Six weeks after announcing a 1.9-gigawatt clean energy deal with Xcel Energy that was widely celebrated as a model for how artificial intelligence can scale without fossil fuels, Google is reportedly exploring on-site gas generation for its 933-MW Goodnight data center campus in Texas, in partnership with Crusoe Energy. While Google doesn’t yet have an offtake agreement in place with Crusoe for the gas, negotiations are reportedly “ongoing.”
The news landed differently than similar moves by other hyperscalers in recent years. Google has spent decades positioning itself as the industry’s clean energy leader — popularizing corporate clean energy PPAs during cleantech 1.0 and more recently pioneering the clean transition tariff, a novel structure that lets large loads fund advanced technologies like enhanced geothermal without burdening ratepayers.
But the Goodnight news is not a first for Google. The company announced last October that it has also agreed to buy power from a 400-MW Illinois gas plant with carbon capture, and the Flatwater Free Press reported in March that the company is eyeing a utility-scale gas facility in Nebraska, a project that Google has not publicly acknowledged.
That said, of the five major hyperscalers and cloud providers that have made significant power announcements in 2026, only Google has made a substantial clean energy commitment. The majority of the others have announced a bunch of fossil gas.
Meta’s Hyperion campus in Louisiana just expanded to 7.46 GW of gas generation after Entergy agreed in late March to build seven additional plants. Microsoft signed a letter of intent for 1.4 GW of behind-the-meter gas in West Virginia and is simultaneously in exclusive talks with Chevron and Engine No. 1 over a proposed 2.5 GW plant in West Texas. Oracle, which is building out AI infrastructure at scale through its Stargate partnership with OpenAI, is powering its Project Jupiter campus in New Mexico with gas turbines.
Amazon is the closest thing to an exception. The company announced 200 MW of solar as part of a broader $12 billion Louisiana data center deal in February, making it the second largest clean energy commitment of the group this year. It has made no significant gas announcement.
Is gas a bridge or foundation?
What makes Google’s moment instructive is less about the company itself and more about the expectations that the market has had for the hyperscaler. Google’s years of clean energy work set a bar its peers never reached for, and that bar is now colliding with the same constraints everyone else is navigating: multi-year grid interconnection queues, years-long turbine backlogs, and the pressure to get gigawatt-scale compute online fast enough to stay competitive.
The Goodnight campus captures this tension directly. One phase of the build-out will co-locate with an onshore wind farm, making it the first test case for co-located generation rules under Texas’ Senate Bill 6, as Latitude Media reported last week. The same site will ultimately house both clean energy and nearly a gigawatt of gas, because that is what the economics currently require.
However, the more important question isn’t whether hyperscalers are using gas: It is what kind of gas commitment they are making: Is it a bridge to relying on more clean power for decades to come, or is it the foundation of their long-term strategy? Most of the capacity in the chart above doesn’t come via a PPA that expires or can be renegotiated. It is physical infrastructure that hyperscalers are financing and building themselves, behind the meter, bypassing the grid entirely.
The Goodnight plant alone would emit as much as 4.5 million tons of carbon dioxide annually, more than the entire city of San Francisco, according to Crusoe’s permit application. Gas plants have operating lives of 20 to 30 years, so once the permits are filed, the turbines are installed, and the capital is deployed, those plants run for decades.
That is what makes the bridge versus foundation question so consequential. Hyperscalers have framed gas as a temporary solution while the grid catches up and cleaner firm power technologies mature, and that framing may be sincere. But bridges built at the gigawatt scale, with sunk costs and 30-year asset lives, tend to become permanent infrastructure.
The industry has not been explicit about what these projects are bridging to, or on what timeline. So far that conversation has played out through permits and satellite images more than in public.
While Google is the only major player that entered 2026 making both kinds of bets — clean and fossil fueled — at scale, the broader pattern is harder to ignore. Four months into the year, the dominant story in hyperscaler power is gas, and most of it is infrastructure that the major players are building themselves.
A version of this story was published in the AI-Energy Nexus newsletter on April 8, 2026, with reporting help from Bianca Giacobone. Subscribe to get pieces like this — plus expert analysis, original reporting, and curated resources — in your inbox every Wednesday.
