Shares of Flower energy jumped on Monday after striking a deal with Brookfield Asset Management to install fuel cells in artificial intelligence data centers.
Brookfield will spend up to $5 billion to deploy Bloom Energy's technology, the first investment in its strategy to support large-scale AI data centers with power and computing infrastructure. Bloom's fuel cells are “fuel flexible” and can run on natural gas, biogas or hydrogen, the company said.
Brookfield and Bloom are working together to design and build so-called “AI factories” around the world, including a site in Europe that will open before the end of the year.
Shares of Bloom Energy rose more than 20%. Bloom's fuel cells provide on-site power that can be installed quickly because they do not rely on being connected to the grid.
Bloom has already placed hundreds of megawatts of fuel cells through contracts with utilities American Electricity and data center developers like Equinix And oraclesaid the company.
The AI industry's data center plans are becoming increasingly extensive. Nvidia and OpenAI, for example, recently announced a partnership that aims to build 10 gigawatts of data centers, equivalent to New York City's electricity consumption in mid-summer.
However, the AI companies' plans run up against an aging U.S. power grid that is often slow to provide additional power capacity. Data centers are also threatening rising electricity prices for private customers.
Deploying energy solutions “behind the meter” or off the grid “is critical to closing the grid gap for AI factories,” said Sikander Rashid, global head of AI infrastructure at Brookfield, in a press release announcing the deal.
“AI infrastructure needs to be built like a factory – targeted, fast and scalable,” said KR Sridhar, CEO of Bloom Energy, in the press release.
Nvidia CEO Jensen Huang told CNBC last week that the artificial intelligence industry needs to build power from the grid to quickly meet demand and protect consumers from rising electricity prices.
“Data center self-generated power could move much faster than sending it to the grid, and that’s what we need to do,” Huang said.