Big Tech’s $650 Billion AI Gamble: The Stakes Have Never Been Higher

When Greg Jensen sat down to write his latest letter to Bridgewater Associates clients, he wasn’t mincing words. The artificial intelligence boom, he warned, has entered what he called a “more dangerous phase.” The numbers back him up: Alphabet, Amazon, Meta, and Microsoft are collectively preparing to spend approximately $650 billion on AI infrastructure this year alone—a 58% jump from the $410 billion they invested in 2025.

“Compute demand continues to significantly outpace supply, driving hyperscalers to invest even more rapidly to try to someday get ahead of the demand.” — Greg Jensen, Bridgewater Associates

The Scale of the Bet

To put that $650 billion figure in perspective, it exceeds the annual GDP of countries like Poland, Belgium, and Sweden. It’s more than the entire market capitalization of most Fortune 500 companies. And it’s being spent by just four technology giants on a single technology category—artificial intelligence infrastructure.

The spending isn’t just about building bigger data centers or buying more GPUs. It represents a fundamental restructuring of how these companies allocate capital. All four have already begun curbing share buybacks—traditionally a favored method of returning cash to shareholders—to help fund this unprecedented surge in capital expenditure.

The infrastructure arms race is being driven by a supply-demand imbalance that shows no signs of easing. Every major AI model release seems to trigger another wave of demand for compute resources. Training runs for frontier models now require tens of thousands of GPUs running for months. Inference—the actual serving of AI responses to users—consumes even more resources as adoption scales.

What Could Go Wrong

Jensen’s warning about the “dangerous phase” isn’t just rhetoric. The scale of spending creates significant downside risks if anything goes wrong. A severe stock market correction could undermine growth and limit companies’ ability to raise capital, echoing patterns seen during the Dot-com bubble in 2000.

“It is no longer possible for AI leaders to satisfy their investors’ expectations without creating existential risks to other sectors like software.” — Greg Jensen, Bridgewater Associates

Valuation pressures are already mounting for AI labs like Anthropic and OpenAI. Both will need major product breakthroughs to secure backing for massive final fundraisings ahead of potential IPOs. Without a credible path to outsized profits, they could struggle to justify lofty valuations and heavy capital demands.

The ripple effects extend beyond AI companies themselves. Software companies and data providers are already feeling the pressure, as evidenced by the recent selloff in software stocks. As hyperscalers pour resources into building their own AI capabilities, they’re increasingly competing with the very software vendors they once partnered with.

The Economic Ripple Effects

Beyond stock markets, the spending boom has tangible macroeconomic implications. Bridgewater estimates that tech investment added about 50 basis points to U.S. GDP growth in 2025 and could provide around 100 basis points of support this year. That’s a meaningful contribution to overall economic growth.

But there are inflationary pressures too. The spending boom may lift prices for technology and communications equipment. More significantly, it could push up electricity prices in some regions as data center power demand surges. Northern Virginia, home to the world’s largest concentration of data centers, is already grappling with grid constraints.

Regional disparities are emerging as a key concern. Areas with abundant renewable energy and favorable regulatory environments—think Texas, Arizona, and parts of the Midwest—are attracting disproportionate investment. Meanwhile, regions with constrained power grids or restrictive policies risk being left behind in the AI economy.

The Path Forward

For investors, the key question isn’t whether AI infrastructure spending will continue—it’s whether the returns will justify the investment. The hyperscalers are essentially making a collective bet that AI will become as fundamental to computing as cloud services are today.

History offers mixed precedents. The build-out of cloud infrastructure in the 2010s ultimately proved to be a massive value creator. The fiber optic and telecom infrastructure spending of the late 1990s, on the other hand, destroyed enormous amounts of capital before eventually finding productive uses.

The difference may come down to timing. If AI capabilities continue to improve rapidly and adoption accelerates across industries, the infrastructure investments will look prescient. If progress stalls or adoption plateaus, the $650 billion could start looking like an expensive lesson in the dangers of herd behavior.

For now, the money keeps flowing. And the hyperscalers keep building, racing to stay ahead of demand that always seems to be one step ahead of supply.

This article was reported by the ArtificialDaily editorial team. For more information, visit Bridgewater Associates and Economic Times.