$650 Billion for Intelligence: Why Big Tech’s AI Race Is Turning Into an Infrastructure War

In 2026, four US tech giants — Alphabet, Amazon, Meta, and Microsoft — are set to pour around $650 billion into artificial intelligence, according to analysts at Bridgewater Associates. That’s almost 60% more than the roughly $410 billion spent in 2025 and a clear sign that AI has stopped being a “software experiment” and has turned into a construction project of historic scale.

AI as the build‑out of the century

In a letter to clients, Bridgewater co‑CIO Greg Jensen writes that the AI boom has entered a “more dangerous phase,” where capital expenditures on physical infrastructure are growing exponentially and companies are increasingly leaning on external financing. Demand for compute is “much stronger than supply,” so the hyperscalers are racing ahead with data‑center construction, chip purchases, and upgrades to networks and power infrastructure, essentially betting they’ll “catch up with demand later.”

Bridgewater estimates that AI infrastructure alone (data centers, hardware, networks) could exceed $500 billion in cumulative investment by the end of 2026, with total AI capex by the “big four” reaching around $650 billion. The firms are already sharply cutting back on share buybacks to redirect cash into building “intelligence factories” — the massive clusters where next‑generation models and agents will run.

Bubble risks and collateral damage

On the bright side, this investment wave is giving a tangible boost to the economy: Bridgewater estimates that tech spending added about 0.5 percentage points to US GDP growth in 2025 and could contribute roughly 1 percentage point in 2026. But there’s a darker side. Jensen warns that the sheer scale of spending creates “substantial downside risks” if anything goes wrong — from a slowdown in demand for AI services to weaker‑than‑expected monetization of models.

He singles out Anthropic and OpenAI: to raise money at large pre‑IPO valuations, they need fresh product breakthroughs and a “credible super‑profit story”; without that, it will be hard to justify current valuations and capital needs. At the same time, AI is already creating existential risks for adjacent sectors. Bridgewater points to the recent slide in software stocks and bluntly says the AI leaders “cannot meet investor expectations without posing a threat to other sectors like software and data providers.”

From software race to a battle over hardware and power

Taken together, Bridgewater’s analysis and market commentary point to an important shift: this is no longer just a race for the smartest models, but a war over physical infrastructure. The winners will be not only those who write the best code, but those who control compute capacity, data centers, and access to cheap energy.

Economists warn that such a concentration of investment in one tech cluster looks a lot like the late stage of the dot‑com bubble — with one key difference: this time, the stakes are tied to very tangible assets, from chips to power plants. If AI revenue growth lives up to expectations, the $650‑billion bet could become the foundation of a “second internet”; if it doesn’t, markets risk a painful correction, higher energy costs, and a long freeze in capital availability for the rest of the IT sector.