Big Tech is spending $650 billion on AI in 2026. Guess who's actually profitable? Not the companies making that bet.

Nvidia is printing money. The chip manufacturer reported $130.5 billion in revenue for fiscal 2025, up 114% year-over-year, with a jaw-dropping 73.6% gross margin on data center sales. Everyone needs GPUs, and Nvidia makes the best ones. As one analyst put it: "The clouds are sold out and our GPU installed base is fully utilized."

Meanwhile, the four hyperscalers—Amazon, Microsoft, Google, and Meta—are hemorrhaging cash. Amazon is projecting negative free cash flow of $17-28 billion. Google's free cash flow is expected to drop 90%. Meta is spending $115-135 billion on AI with no direct revenue line to show for it. Wall Street noticed, wiping roughly $950 billion off their combined market value.

The model makers aren't faring much better. OpenAI hit $20 billion in annual revenue but faces projected operating losses of $14 billion in 2026. Deutsche Bank estimates the company will burn through $143 billion in cumulative negative cash flow through 2029. That's not a typo.

So who's quietly getting rich? Energy companies. By 2030, U.S. data centers will consume electricity equivalent to Japan's entire annual demand. Electricity prices near data center clusters have risen up to 267% since 2020. AI doesn't just need chips. It needs power. Lots of it.

The structural reality is stark: only Nvidia and energy infrastructure providers are reliably profitable right now. Everyone else is betting that enterprise AI adoption will eventually justify historic capital expenditures. It's the biggest tech bet since the cloud. And unlike the cloud, which had an obvious business model, AI's path to profitability is still theoretical for most companies.

The technology is impressive. The economics are terrifying.