Morgan Stanley has again raised its forecast for the capital expenditures of the United States’ five largest hyperscalers, showing that the AI infrastructure investment boom is not only not cooling down, but is instead accelerating further. Finance commentator Holger Zschaepitz cited Morgan Stanley research indicating that Amazon, Alphabet, Meta, Microsoft, and Oracle’s 2026 capex estimates have been revised upward from the original $765 billion to approximately $805 billion; the 2027 estimate has also been sharply raised from $951 billion to above $1.1 trillion.
The five hyperscalers’ capital expenditure equals the total of all non-tech companies in the S&P 500
The five hyperscalers’ capex totaled approximately $261 billion in 2024, rising to $449 billion in 2025, while Morgan Stanley currently estimates 2026 will reach $805 billion and 2027 will increase further to $1.116 trillion. In other words, AI and cloud infrastructure investment in 2026 will be nearly twice that of 2025, and about three times that of 2024.
Holger Zschaepitz specifically pointed out that, by scale comparison, the capital expenditure of these five companies alone in 2026 is roughly equivalent to the combined total capex of all non-tech companies in the S&P 500 in 2025. This means AI capex is no longer just an investment cycle belonging to technology companies; it has become a macro variable large enough to reshape the structure of corporate investment in the United States.
Amazon, Google, Meta, Microsoft, Oracle all raise capex across the board
According to the chart, Morgan Stanley currently estimates that the five hyperscalers’ capex in 2027 will reach $1.116 trillion, including Microsoft at $276 billion, Alphabet at $299 billion, Amazon at $268 billion, Meta at $165 billion, and Oracle at $108 billion.
Looking at compound annual growth rates from 2024 to 2027, Oracle has the highest growth rate at 116%; Alphabet at 69%; Meta at 59%; Microsoft at 54%; and Amazon at 48%. This also reflects how AI infrastructure spending has expanded from initial purchases by cloud giants and GPUs into data centers, compute leasing, enterprise AI workloads, and token generation infrastructure.
David Sacks: This year AI capex could contribute 2.5% to GDP growth; over 3% next year
David Sacks further interpreted Morgan Stanley’s data from a macroeconomic perspective. He said he has previously argued multiple times that AI capex this year would provide a tailwind of about two percentage points to U.S. GDP growth; but according to Morgan Stanley’s latest report, the actual figure could be stronger—about 2.5% this year, and even above 3% next year.
Sacks believes this number may even still be underestimating AI’s impact on the economy for two reasons. First, Morgan Stanley’s estimate covers only the five hyperscalers and has not accounted for AI startups, nor other companies’ investment in AI. Second, capex is only used to build the so-called “token factories,” the infrastructure that produces AI tokens and provides inference and computing capability, but it has not yet included the economic activity generated by these tokens.
Data center, GPUs, cloud, and power infrastructure are only the first layer of investment; the real multiplier effect will come from using AI tokens to generate code, build customized software, improve enterprise efficiency, and spread into overall productivity. That is why he believes the ROI on AI capex is likely to be far higher than capex itself, which is precisely the reason companies keep expanding their investment.
“AI accounts for 75% of Q1 GDP growth”: Is AI investment becoming the main engine of the U.S. economy?
David Sacks also said that in the first quarter this year, AI already accounted for 75% of U.S. GDP growth, and he believes this trend may continue. He stressed that technological leadership has always been the United States’ biggest advantage, and AI is pushing the U.S. economy forward. His conclusion is also quite direct: although polls might show that AI is not popular, economic growth is popular; at this stage, trying to stop AI development is essentially the same as bringing the U.S. economy to a halt.
This also highlights the two-sided nature of the AI investment debate. On the one hand, the market is concerned that hyperscalers’ capex spending may expand too quickly, potentially squeezing free cash flow and even forming another round of a technology bubble. On the other hand, supporters argue that AI infrastructure is becoming the next generation of the economy’s underlying layer; while short-term spending is large, the productivity improvements and software generation capabilities it delivers long term may far outweigh traditional data center investment.
AI capex is no longer just a tech stock story—it’s a U.S. economy story
Judging from Morgan Stanley’s upward revisions, the market’s understanding of AI capital expenditures is changing. This is no longer just the story of “a few tech giants buying more GPUs,” but rather a long-cycle investment landscape that includes data centers, power, advanced packaging, semiconductors, cloud platforms, enterprise software, and a reassessment of productivity.
Last week, major U.S. tech giants all updated their earnings reports—Apple, Alphabet, Microsoft, Meta, and Amazon sent a very clear shared signal: AI demand remains strong, most core businesses are performing above market expectations, and there has been no obvious slowdown in cloud, advertising, device, and services revenue.
But the market reaction also shows that investors are no longer satisfied with the narrative that “AI will bring growth.” They are now scrutinizing more rigorously whether accelerating AI capital spending will squeeze free cash flow too fast, whether cloud and advertising businesses can support higher future depreciation pressure, and whether memory, advanced process nodes, data center power, and supply-chain bottlenecks could become key variables for next quarter’s guidance.
In other words, the AI war among the five tech giants has entered the next phase: in the first phase, it was about who is willing to invest and who can secure GPUs and data center capacity first; in the second phase, it will be about who can turn these investments into revenue, gross margin, and free cash flow.
AI demand still exists, and the scale may still be expanding; but from now on, investors will care more about AI’s unit economics. Whoever can prove that every dollar of AI capex ultimately turns into higher cloud revenue, stronger advertising conversion rates, higher service gross margins, or more stable cash flow will become the true winner of the next AI cycle.
If David Sacks’ view holds, AI capex is not a bubble, but a new round of GDP growth and a capital expenditure cycle for U.S. technology dominance. However, if demand lags behind supply, or if the monetization pace of AI applications fails to keep up with the expansion of infrastructure, the market may also re-examine the logic behind this trillion-dollar AI buildout.
This article says AI accounts for 75% of U.S. Q1 GDP growth, and by 2027 the capital expenditures of the five giants may exceed $1.1 trillion. First appeared on Lian News ABMedia.
