelerating in the second half of 2026. This slowdown, paradoxically, could be good news for investors. A deceleration in capex growth would provide visibility into where free cash flow bottoms out, allowing analysts and investors to value these companies on earnings again rather than on reinvestment narratives. It would also signal that the infrastructure complex benefiting from all this spending is beginning to mature.
What is really at stake here is a fundamental bet about the future. Every dollar hyperscalers redirect from shareholder returns into AI infrastructure is a wager that the revenue and margins generated by artificial intelligence will eventually justify the extraordinary investment. The companies that once epitomized capital-light efficiency are now betting their near-term returns on the promise of long-term dominance. For investors, the question is whether that trade-off will pay off—or whether Big Tech is simply spending money it may never recoup.
Bemerkenswerte Zitate
This dynamic has dramatically reduced hyperscaler free cash flows.— Goldman Sachs analysts
Instead of optimizing for shareholder distributions, management teams are optimizing for scale in AI.— Analysis in source material
Möchten Sie die ganze Geschichte?
Das Original lesen bei Benzinga ↗
Economy & Finance