Google confirmed on November 5 a long-term, $17 billion investment in its first Wyoming data center, in Laramie County near Cheyenne, the same figure Bloomberg reported that day. The money is the headline; the wiring is the story. Much of the campus's power will come from generation built nearby for it, including a 702 MW natural-gas fuel-cell plant with carbon capture that Google calls the largest planned project of its kind. For engineers who have watched data-center schedules bend around grid interconnection queues, this is what going around the queue looks like at gigawatt scale.
Three power plants and a thin wire
The fuel cells are not Google's. AEP announced on October 14 that it would invest about $4 billion in the 702 MW plant, owned by Cheyenne Fuel Cell LLC (about 92% AEP, nearly 8% funds advised by Basalt Infrastructure Partners), and sell the output to Google under a 20-year power purchase agreement. Construction starts in 2026 and full operation is expected by the end of 2028. Data Center Frontier works that out to about $5.7 million per megawatt and notes that the PPA's pricing is not public.
The fuel cells are one layer of three. Black Hills Energy's agreements with Google, effective September 30, cover a portfolio of about 2.7 GW including reserve margins: up to 590 MW of grid-connected service, plus about 2.1 GW of third-party Wyoming resources run through a private microgrid. Black Hills will itself spend $1.8 billion on 564 MW of new gas generation at its Cheyenne Prairie station from 2027 to 2029, backed by $399 million of refundable advances from Google. The third layer is Tallgrass's Cheyenne Power Hub, a project of more than $7 billion whose first phase uses two Mitsubishi M501JAC turbines for about 1,150 MW.
These figures overlap and should not be added up; the public record does not say which megawatts sit inside the microgrid. The shape is clear enough: a large private network with a comparatively thin tie to the public grid. A FERC filing caps flow onto the transmission system at 600 MW, per Measured AI's reading of the filings, against an announced load of 1.8 GW and a design ceiling of 10 GW.
Fuel cells, minus the folklore
Bloom's units are solid oxide fuel cells. They turn natural gas into electricity electrochemically, without combustion, in AEP's description, a distinction from a turbine that is often fumbled: one write-up of the deal says the cells "still burn natural gas." They do not burn it. They do still consume it, and they still emit carbon dioxide. Bloom's argument, as Data Center Frontier relays it, is that the exhaust is a more concentrated CO2 stream, which makes capture simpler than scrubbing turbine flue gas. The hardware can also run on biogas or hydrogen, an option rather than a plan. No efficiency figure has been published for this project.
The practical appeal is lead time. AEP pitches the cells as rapidly deployable power for large loads, and the hardware has been in the pipeline for two years. Bloom and AEP agreed in November 2024 on up to 1 GW of equipment, and in January 2026 an AEP unregulated subsidiary disclosed about $2.65 billion committed to fuel-cell equipment, part of this project's history rather than an addition to the $4 billion. Ordering equipment before the customer signs is how you shorten a timeline.
The price of that speed shows in the per-megawatt numbers. Black Hills' $1.8 billion for 564 MW of gas generation works out to roughly $3.2 million per megawatt by my arithmetic, against about $5.7 million for the fuel cells. The comparison is loose: nobody has said whether the $4 billion includes carbon capture. Google says it will "pay for 100% of the power we use" and cover infrastructure its growth requires, a line aimed at local ratepayers.
The carbon ledger has blanks
Google says a large majority of the fuel-cell plant's emissions will be captured and stored permanently in geologic formations at Tallgrass's nearby Eastern Wyoming Sequestration Hub, and that it is studying capture for Tallgrass's combined-cycle turbines. What nobody has published is a capture rate, a capture-equipment supplier, or the terms for CO2 transport and storage, all listed as undisclosed by Data Center Frontier. The hub itself is still being developed for commercial-scale storage. The turbines, for now, are plain gas. One plainly good detail: the campus will be air-cooled, to spare local water supplies.
A schedule with history
The site has changed hands once already. It began as Crusoe's project with Tallgrass, pitched to scale eventually to 10 GW; Google, after reportedly raising concerns about cost and timeline, took over development, and Crusoe confirmed the sale on November 6 at an undisclosed price. Under Google the plan shrank from five buildings to four on about 716 acres. Black Hills expects initial service in late 2027 and peak demand in 2030, with the fuel cells at full output by the end of 2028.
The risks are ordinary ones at unusual scale: two decades of gas-price exposure under a contract whose price is secret, carbon capture that is designed but not specified, a storage hub not yet in commercial operation, and a community that has started asking questions. More than 200 Laramie County residents turned out at one meeting, and the partners are forming a Community Council in response.
The 30,000-foot read
The grid has become the slowest component in the AI stack, and the hyperscalers have stopped waiting for it. Google's Wyoming campus is less a data center with a power contract than a power system with a data center attached: utility-scale generation, owned by partners, wired through a private microgrid with a modest tie to the public network. That arrangement buys time and, by Google's account, keeps the bill off other ratepayers. The price is a fossil fuel locked into a 20-year contract and a capture system whose specifics remain unpublished.
What Google controls is the money and the contracts. What it does not control is the gas price in 2040, the performance of a sequestration hub that is still being built, or the patience of Cheyenne. Watch three things: AEP breaking ground before year-end, a published capture rate, and whether the first Black Hills power flows in late 2027 as promised. Projects this size rarely fail all at once. They slip a quarter at a time.
Sources: Google blog, "Google's data center investment in Wyoming" ($17 billion figure, partners, 702 MW fuel cells with CCS, Eastern Wyoming Sequestration Hub, turbine FEED study, air cooling, 100% power pledge); Bloomberg (original November 5 report); American Public Power Association (AEP's October 14 announcement: $4 billion, 702 MW, ownership split, 20-year PPA, 2026 start, end-2028 operation, no-combustion description); Data Center Frontier (solid oxide technology, $5.7 million per MW, Black Hills and Tallgrass figures, Bloom procurement history, undisclosed CCS details, campus revisions); Measured AI (600 MW FERC transmission cap, 1.8 GW load, 10 GW design ceiling); TechCrunch (Crusoe sale confirmation, Google's reported cost and timeline concerns); Interesting Engineering (sequestration hub status, community meeting, the "burn" wording). CAVEATS: the Bloomberg article is paywalled and was not read; the $17 billion figure is confirmed by Google's own post. AEP's primary release and the Black Hills release could not be opened in this session, so their figures come via APPA and Data Center Frontier. The 600 MW transmission cap is from Measured AI's reading of FERC filings and was not checked against the docket. The $3.2 million per MW figure for Black Hills is my arithmetic from its reported cost and capacity. Capacity figures for the three power sources overlap and should not be summed. No efficiency, capture-rate or PPA-price figures have been disclosed.