Chevron just moved from selling fuel into selling the operating foundation of artificial intelligence.
On June 22, 2026, Chevron, through its wholly owned subsidiary Energy Forge One, signed a 20-year agreement to build a dedicated power facility for a Microsoft-operated data center in Reeves County, Texas. The development, known as Project Kilby, is expected to deliver about 2.67 gigawatts of capacity in phases, with most of the power coming from GE Vernova turbines and extra capacity supplied by Solar Turbines, a Caterpillar subsidiary.
This is not a side project. It is a structural shift in the AI economy. Microsoft needs firm electricity for cloud and advanced compute. Chevron has Permian Basin gas, capital discipline, project-development experience, and a direct route into a new high-demand customer base. Big Tech needs power that does not blink. Big Oil wants a cleaner growth story than crude cycles alone.
The deal turns West Texas gas into a strategic input for AI infrastructure. It also reveals a harder truth: the AI boom is no longer limited by chips, models, or software talent. It is being shaped by electricity access, turbine supply, grid congestion, gas economics, and local infrastructure politics. Readers tracking AI energy infrastructure are now watching the next phase of the market form in real time.
From Cloud Campus To Power Foundry
Project Kilby is planned near Pecos in Reeves County, close to the Waha Hub, one of the most important gas pricing points in the Permian Basin. That location is the business logic.
The Permian produces huge volumes of associated gas alongside oil. At times, local gas prices have been weak or even negative when pipeline capacity fails to keep pace with production. Chevron’s agreement with Microsoft creates a new demand source near the basin itself: burn the gas locally, generate power on site, and feed that electricity directly into a colocated data center campus.
That model cuts around one of the biggest delays in the U.S. power market: waiting for grid interconnection. AI data centers need enormous loads, and traditional utility planning timelines can stretch for years. A behind-the-meter power plant gives Microsoft a dedicated supply path. It gives Chevron a long-term buyer. It gives the region a project that could bring construction jobs, operating roles, and tax revenue.
Local reporting on the Project Kilby power deal said the development is forecast to create more than 6,000 construction jobs at peak build-out, hundreds of permanent operational jobs, and more than $10 billion in state and local tax revenue. Those numbers explain why AI infrastructure is now an economic-development weapon.
The Microsoft Demand Problem
Microsoft’s AI strategy has created a power problem hiding inside a growth story.
The company is building around Azure, Copilot, OpenAI workloads, enterprise AI tools, and advanced compute contracts. Every layer of that stack needs data centers. Every data center needs electricity. The newest AI workloads need dense compute halls that draw far more power than older cloud workloads.
Microsoft has spent years buying renewable power and pushing carbon goals. The Chevron deal does not erase that strategy, but it does complicate the story. AI demand is now large enough that hyperscalers need every credible source of firm capacity: renewables, batteries, nuclear contracts, grid purchases, and gas-fired plants close to data center campuses.
That is the pivot. Renewable energy can be cheap and clean, but data centers need constant uptime. Batteries can smooth gaps, but long-duration supply is still hard. Grid interconnection can take too long. Gas turbines can be built closer to load, dispatched around demand, and fed by domestic fuel.
For Microsoft, the value is reliability. For Chevron, the value is customer lock-in. For GE Vernova and Caterpillar, the value is a new order cycle tied to AI buildouts rather than traditional utility demand.
The Hardware Of Gas-Powered AI
The phrase “gas-powered AI” sounds like a contradiction until the project map is visible. Behind the cloud interface sit turbines, substations, pipelines, water systems, emissions controls, transformers, switchgear, and land agreements.
Chevron and Engine No. 1 previewed this strategy in January 2025 with GE Vernova, saying their joint development aimed to deliver up to 4 GW of power for U.S. data centers using natural gas. The companies described the projects as “power foundries” built for colocated data centers, with early plans involving seven U.S.-made GE Vernova 7HA gas turbines.
Project Kilby now gives that concept a named customer and a major location.
| Project Element | Reported Detail | Strategic Meaning |
|---|---|---|
| Customer | Microsoft | Big Tech demand anchors the project |
| Energy Developer | Chevron subsidiary Energy Forge One | Big Oil moves into direct power supply |
| Location | Reeves County, Texas, near Pecos | Permian gas becomes local AI fuel |
| Capacity | About 2.67 GW | Utility-scale load for advanced compute |
| Main Equipment | GE Vernova turbines | Turbine supply becomes part of AI infrastructure |
| Added Capacity | Solar Turbines, a Caterpillar subsidiary | Industrial suppliers gain AI exposure |
| Agreement Term | 20 years | Long-duration cash flow replaces spot-market volatility |
This is why oil and gas companies are looking at data centers with fresh interest. AI workloads need long-term energy contracts. Energy producers want predictable demand. The overlap is obvious: the model converts gas molecules into compute capacity.
Why The Grid Is No Longer Enough
The American power grid was not built for a sudden AI land rush.
Data centers already strain regional systems in places such as Northern Virginia, Texas, Arizona, Georgia, and parts of the Midwest. New AI campuses arrive with loads that can look like small cities. Utilities must plan generation, transmission, transformers, substations, and reserve margins around customers that want speed now.
The International Energy Agency’s AI electricity outlook projects global electricity supply for data centers rising from 460 terawatt-hours in 2024 to more than 1,000 terawatt-hours in 2030 in its base case. That is the scale behind Chevron’s move. AI is turning data centers from real estate assets into power-market actors.
Behind-the-meter projects answer one question: how does a hyperscaler get capacity without waiting for the grid to catch up? The answer is to build energy and compute in the same place.
That carries political risk. A private AI campus can reduce pressure on the grid at first, but it can still reshape local water use, land values, gas demand, emissions, construction traffic, and long-term power planning. If the plant later connects to the regional system, it becomes part of a wider market that affects more than Microsoft alone.
The Climate Math Gets Harder
Microsoft has pledged to be carbon negative by 2030. Chevron’s core business remains hydrocarbons. Project Kilby sits directly between those two realities.
Chevron and its partners have said the broader gas-power model can be built with the flexibility to integrate lower-carbon options such as carbon capture and storage, along with renewable resources. Project Kilby’s local reporting points to emissions controls, minimal water-use design, brackish non-potable water, and potential produced-water use.
Those details matter, but they do not remove the central issue. Gas-fired power still produces emissions. Carbon capture can lower the footprint, yet cost, capture rates, permitting, transport, and storage all matter. Microsoft now faces the optics of using natural gas to support AI growth at a time when tech companies are trying to keep climate commitments credible.
The market may tolerate that trade-off if power scarcity becomes the main bottleneck. Investors are already rewarding infrastructure capacity more than clean branding alone. The AI race is moving too fast for a single energy source to carry the load. That does not make gas clean. It makes gas useful.
Chevron’s New Growth Lane
For Chevron, Project Kilby is more than a power plant. It is a hedge against the old commodity cycle.
Oil and gas earnings swing with prices. Data-center power contracts can create longer-duration revenue tied to electricity demand from elite corporate customers. A 20-year agreement with Microsoft has a different risk profile than selling into volatile commodity markets.
The strategic gain is vertical movement. Chevron can supply gas, develop the power facility, and sell dedicated electricity to a hyperscaler. That pulls the company closer to the end customer and deeper into infrastructure. It also gives Chevron a story that aligns with American energy security, AI leadership, industrial jobs, and domestic manufacturing.
ExxonMobil has shown interest in similar opportunities. Utilities, independent power producers, nuclear developers, renewable firms, and fuel-cell companies are all chasing the same customer base. The new AI supply chain has room for chipmakers and cloud providers, but it also has room for drillers, turbine manufacturers, construction firms, and land-rich regions with energy access.
| Winner | Why The Deal Matters |
| Microsoft | Secures dedicated large-scale power for AI and cloud growth |
| Chevron | Enters long-term data-center power supply with Permian gas leverage |
| GE Vernova | Gains turbine exposure tied to AI infrastructure demand |
| Caterpillar | Participates through Solar Turbines equipment |
| Reeves County | Receives jobs, tax revenue, and industrial investment |
| Rival Energy Developers | Gets a clear template for colocated AI power projects |
The New AI Stack Runs On Fuel
The old AI stack was chips, models, data, and cloud. The new AI stack adds turbines, pipelines, water sourcing, land rights, and local politics.
Chevron’s deal with Microsoft makes that shift impossible to miss. The frontier model race still matters. GPU access still matters. Software distribution still matters. Yet the next bottleneck may sit outside the data hall: can the company get enough power, fast enough, at a cost investors accept?
Project Kilby is a template for the next wave. Colocated gas plants can get built near fuel sources. Data centers can be designed around dedicated generation. Energy companies can move into direct compute infrastructure supply. Tech companies can turn to oil majors when utilities cannot deliver capacity on AI timelines.
That creates a new market map. Big Oil is not just watching the AI boom from the sidelines. It is becoming part of the machine.
