Alphabet buys Intersect for $4.75B as AI power crunch grows

AI’s biggest constraint is starting to look less like software and more like infrastructure. Now that “always on” AI functions are becoming more widespread, the competitive edge is shifting toward who can secure reliable power fast enough to keep data centers expanding.

That helps explain Alphabet’s move this week.

On December 22, 2025, the parent company of Google, which builds the Gemini AI models, announced that it will buy Intersect Power for $4.75 billion in cash, plus assumed debt, positioning the deal as part of Google's push to accommodate surging AI demand within its data centers. Reuters reports that this move is designed to secure more power capacity for Google’s expanding footprint.

Intersect is not a model company. It is an energy and infrastructure developer with a build pipeline tied to data center expansion. Reuters reports the company’s pipeline is expected to reach 10.8 gigawatts by 2028. That’s grid-shaping scale, not what you find in a typical corporate power purchase agreement. In other words, Alphabet is acquiring speed and optionality—projects and momentum that can bring capacity online on timelines that match AI product cycles.

The structure matters, too. Reuters and the Financial Times note that Intersect’s operating assets in places like Texas and California remain separate under its existing investors, while Alphabet takes ownership around the development and construction engine. That split reinforces what Alphabet appears to value most: future capacity and execution, not necessarily the existing operating portfolio.

One detail making the rounds is a comparison to the Hoover Dam, and it is worth explaining clearly. The nameplate capacity of Hoover’s powerplant is about 2.08 GW. On capacity alone, 10.8 GW is roughly five times Hoover. 

The “20×” framing you may see is based on Hoover's average annual electricity generation (about 4B kWh), which translates to a much lower average output. Different yardsticks, same takeaway: Intersect’s pipeline is big enough to be a strategic asset in an AI buildout.

This deal also aligns with a broader Google narrative: co-locate data center load with new clean power so projects can come online faster and require less new transmission. Intersect previously announced a partnership with Google and TPG Rise Climate aimed at scaling “power-first” data center development, with Intersect pointing to a $20 billion investment roadmap through the end of the decade and early projects slated for the 2026–2027 window.

Zoom out, and the implications for AI automation become clear. Sales and marketing automation is shifting from batch-based campaigns to continuous personalized and agent-based outreach. 

In healthcare and legal domains, AI applications are being increasingly integrated into their day-to-day tasks such as documentation, summarization, drafting, and review, where availability and costs become as important as performance. These tasks don't act like training exercises. These act like services that depend on reliable and scalable power every single day.

Alphabet is signaling that the next AI race is not just for better models or more chips. It’s for the infrastructure that makes AI cheap and dependable at scale. The company has also been expanding energy partnerships elsewhere—Reuters notes this is an indication that power procurement is becoming a core AI strategy, not a side project.

What to watch next: whether this becomes a template. If other AI leaders start buying or tightly controlling “power-first” development pipelines, the market may split between companies that can build great AI and those that can run it at the volumes customers actually want.