The Real Winners of the AI Gold Rush

$66.8 billion. That's how much revenue CoreWeave has locked up in its backlog—more than the GDP of most countries. While everyone obsesses over ChatGPT and AI models, the real money flows to whoever keeps the lights on.

Picks and Shovels 2.0

During the 1849 California Gold Rush, the biggest fortunes went not to prospectors but to merchants selling pickaxes and shovels. The 2026 AI boom follows the same playbook. CoreWeave just posted $1.57 billion in Q4 revenue, up 110% year-over-year, by being the infrastructure backbone that OpenAI, Google, and others depend on.

Training a single AI model consumes as much electricity as a small city. Someone has to provide that power, and CoreWeave is cashing in. The company now operates 850 megawatts of active capacity with 3.1 gigawatts under contract—equivalent to three mid-sized nuclear plants.

Yet despite crushing revenue expectations, shares dropped 8% in after-hours trading. Wall Street's message: growth is great, but at what cost?

The Infrastructure Bet

CoreWeave's success story reads like a masterclass in timing. Founded as a cryptocurrency mining company, it pivoted to AI infrastructure just as demand exploded. Now it's the fastest cloud platform ever to hit $5 billion in annual revenue.

The company's $10.31 billion capex budget for 2026—lower than expected but still massive—will fund more data centers and GPU clusters. This quarter alone, CoreWeave launched object storage services to compete with Amazon Web Services and expanded its credit facility to $2.5 billion.

But growth comes with baggage. The company carries $21.37 billion in debt since going public last March. That's a hefty load, even with a $66.8 billion revenue backlog providing cushion.

Market Reality Check

While CoreWeave shares surged 36% this year, the broader tech-software ETF fell 22%. Investors are clearly betting on infrastructure over applications. The recent Anthropic announcements that spooked AI software stocks barely touched CoreWeave.

This divergence reveals a harsh truth: building AI models is expensive and uncertain. Providing the infrastructure to build them? That's a utility business with predictable cash flows and multi-year contracts.

CoreWeave's adjusted EBITDA of $898 million missed estimates, but the revenue backlog tells a different story. Companies are literally paying billions upfront for future computing capacity—the ultimate vote of confidence.

The Commodity Question

Here's what keeps CoreWeave executives up at night: will AI infrastructure become commoditized? Amazon, Microsoft, and Google all have deeper pockets and broader service portfolios. CoreWeave's advantage lies in specialization, but specialization can become a trap if the market shifts.

The company's object storage launch signals awareness of this risk. By expanding beyond pure compute into storage and networking, CoreWeave aims to become stickier with customers. But competing with AWS on storage is like challenging Walmart on price—possible, but painful.