Chevron Lands 20-Year Deal To Power Microsoft's AI Expansion

Microsoft has signed a 20-year agreement with Chevron to power a massive new AI-focused data center campus in West Texas, underscoring the growing race among tech companies to secure reliable energy supplies, according to Bloomberg.

The project, known as Project Kilby, is expected to begin generating power in 2028 and eventually reach 2.67 gigawatts—enough electricity for more than 530,000 Texas homes.

Chevron is developing the project with Engine No. 1 and expects to make a final investment decision later this year. Despite the enormity of the deal and the inroads into powering AI directly, Chevron stock was little changed after the cash open:

Bloomberg writes that the site near Pecos, Texas, will use natural gas from the Permian Basin to fuel GE Vernova turbines and generate electricity directly for Microsoft's planned data center campus. Because the facility will produce its own power, it will not draw from the grid.

“Consumers are concerned about and are already feeling the effect of power-demand growth,” said Jeff Gustavson, Chevron’s president of New Energies. “We specifically designed this, in this part of the country, to avoid any of that.”

The agreement comes as Microsoft accelerates its AI infrastructure buildout to compete with Alphabet and Amazon. The company has said it plans to double its data center footprint over the next two years, driving demand for large-scale, dependable power sources.

Chevron argues the project also creates a productive use for abundant Permian natural gas that is often wasted because pipeline capacity is limited. “This is the most abundant gas basin in the country, maybe the world,” Gustavson said.

Chevron and Engine No. 1 have already secured orders for seven GE Vernova gas turbines. While Chevron has not disclosed the project's cost, people familiar with the matter previously estimated it at roughly $7 billion.

The deal reflects a broader trend across the U.S., where data center capacity is expected to double to 77 gigawatts by 2030. Texas alone has 33 gigawatts of planned data center power projects, the most of any state, highlighting how AI is reshaping energy demand and infrastructure investment.

At a higher level, the deal highlights a growing shift in how AI infrastructure is being financed and powered. Rather than relying solely on utilities and the existing grid, hyperscalers are increasingly partnering directly with energy producers to secure dedicated generation capacity. That model could accelerate data center development while reducing grid constraints, but it also raises questions about the long-term balance between AI-driven power demand, decarbonization goals, and energy affordability.

As we have constantly written about over the last 2 years, as AI becomes a larger share of electricity consumption, access to reliable power may emerge as a competitive advantage on par with access to chips and computing infrastructure.