Big Tech Will Spend $650 Billion on AI This Year - Bridgewater Says the Boom Just Got Dangerous

Alphabet, Amazon, Meta, and Microsoft will collectively spend roughly $650 billion on AI infrastructure in 2026, according to a new analysis from Bridgewater Associates published today. The figure represents a 67% jump from the $381 billion the four companies spent in 2025, and it comes with a warning: Bridgewater co-chief investment officer Greg Jensen says the AI investment cycle has entered a "more dangerous phase."

The numbers are staggering even by Big Tech standards. To put them in context, $650 billion exceeds the combined annual output of Hollywood and the global gaming industry. It is roughly 15-20% of the four companies' combined annual revenue. And it is now responsible, by Bridgewater's estimate, for roughly one-third of total U.S. economic growth.

TL;DR

Big Tech's combined AI capex hits ~$650B in 2026, up from $381B in 2025
Amazon leads at $200B; Alphabet $175-185B; Microsoft $145B; Meta $115-135B
Bridgewater warns the boom is in a "dangerous phase" - growing reliance on external capital, stretched valuations, and concentration risk
AI capex now drives roughly one-third of U.S. economic growth, creating a single point of failure

Who Is Spending What

The company-by-company breakdown, drawn from Q4 2025 and Q1 2026 earnings guidance, reveals an arms race where nobody is willing to blink.

Company	2026 Capex (est.)	YoY Change	Primary Focus
Amazon	$200B	~50% increase	AWS AI infrastructure, robotics, LEO satellites
Alphabet	$175-185B	Significant increase	Gemini models, Vertex AI, Google Cloud
Microsoft	$145B (run rate)	Sharp increase	Azure data centers, OpenAI partnership
Meta	$115-135B	Substantial increase	Llama open-source models, AI ad infrastructure
Combined	~$635-665B	67-74% vs 2025

Amazon is the biggest spender at $200 billion, reflecting a near-50% year-on-year increase. Alphabet is close behind at $175-185 billion. Microsoft's quarterly run rate of $37.5 billion puts it on pace for $145 billion annually, with the notable detail that 45% of its $625 billion in expected future cloud contracts come from a single customer: OpenAI. Meta rounds out the group at $115-135 billion, riding the momentum of crossing $200 billion in annual revenue driven largely by AI-powered advertising.

"None of them are willing to lose," said Gil Luria, an analyst at DA Davidson. He characterized the competition as "the next winner-take-all or winner-takes-most market."

Why Bridgewater Says This Is Dangerous

Jensen's warning is not that AI is a bad bet. It is that the nature of the bet has changed.

From Digital to Physical

A year ago, the AI boom was primarily a story about software - models getting smarter, APIs getting cheaper, startups multiplying. Now it is a story about concrete and copper. The exponential growth in compute demand means the boom "now requires a huge physical component that is also growing exponentially and facing many new constraints," Jensen wrote.

The evidence is already showing up in supply chains. Western Digital's CEO confirmed during the company's Q2 earnings call that the firm is "pretty much sold out for calendar 2026," with firm purchase orders from its top seven customers and long-term agreements stretching into 2027 and 2028. Hard drive prices have surged roughly 46% since September 2025. Memory chips are sold out through 2027. Data centers are hitting a "soft wall" on power availability, with companies deploying behind-the-meter natural gas turbines to work around grid constraints.

From Savers to Seekers

Perhaps the most structurally significant shift: the four companies have transitioned from being net capital accumulators - running massive buyback programs - to being net capital seekers. Exponential capex growth now requires external funding sources. Bridgewater projects that by 2027, the AI sector's capital needs could exceed total U.S. corporate bond market issuance for that year.

From Cheap to Priced In

A year ago, Bridgewater's view was that the AI ecosystem appeared "measurably cheap relative to most likely outcomes." That is no longer the case. Forward valuations among global AI-related companies have risen sharply, and more of the exponential growth assumption is already baked into stock prices.

One-Third of U.S. Growth

This may be the most underappreciated risk. Bridgewater estimates AI-related capex contributed about 50 basis points to U.S. GDP growth in 2025 and could add 100-140 basis points in 2026, with 150 basis points projected for 2027. Those are "massive sums, on par with the contribution of business investment to growth in the tech bubble," the firm noted. If this single growth driver falters, the macro consequences would be severe.

The Market's Verdict So Far

Investors have not taken the spending announcements well. Amazon shares fell 11% after its earnings disclosure. Microsoft dropped 18% from its late January peak, making it the worst initial performer of the four. Alphabet dipped 4% before recovering. Meta initially rose 10% on strong revenue numbers before getting pulled down in the broader tech selloff. Across Amazon, Alphabet, and Microsoft alone, roughly $900 billion in market capitalization evaporated.

Counter-Argument

Not everyone sees a bubble. Shai Luft, CEO of Bench Media, argues that the spending "represents roughly 15-20% of combined annual revenue" and that "the bigger risk isn't overconfidence, it's under-investing and leaving core products exposed to disruption." The companies, after all, are not speculative startups - they are the most profitable businesses on earth, and they are spending money they largely have.

The comparison to the dot-com era has limits. Fed officials have noted that today's AI investors are established firms with actual earnings, not cash-burning startups with PowerPoint decks. And unlike the late 1990s, the underlying technology - large language models, agentic AI, multimodal systems - is already generating real revenue. OpenAI alone is on pace for $13 billion in 2025 revenue, targeting $280 billion by 2030.

But Leela Nair, a senior analyst at Ebiquity, offers a note of caution: "This isn't a typical investment cycle, it's an AI arms race where infrastructure doesn't automatically translate to revenue."

What the Market Is Missing

The risk Bridgewater is flagging is not that AI will fail. It is that the American economy has become dangerously dependent on a single growth engine. One-third of U.S. economic growth riding on four companies' infrastructure spending is not a sign of strength - it is a concentration risk that would make any portfolio manager wince.

Jensen outlines four scenarios that could break the cycle: capital-raising difficulties tightening liquidity; a technological breakthrough that renders current data centers obsolete; the LLM paradigm stalling and requiring fundamentally new approaches; or populist backlash triggering aggressive regulation, potentially becoming a central issue in the 2028 election.

None of these are his base case. The bubble, he believes, "likely lies ahead" rather than behind. But when the four largest companies on earth are spending more on AI infrastructure than most countries spend on their entire economies, and when that spending is now the primary engine of U.S. GDP growth, the margin for error has gotten very thin.

The money is flowing. The question is whether the returns will follow fast enough to justify it - or whether we are building the most expensive infrastructure in human history on the assumption that they will.

Sources: