Project Kilby Plans Texas Power Plant

The update mainly adds that Chevron has not yet approved the project and broadens the framing to unresolved execution and community-impact issues. It also confirms the same West Texas power deal while slightly extending Chevron’s parallel search for similar data-center power arrangements.

Project Kilby is now more specifically defined as a colocated 2.67-gigawatt natural-gas facility for Microsoft, with a clearer execution path and broader implications. The update also adds Chevron’s wider search for similar data-center power deals and sharper environmental scrutiny.

• Project Kilby is planned beside Microsoft’s data center campus.
• The facility may support roughly 2 gigawatts of additional Microsoft capacity.
• Chevron is evaluating similar deals in the Midwest, Gulf Coast, Rockies, Utah, and elsewhere.
• Environmental groups have raised emissions concerns.
• Chevron plans a final investment decision by the end of 2026.

The story has narrowed from a broad set of AI data center buildouts to a more specific emphasis on power-constrained execution in West Texas, led by Microsoft’s Project Kilby and its long-term Chevron-backed gas supply plan. It also adds a separate Michigan expansion by Hyperscale Data, including a new AI compute contract and a pending local noise lawsuit.

• Project Kilby is described as a roughly 2.7 GW power arrangement.
• Chevron has a long-term natural gas supply role for the project.
• GE Vernova and Caterpillar turbines are expected to supply most generation.
• First power for the West Texas project is targeted for 2028.
• Hyperscale Data signed a 20 MW AI compute agreement in Michigan.

The story broadens from Microsoft-centric power deals into a wider industry pattern of mega-campus AI infrastructure, with OpenAI, Nvidia, SB Energy, and Constellation now part of the financing and supply picture. The emphasis also shifts from a single West Texas project to a larger buildout pipeline spanning multiple states and grid arrangements.

• OpenAI’s Ohio plans may require federal land and private parcels.
• Nvidia is reportedly a financing backstop in campus negotiations.
• Constellation Energy is now tied to Microsoft’s PJM supply arrangement.
• Microsoft’s expansion includes land purchases in both Wyoming and Texas.
• The story now spans Michigan alongside Texas, Wyoming, Ohio, and PJM.

The story has sharpened around Microsoft's Project Kilby in West Texas, moving from a general pattern of AI power deals to a more specific build centered on dedicated gas generation, co-located infrastructure, and phased delivery. The newest material also adds clearer execution details on water sourcing, supply-chain partners, and a separate large Ohio lease that reinforces the capital intensity of the sector.

• Project Kilby is described as roughly 2.67-2.7 GW.
• First power is targeted for 2028, with phased buildout into the 2030s.
• Water planning now includes closed-loop cooling and produced-water reuse.
• Joulent, GE Vernova, Caterpillar, and Engine No. 1 appear in the supply chain.
• OpenAI is reportedly pursuing a 10 GW Ohio campus lease backed by Nvidia and SB Energy.

The story has expanded from mostly Microsoft/OpenAI land-and-power buildouts into a broader, more concrete view of AI infrastructure sourcing, with Chevron’s gas-backed Project Kilby now the clearest new example of dedicated generation for data centers. The current version also shows stronger emphasis on implementation details—power contracts, phased delivery, and site control—rather than just announced scale.

• Chevron signed a 20-year natural gas agreement for Project Kilby.
• Project Kilby is a dedicated behind-the-meter generation setup for AI demand.
• West Texas, Reeves County, and Pecos are now central locations.
• Projects face explicit permitting, FID, and phased-delivery milestones.
• Power procurement now spans gas, renewable, and nuclear-linked contracts.

The story has broadened beyond Microsoft-led land and power buildout to include a much larger, more speculative OpenAI campus in Ohio, backed by Nvidia and structured around federal-land leasing. That shift raises the apparent scale and financing intensity of the AI infrastructure wave dramatically.

• OpenAI is reportedly pursuing a 10-gigawatt Ohio campus.
• The Ohio campus would be leased on federal land.
• Nvidia is tied to both supply and financing for the buildout.
• Reported project costs reach hundreds of billions of dollars.
• The story now includes Southern Ohio and federal-land leasing.

The story has become more specific and more Microsoft-centric: Cheyenne is now the clear hub, and the power story is sharpened by the PJM-linked Three Mile Island restart. A smaller but new thread also shows concrete leasing momentum for a Michigan hyperscale AI campus.

• Cheyenne expansion now includes additional parcels totaling about 420 acres.
• Three Mile Island restart is linked to PJM data-center load.
• Microsoft reported sharply higher commercial remaining performance obligations.
• Michigan campus is moving toward power and infrastructure agreements.
• Cheyenne faces industrial land-use and community scrutiny.

The framing has become more concrete and execution-oriented: Microsoft’s AI buildout is now tied explicitly to land acquisition, power procurement, and campus planning, while Hyperscale Data’s Michigan project has advanced to visible tenant-demand indicators. The story also adds a sharper emphasis on persistent hardware and grid constraints as the binding limits on converting capex into deployed capacity.

The story shifts from a general account of AI data-center expansion to a more operational picture: operators are now actively locking up power, land, and equipment rather than simply signaling future demand. Microsoft remains the clearest driver, but the framing also emphasizes that smaller projects are moving into concrete leasing and site-prep steps.

The story now extends beyond Microsoft’s demand-led buildout to include a second, smaller-scale operator, Hyperscale Data, showing that AI infrastructure expansion is still constrained but also actively moving into execution. Power procurement and site control remain central, but the framing shifts slightly toward broader industry-wide scaling activity.

Microsoft is expanding AI infrastructure at unusually large scale because demand is already booked and current compute supply remains constrained. The company is pairing heavy capex with land, data center, and power commitments to turn contracted AI workloads into usable capacity. The main significance is that Microsoft’s AI growth story is increasingly shaped by physical bottlenecks, especially chips, storage, and electricity.