The AI boom just hit the Permian Basin. Microsoft and Chevron announced today a massive 20-year power purchase agreement to fuel one of the largest co-located natural gas power and data center developments in the United States — a move that underscores Big Tech's insatiable hunger for energy and the uncomfortable trade-offs that come with it.
Meet Project Kilby — a 2-gigawatt AI and cloud data center campus rising in Pecos, West Texas, powered by a dedicated Chevron-built plant delivering up to 2.67 GW of electricity. It's the largest collaboration to date between a U.S. oil and gas giant and a tech company, and it marks a pivotal moment in the AI infrastructure arms race.
Inside Project Kilby: What the Deal Looks Like
Chevron will construct, own, and operate a natural gas power plant hub drawing from the nearby Permian Basin — America's largest oil field. According to TechCrunch, the plant will use at least seven GE Vernova turbines plus secondary turbines from Caterpillar subsidiary Solar Turbines. Microsoft will operate the data center itself, using the dedicated electricity to train and run its next-generation AI models including Copilot and future OpenAI workloads.
| Detail | Value |
|---|---|
| Project Name | Kilby |
| Location | Pecos, Texas |
| Data Center Capacity | 2 GW (scalable) |
| Power Plant Capacity | Up to 2.67 GW (scalable to 5 GW) |
| Deal Duration | 20-year PPA |
| Turbines | 7+ GE Vernova + Solar Turbines |
| Chevron Investment | ~ billion |
| First Power Target | Late 2027 / Early 2028 |
| Construction Jobs | 6,000+ at peak |
Microsoft expects the Pecos campus to add 2 GW of data center capacity over the next five to seven years, with Chevron's plant built to scale to 5 GW. By comparison, a typical large nuclear reactor produces about 1 GW. This is not a pilot project — it's industrial-scale infrastructure.
Why AI Is Driving Big Tech Deeper Into Fossil Fuels
The deal underscores a stark reality the tech industry is facing head-on: AI doesn't just need better models — it needs massive amounts of electricity. Training a single frontier AI model can consume as much power as hundreds of homes use in a year. Running inference for billions of users multiplies that demand many times over.
This comes just weeks after the Federal Energy Regulatory Commission ordered a historic grid overhaul specifically to accommodate AI data center demand — a story we covered in depth on TekMag. The FERC ruling gave AI data centers priority access to the grid, but Microsoft and Chevron are going a step further: bypassing the grid entirely with dedicated behind-the-meter generation.
Microsoft is far from alone in this energy land grab:
- Alphabet recently announced an 0 billion stock offering to fund AI compute and data center expansion, as reported by IMFounder.
- Anthropic secured a deal with SpaceX's Colossus 1 supercomputer, paying an estimated .25 billion per month for 300+ megawatts of compute, per CNBC.
- Nvidia is pushing into the CPU market with "AI Agent PCs" that can run advanced workloads locally, shifting some compute off the grid entirely.
The Climate Contradiction
Here's where things get complicated. Microsoft has publicly pledged to be carbon negative by 2030. This 20-year natural gas deal — running until roughly 2046 — makes that goal substantially harder to reach.
According to the Environmental Integrity Project, the Chevron plant is expected to emit over 13 million tons of CO₂, 3,200 tons of criteria air pollutants, and 278,000 pounds of hazardous air pollutants annually. That's a significant emissions footprint for a company that promised to eliminate its carbon impact this decade.
Microsoft is quick to highlight that the data center itself uses a closed-loop cooling system requiring only an initial water charge — claiming the facility's lifecycle water use is less than that of a typical fast-food restaurant. But critics note the press release conveniently omits the water consumption of the gas power plant itself, which uses significant amounts for steam generation and cooling.
The tech giant has been telegraphing this move for months. For now, Microsoft seems to be making a calculated bet: secure power now, worry about offsets later.
Why This Matters for the AI Industry
The Microsoft-Chevron deal may well become a template for how Big Tech powers the AI era — dedicated, behind-the-meter energy supply built to order, with climate consequences handled downstream. Reuters notes this is the largest collaboration between a U.S. oil and gas major and a tech company, and it likely won't be the last.
AI infrastructure spending is approaching trillion-dollar levels over the next decade, and everyone is scrambling to secure the energy, chips, cooling, and fiber needed to power it. The winners of the AI era won't just build better models — they'll secure the physical resources to run them.
First power from Project Kilby is expected by late 2027 or early 2028 , pending regulatory approvals. At peak construction, the project will employ over 6,000 workers and create hundreds of permanent operational roles in the Pecos region.
Whether Project Kilby becomes a blueprint for progress or a climate setback depends entirely on what happens next. One thing is certain: the era of AI running on wishful energy thinking is over.
Originally published on TekMag
Top comments (0)