Microsoft has entered a 20-year agreement with Chevron tied to a dedicated power project in West Texas, a sign that AI data center power is becoming as strategic as chips and cloud capacity. The planned site, known as Project Kilby, is expected to begin generating electricity in 2028.
At full buildout, the project is slated to reach 2.67 gigawatts of capacity, enough to serve more than 530,000 Texas homes. That scale places the agreement among the clearest examples yet of a major technology company securing long-term, direct energy supply for AI expansion.
The structure also stands out because the facility is designed to generate power for Microsoft’s planned campus without drawing electricity from the grid. For investors, the deal offers an early view of how hyperscalers may navigate rising power constraints as AI workloads accelerate.
Key Facts
- Microsoft signed a 20-year agreement linked to Chevron’s Project Kilby in West Texas.
- The project is expected to start generating power in 2028 and eventually reach 2.67 gigawatts of capacity.
- Estimated cost for the development has previously been placed at about $7 billion.
- Chevron and Engine No. 1 have already secured orders for seven GE Vernova gas turbines.
- U.S. data center capacity is projected to double to 77 gigawatts by 2030, with Texas hosting 33 gigawatts of planned projects.
AI Data Center Power
The agreement reflects a wider shift in the economics of artificial intelligence infrastructure. As companies such as Microsoft, Alphabet, and Amazon scale training and inference capacity, the constraint is no longer limited to semiconductors, networking gear, or land. Electricity supply, particularly reliable baseload power available on a predictable timeline, is moving to the center of capital planning.
Project Kilby is designed around natural gas sourced from the Permian Basin and electricity generated by GE Vernova turbines near Pecos, Texas. By keeping production off-grid and directly tied to the campus, the model aims to reduce exposure to congestion, permitting delays, and regional power shortages. Chevron’s Jeff Gustavson described the design as a way to avoid adding strain to consumers already facing the effects of rising electricity demand.
The arrangement matters because it shows how large technology buyers are beginning to partner directly with energy producers rather than relying only on utilities. That could speed deployment of new data centers, especially in regions where interconnection queues and transmission bottlenecks have become major obstacles. It also creates a new lane of demand for energy companies that can combine fuel access, generation assets, and long-duration contract structures.
Reliable power is rapidly becoming a strategic input for AI, not just an operating expense.
Why West Texas Fits the Model
West Texas offers a rare combination of advantages for large-scale power-intensive development: abundant natural gas from the Permian Basin, available land, and proximity to one of the most active regions for new energy infrastructure. Chevron has argued that the project creates a productive outlet for gas that can otherwise be stranded or discounted when pipeline capacity is tight.
That practical benefit helps explain why natural gas remains central to some AI-related infrastructure plans despite broader decarbonization goals. For operators building multi-gigawatt campuses on aggressive schedules, gas-fired generation can offer dispatchable power with clearer timing than many alternatives. The trade-off is that investors will likely keep watching how companies balance speed, cost, emissions, and future regulatory pressure.
Implications for Investors
For Microsoft, the deal reinforces the scale of its AI capital cycle. The company has said it plans to double its data center footprint over the next two years, and long-term power agreements can reduce a major execution risk in that expansion. Investors evaluating cloud and AI leaders may increasingly need to assess energy procurement strategy alongside GPU supply, software monetization, and enterprise demand.
For Chevron, the agreement points to a potential growth avenue beyond traditional upstream and refining narratives. If direct power supply for data centers becomes a durable market, integrated energy companies with fuel access and development expertise could capture new long-term contracted cash flows. Engine No. 1 and GE Vernova also gain relevance as beneficiaries of a buildout that requires generation assets, turbines, and project execution capability.
The broader market implication is that power infrastructure may become one of the most important second-order beneficiaries of AI spending. Developers of gas generation, transmission assets, grid equipment, and select utility-scale energy projects could all see stronger demand. At the same time, investors should watch for risks around permitting, fuel price volatility, environmental scrutiny, and whether a wave of dedicated generation projects changes competitive dynamics in merchant power markets.
Looking ahead, the final investment decision expected later in 2026 will be a key milestone. If Project Kilby moves forward as planned, it may become a template for how future AI campuses in the U.S. are financed, powered, and brought online.
VIP Algorithmic Setups
