Microsoft is securing long-term power agreements with energy companies to fuel its massive AI infrastructure buildout, a shift that suggests the company's AI spending will translate into sustained revenue growth. Rather than hoping energy providers will build capacity and attract customers later, Microsoft is already the guaranteed buyer on 20-year contracts, reducing the speculative risk that has plagued the data center industry.

Why Are Energy Companies Betting on Microsoft's AI Data Centers?

Two major energy deals announced this week reveal how Microsoft is de-risking the AI infrastructure race. Chevron announced Project Kilby, a 20-year power agreement with Microsoft covering 2.67 gigawatts of behind-the-meter capacity in West Texas. The project is moving toward a final investment decision later this year, with Chevron expecting mid-teens returns on the investment. Constellation Energy is pursuing a similar strategy, anchoring a long-term power arrangement with Microsoft through the restart of the Crane Clean Energy Center, which will operate under a 20-year agreement to supply Microsoft's data centers once it returns to service.

"Few competing data center power projects have locked in long-term customer commitments as Kilby has," stated Jeff Gustavson, Constellation Energy's New Energies president.

Jeff Gustavson, New Energies President at Constellation Energy

What makes these deals significant is that Chevron isn't speculatively building power capacity and hoping a hyperscaler shows up later. Microsoft is already the counterparty on a two-decade contract, which is about as far from speculative as an energy deal gets. This approach stands in stark contrast to the broader data center boom, where thousands of projects sit in grid queues waiting for power that may never materialize.

How Is Microsoft Diversifying Its Power Supply?

Microsoft's strategy reveals a sophisticated approach to managing energy risk. By securing power from two very different corners of the energy sector, the company is ensuring that its AI buildout doesn't depend on a single bet. Chevron's gas-fired capacity and Constellation's nuclear restart represent fundamentally different energy sources, each with distinct advantages and regulatory pathways.

• Gas-Fired Power: Chevron's Project Kilby provides flexible, dispatchable capacity in West Texas, allowing Microsoft to scale power consumption as AI workloads demand.
• Nuclear Energy: Constellation's Crane Clean Energy Center restart offers carbon-free, baseload power with minimal operational variability, ideal for continuous AI training and inference workloads.
• Geographic Diversification: By securing power in multiple regions, Microsoft reduces the risk of grid congestion or regional supply constraints affecting its operations.

This diversification strategy contrasts sharply with the broader data center crisis unfolding in states like Texas and New York. Texas has 1,800 data center projects sitting in the grid queue, more than five times the state's peak electricity demand. New York developers are chasing enough power to cover a third of what the whole state uses in a year. The sheer volume of speculative projects has prompted both governors to pause new data center connections while they audit impacts on water supplies and grid stability.

What Does This Mean for Microsoft's Financial Outlook?

Microsoft's earnings report on July 29 provided the proof that investors had been waiting for: AI spending is translating into revenue. The company posted fiscal fourth-quarter revenue of $90 billion, up 18 percent year over year, with Azure revenue climbing 43 percent. Full fiscal 2026 revenue topped $331 billion. Microsoft 365 Copilot crossed 30 million paid seats, and the company's commercial remaining performance obligation, essentially its contracted backlog, grew sharply.

However, the financial picture comes with a trade-off. Microsoft delivered $19.6 billion in free cash flow in its latest quarter, down 23 percent year over year. Management now expects roughly $175 billion in fiscal 2027 capital expenditures, and it's changing how it accounts for some of that spending. Data centers and office buildings will now be depreciated over 25 years instead of 15, and more future leases will be booked as operating leases rather than finance leases.

The accounting shift is important because it affects how the free cash flow picture looks going forward. For investors, the setup is straightforward: the Chevron and Constellation deals suggest the demand side of the AI story is real and multi-year in nature, while the free cash flow trend suggests the cost side is real too and isn't going away next quarter.

How Are States Responding to the Data Center Boom?

While Microsoft is securing power through long-term contracts, state governments are pumping the brakes on the broader data center industry. Texas Governor Greg Abbott announced a data center pause to audit how they will impact water supplies and the power grid, a dramatic reversal from just a year ago when Abbott was championing Texas as the epicenter of the AI industry. New York Governor Kathy Hochul signed an executive order pausing approval of the largest data centers in New York for a year, anything over 50 megawatts.

The political shift reflects widespread public concern about data center impacts, though the trust deficit runs deeper than specific environmental or grid concerns. The backlash has become bipartisan, with governors from both parties facing pressure to look tough on data centers ahead of elections. Abbott is in a statistical dead heat with his challenger, and the polling on data centers has flipped dramatically in Texas.

The key question now is whether these pauses will become opportunities for states to upgrade their grids and deploy clean energy infrastructure, or whether they will simply delay projects indefinitely. Microsoft's approach of locking in power through long-term contracts suggests that the most sophisticated players in the AI infrastructure race are already moving beyond the speculative phase and into the era of committed, multi-decade buildouts.