The AI industry’s biggest bottleneck isn’t chips anymore. It’s power. And Big Tech is responding by doing something the energy sector never expected: building its own energy infrastructure from scratch, according to The Information.
The shift is striking. Companies that once simply plugged into the grid are now signing multi-billion-dollar deals for nuclear plants, developing gigawatt-scale solar campuses, and pioneering battery storage systems that would make traditional utilities jealous. The Information reports that this dynamic is fundamentally changing the relationship between Silicon Valley and the energy industry, with tech companies increasingly dictating terms.
The Numbers Tell the Story
The scale of investment is staggering:
- Total AI hyperscaler capex is projected to hit $530 billion in 2026
- Global data center infrastructure spending is expected to reach $7 trillion by 2030, with $1.3 trillion going toward power generation
- U.S. data center demand could reach 74 GW by 2028, according to Morgan Stanley Research, with a projected shortfall of about 49 GW
That 49 GW gap is roughly the output of 49 nuclear reactors. It’s a problem that can’t be solved by plugging into existing grids.
Who’s Doing What
Meta has locked in deals with Vistra, TerraPower, and Oklo for up to 6.6 gigawatts of nuclear capacity by 2035. Microsoft signed a 20-year agreement with Constellation Energy to restart the Three Mile Island nuclear plant. Amazon secured 1.9 gigawatts of nuclear power from the Susquehanna plant plus agreements for small modular reactor projects.
Google is taking a different angle. It partnered with NextEra Energy to co-develop multiple gigawatt-scale data center campuses and recently announced a Minnesota data center powered by a blend of wind, solar, and a massive 30 gigawatt-hour battery from Form Energy.
Alphabet’s CFO Ruth Porat has said Google is committed to adding data center capacity using nuclear power, carbon capture, and other clean energy solutions. What stands out here is that these aren’t PR announcements. They’re binding, long-term infrastructure commitments.
Bring Your Own Power Is Now a Business Requirement
Bloom Energy’s CEO captured the shift perfectly: “Bring-your-own-power has shifted from a slogan to a business necessity for AI hyperscalers.” He described it as a “secular and growing” trend.
This is significant because it flips the traditional utility model. Instead of energy companies selling power to tech firms, tech firms are now co-developing, financing, and sometimes owning their own generation capacity. They’re not waiting for utilities to build out infrastructure. They’re doing it themselves, often faster and at larger scale.
The Trump administration has noticed. The White House negotiated a “Rate Payer Protection Pledge” with major tech companies, ensuring they pay for their own power needs rather than pushing costs onto consumers. It’s a sign that AI’s energy appetite has become a political issue, not just a business one.
Why This Matters for AI Practitioners
Energy access is quickly becoming a competitive moat. Companies that secure cheap, reliable power will train bigger models faster and run inference more cheaply. Those that don’t will face higher costs or simply won’t be able to scale.
For startups and mid-size AI companies, this creates a real challenge. The hyperscalers are locking up energy capacity years in advance. Smaller players can’t sign 20-year nuclear deals. This could accelerate the concentration of AI capabilities among a handful of companies that control both the compute and the power.
What Comes Next
Expect two trends over the next 12-18 months:
- Nuclear becomes mainstream in tech: Small modular reactors and fusion investments will accelerate. The stigma around nuclear in the tech world is gone.
- Energy-as-a-service for AI: New intermediaries will emerge, offering power procurement and management specifically for AI workloads, giving smaller companies access to the energy deals they can’t negotiate alone.
The AI race was always going to become an energy race. It just happened faster than anyone predicted. For more details on how this power dynamic is reshaping the industry, check the original reporting at The Information.