Petra Power fuel cells promise lower fuel bills for data centers

A regular generator takes the long way round. It burns fuel, the burning makes heat, the heat drives moving parts, and the motion spins a generator to produce electricity. Founder Aaron Goodman says each of those steps wastes energy, which is why combustion ends up inefficient.

A fuel cell skips most of that. It uses a chemical reaction between the fuel and oxygen to pull electrons off the fuel directly, and those moving electrons are the electricity. Goodman calls it one step with no loss, at least in theory. In practice no machine is perfect, but cutting out combustion removes most of the places where energy escapes as waste heat and noise.

Petra, founded in 2017, says its cells use much less fuel and produce far lower emissions than traditional power sources. They’re also compact and light, so they take up less room, which matters both in a crowded data center yard and inside a military truck.

Speed is the big selling point. Fuel cells can be installed next to a data center without waiting years for a grid connection. The leader in the field, Bloom Energy, says it got power flowing at an Oracle site in Utah in less than two months. Oracle plans to buy up to 2.8 gigawatts of Bloom systems, with 1.2 gigawatts already under contract.

The market is growing fast. Energy research firm Rystad Energy expects investment in fuel cells for data centers to climb to about $30 billion by 2030, up from $2.8 billion in 2025. It also predicts that roughly 40 percent of planned US data center capacity in 2030 could use dedicated on-site power instead of relying only on the grid.

That’s the opening Petra is chasing. Goodman says it’s working with neoclouds (smaller cloud companies that rent out AI computing power) and other infrastructure providers a step below the biggest players. He won’t name customers yet. The giant hyperscalers, like the biggest cloud providers, are Petra’s ideal customers and it talks to them often, but there’s no firm deal with any of them so far.

Petra’s second customer is the Defense Department. The plan is for the cells to keep a ground vehicle’s onboard gear running while its main engine is switched off, a job known as auxiliary power. Running a big engine just to keep electronics alive wastes fuel and makes noise, so a quiet, efficient alternative has obvious appeal.

So far the company has received nearly $9 million in Defense Department contracts. But Goodman is clear that its cells haven’t been installed on a live vehicle. They’re still in testing, and the government is deciding whether to go further.

You won’t buy a Petra fuel cell for your home. But the shift it represents could touch your life in a few ways:

• Grid strain: if more data centers make their own power, they lean less on the shared grid that also serves homes and businesses.
• Air quality: fuel cells avoid combustion, so they give off lower local emissions than gas-fired generators. They still run on fossil gas in most cases, though, so they’re cleaner, not carbon-free.
• Speed of AI growth: faster on-site power means AI services can expand more quickly, for better or worse.

There are limits. Fuel cell projects still need gas connections, permits, transformers and switching equipment, and those often slow things down more than the cells themselves. Industry manufacturing capacity is also small today, at about 1.8 gigawatts a year, which Rystad expects to rise to 4 gigawatts by 2030.

Petra is aiming for its first data center deployments in 2028, with full-scale production hoped for in 2029. The team is still small but growing, and the company wants to expand its defense work significantly in the next few years. It was picked for TechCrunch’s Startup Battlefield 200, which earned it a spot at the Disrupt conference in San Francisco in October.

The test will be whether a young company can turn lab efficiency into reliable machines at a price that competes with bigger, established fuel cell makers already signing multi-gigawatt deals.