Investors are financing power plants, fuel cells and other on-site systems to get electricity to AI data centres faster than the grid can deliver it. The deals bring new money to a key AI bottleneck, but can tie data-centre operators to long contracts even as computing needs change.
Watch Desk analysis
What happened
The Information describes several ways investors are backing power for data centres: funding individual projects, taking stakes in power developers and financing equipment for customers. It cites a Blackstone-led group’s $5.3 billion investment for a 49% stake in five gas-fired projects Williams is developing for data centres. Williams said its first 200 megawatts of on-site generation came online less than 18 months after commercialisation; the article contrasts that with Fitch Ratings’ estimate that grid-connected power plants and high-voltage lines can take five to 10 years to build.
Other examples include Brookfield financing future Bloom Energy projects, with the potential amount expanded to $25 billion, and a $1.7 billion investment to install Bloom fuel cells at Nebius data centres. These are distinct arrangements, not one shared funding pot.
Why it matters
For AI developers and cloud providers, securing power sooner could help move data-centre projects ahead while grid infrastructure catches up. On-site power can cost more, and the financing structures matter: investors want predictable returns, while operators may want flexibility if grid access improves or their power needs change.
The Information says investors are increasingly favouring customer commitments of 10 to 15 years, with five-year renewal options. That creates a potential mismatch: power assets may operate for decades, while AI computing contracts are shorter and GPUs can become outdated within a few years.
Our read
Power is part of the AI build-out, not an afterthought waiting politely outside the server room. These deals may help projects get electricity sooner, but their long commitments could leave operators paying for capacity after their needs have shifted. The important question is not just who will fund the power, but who carries the risk if the technology or demand changes first.
What to watch
- Whether on-site generation is delivered quickly enough to ease data-centre power constraints.
- How often investors require long-term commitments from cloud and data-centre customers.
- Whether projects can sell power elsewhere if a data-centre customer’s needs change.
Discussion spark: Should AI data-centre operators accept long power contracts to build sooner, or keep flexibility in case grid access and computing needs change?
Sources and evidence
- How Investors Are Financing On-Site Power to Break AI Bottlenecks (29 September 2026, 18:15 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.