Google buys Wyoming data center site as hyperscalers move to own AI capacity

Alphabet is buying the site it could have rented. Google has acquired Crusoe’s southeast Wyoming data center site, a day after announcing a $17B investment in the project (). Crusoe is a private AI infrastructure developer, and the search returned no listed shares for it. The deal’s financial terms were not disclosed. Read alongside Microsoft (MSFT)’s pullback from some neocloud projects, it points to a shift in strategy. The largest buyers of AI capacity increasingly want to own it, because the permitting, power and construction risks of renting it have become harder to price. (Prices as of Nov 6, 2026, about 3:28 PM EST.)

From Landlord to Owner

Crusoe was not simply bought out of a finished project. According to (Jul 2, 2026), Crusoe was pressured to leave the Wyoming campus in June after Google raised concerns about costs and timetable. Crusoe has now handed over the pre-development work, which included site selection, government approvals and power strategy. Google is building with Tallgrass Energy, Black Hills Energy, American Electric Power and Related Digital. Crusoe says it is separately developing another Google campus in Armstrong.

That sequence supports the thesis. Google is no longer simply contracting for capacity delivered by an intermediary. It wants direct control of the timeline.

Power Is the Product

The power setup shows what Google is buying. In October, Alphabet (GOOGL) agreed to directly finance a $1.8B, 564-megawatt natural gas facility with Black Hills (BKH), part of a ring-fenced Wyoming microgrid (Oct 8, 2026). The Wyoming project will also draw on a planned 702-megawatt fuel-cell development supplied by Bloom Energy (BE) (). Control of generation, not just buildings, is what lets an owner promise delivery dates. Alphabet guided 2026 capex to roughly $195B to $205B, so it has the balance sheet to carry that ownership.

The Political Squeeze

The midterms made timing risk worse. Of 41 data-center ballot measures this week, all but one went against data centers, according to Ballotpedia. Candidates opposed to expansion also won key races in Texas and Georgia. In Texas, governor-elect Gina Hinojosa has said she would ban new projects until lawmakers pass rules for the industry. A renter exposed to a developer’s permitting or power delays can end up holding capacity that arrives years late. An owner at least controls which sites to pursue, how to sequence them, and how to arrange power.

Microsoft’s Parallel Retreat

Microsoft has not announced a new strategy, but its recent decisions point the same way:

• It declined to exercise an option to more than double its Nscale Texas lease ().
• It tightened oversight of some projects from Nebius (NBIS), IREN (IREN) and Nscale, and is debating whether to sign new neocloud deals ().
• It withdrew from an Essex anchor deal after power connection delays pushed supply into the 2030s.
• Its fiscal quarter capex of $45.2B missed the $50.7B consensus, partly because of timing of third-party capacity ().

Microsoft’s spokesperson said its approach combines owned, leased and partner capacity, which is a reminder that neoclouds are not ruled out. Microsoft also plans about 38 gigawatts of capacity by 2032, (Sep 10, 2026). It still needs a lot of supply from somewhere.

Market Scorecard

Neocloud shares fell Nov 5, with Nebius down about 14%, IREN about 17% and CoreWeave about 10%, according to Reuters. Today’s gains look like relief rather than a change of trend.

Bull, Bear and Bottom Line

Bull case for ownership: more control over deployment, utilization and unit economics, and less exposure to third-party pricing. If Alphabet delivers Wyoming on time, it gains a cost advantage as rivals queue for power.

Bear case: ownership concentrates execution risk on the owner’s balance sheet. Construction delays, cost overruns or an AI demand slowdown would hit Alphabet directly rather than a landlord. Microsoft’s own caution on neocloud costs also shows that tenants can walk. Owners cannot.

The nuance: the Crusoe deal looks less like a rejection of third-party developers than a division of labor. Crusoe did the pre-development work, and Google took the asset once it was de-risked. Neoclouds that can’t secure customers and power will struggle to keep borrowing. Hyperscalers that can finance and operate their own capacity gain leverage.