Bloom Energy's 1 GW Fuel Cell Deal Lands While GE Vernova Sells Gas Turbine Slots for 2032
Fuel cells sell speed to data centers that can't wait for a gas turbine, while GE Vernova's 2032 order book is priced for the utilities that can.
Fuel cells sell speed to data centers that can't wait for a turbine. GE Vernova's order book is priced for the customers who can.
California Resources and Bloom Energy signed a memorandum on Monday to work toward fuel cell power plants in central and northern California, five days after GE Vernova told investors it had sold gas turbine delivery slots for 2032. Both stories come out of the same shortage. CRC's announcement sets the target at "1 GW or more" of solid-oxide fuel cells (they turn gas into electricity through a chemical reaction, so there's no combustion and nothing spins), with CRC's Carbon TerraVault business storing the captured carbon dioxide underground. The memo is non-binding and names no site, no price and no customer. Data centers appear in a list of possible buyers next to utilities and municipalities. $BE still rose about 4% after hours.
A non-binding memo moves the stock because of 2032. A data center that wants a large gas turbine today is booking for the end of the decade, and fuel cells are the main way to buy onsite power on a timeline measured in months. So the gas power market has split in two. $GEV sells scale, at prices that went up by half this year, and $BE sells time. I think that split puts a ceiling on GE Vernova's pricing for one small slice of its demand and leaves the rest of the order book alone.
The gas turbine wait, in GE Vernova's numbers
GE Vernova's third-quarter release says gas equipment backlog plus slot reservations grew from 116 to 122 GW. A slot reservation is a paid place in the factory schedule, held before the final order is signed. The 10-Q shows how lopsided the quarter was, because the company took 16.6 GW of gas turbine orders while it shipped 5.3 GW. You can see the queue getting longer in the chart below. Orders more than doubled from a year earlier, while shipments barely moved.
A long queue lets the seller raise the price, and GE Vernova has. Scott Strazik, GE Vernova's CEO, said on the call that full-year 2026 gas order pricing per kW is on track to rise 50% from last year. The Power segment's margin went to 21.4% from 13.4% a year earlier, and the 10-Q puts favorable pricing first among the reasons. Customers are paying up front too. Power's contract liabilities (cash customers hand over before delivery) reached $33.2B at the end of September. That's double the December figure.
How fuel cells sell speed to data centers
Bloom's pitch is the calendar. Its data center page offers power "in as little as 90 days," against two to five years or more for a grid connection. For Oracle, Bloom says it delivered a working system in 55 days (see our earlier coverage: Oracle's 2.8 GW fuel cell bet). On the Q3 call, KR Sridhar, Bloom's founder and CEO, said Nebius had its first block of power two months after approvals. The trick is size. Each Bloom server is a 325 kW box, so a campus can add power a block at a time instead of waiting on one 400 MW-class turbine.
The buyers lining up are the ones a turbine can't reach in time. Oracle has contracted 1.2 GW and intends to buy up to 2.8 GW, while Nebius signed for 328 MW. AEP is building a 702 MW fuel cell plant near Cheyenne to sell power to Google, fully running by the end of 2028. It works at small sites too, where 5C's Memphis campus plans to go from 20 to 84 MW within six months on onsite fuel cells, per the Data Center Index.
That demand shows up on one line. Bloom's product revenue was $1.07B in the third quarter, up 179% from a year earlier. Full-year revenue guidance went to $4.4 to $4.6B. Bloom doesn't publish a price per kW, and I haven't found a current independent comparison with turbines.
What fuel cells do to GE Vernova's order book pricing
Strazik said about 20% of GE Vernova's gas demand comes from data centers. The other 80% is utilities and other traditional buyers. A utility building a combined-cycle plant for 2031 load isn't choosing between a heavy-duty turbine and a field of 325 kW boxes. A cloud company that needs power in 2027 is, and I'd guess that's the buyer who'd pay the most for an early slot. So Bloom takes the top off the price GE Vernova can charge a data center for near-dated power, and it does nothing to the 80%.
The overlap is narrower still. The fast product GE Vernova sells is the aeroderivative, a smaller turbine built from a jet engine design that ships sooner and often serves as bridge power. The company took 57 aeroderivative orders in the quarter, and Strazik credited that richer mix for pushing Q3 pricing more than 20% above the fourth quarter of 2025. He also said per-kW pricing will dip next quarter with fewer aeroderivatives in the mix. Those are the GE Vernova turbines a data center weighs against Bloom's boxes, and they carried the quarter's price gain.
The place to watch for Bloom eating into GE Vernova is the 2028 to 2030 slots that don't convert. A reservation isn't a binding order. Slot reservations did fall this quarter, and William Blair's note put that down to reservations converting into firm orders.
The view
I think GE Vernova's pricing holds across the 2032 book, because four out of five gigawatts go to buyers with nothing else to buy at that scale. Bloom's growth comes from the customers a turbine can't serve before 2030. The CRC memo matters once a data center signs a contract under it.
GE Vernova is selling 2032. Bloom is selling the years before it.