Aaron Goodman built Petra Power on a simple idea. Take fuel. Strip electrons directly from it. Generate electricity without the explosions, heat and mechanical losses that plague conventional systems.

The result? Solid oxide fuel cells that promise far lower costs and emissions. Compact enough for tight spaces. Quiet enough to place near people. Petra Power targets two demanding sectors where those traits matter most: data centers strained by artificial intelligence and military vehicles that need reliable auxiliary power.

Founded in 2017, the company employs just 15 people yet has already secured nearly $9 million in Defense Department contracts. Its technology, licensed years ago from NASA, now sits in testing for land vehicles while the firm courts neocloud operators for initial deployments targeted at 2028. Full-scale production? That waits for 2029.

Goodman doesn’t hide the appeal. “Combustion takes fuel and creates explosions,” he told TechCrunch. “Those explosions create heat. That heat creates motion. The motion powers an electromotive force, which eventually creates electricity. It’s a lot of steps, and they all have so much loss in them that you end up with a really inefficient system.”

Petra’s approach skips the middlemen. “What fuel cells do is they take the electrons directly off of the fuel, so they strip the fuel of its electrons, which creates electricity,” Goodman continued. “It’s one step, no loss, and subsequently, in theory, at least, they’re much, much more efficient.”

Companies stand to save significantly on fuel. The cells run on natural gas, hydrogen, propane and more. They operate silently and without odor. Footprint shrinks dramatically compared with generators or turbines.

Data center operators face unprecedented pressure. Projections show electricity demand from these facilities could reach 9.5% to 15% of total U.S. power by 2030, up from about 4.5% today, according to a Lawrence Berkeley National Laboratory study cited by GovCon Wire. AI workloads accelerate the climb. Rack densities that once hovered between 10 and 30 kilowatts now push 80 to 150 kilowatts, with some advanced setups approaching a megawatt per rack.

Utilities cannot respond fast enough. Interconnection queues stretch for years. Transformer shortages persist. Permitting drags on. So hyperscalers and their smaller rivals hunt for on-site solutions. Morgan Stanley analysts see a $240 billion market for behind-the-meter power by 2029, with data centers driving much of the gap. Fuel cells rank high on their scorecard for speed, emissions and load-following ability.

Petra positions itself squarely in that hunt. Its official site claims up to three times the fuel savings versus generators and turbines. It talks of five nines reliability. Systems that deploy in months rather than years. No siting headaches from noise or exhaust. The company even references the next 100 gigawatts of on-site generation arriving for data centers, heavy industry and utilities.

But the customer base remains nascent. Goodman told TechCrunch his firm works with neoclouds and infrastructure providers a few steps removed from the largest hyperscalers. Specific deals stay under wraps. Hyperscalers represent the ideal target given their massive consumption, yet no firm commitments exist yet.

Defense offers a different path. Petra aims to supply auxiliary power units for land vehicles. When the main engine shuts off, onboard electronics and systems still draw electricity. Traditional idling burns fuel wastefully and creates noise. Fuel cells promise efficiency and stealth. The technology currently sits in testing. Government officials weigh whether to advance. Nearly $9 million in contracts already secured, with hopes to expand that work substantially.

The small team continues to grow. Petra graduated from Techstars in 2023 and earned a spot in TechCrunch’s Startup Battlefield 200 this year. Earlier this year it raised a $3.8 million seed round. Those funds support scaling manufacturing and refining the ceramic stacks at the heart of its systems.

Challenges remain familiar to anyone who has watched fuel cell companies. Durability under real-world conditions. Tolerance for fuel impurities. Cost at volume. History shows many promising ventures stumble here. Petra claims its NASA-derived design addresses those historic barriers. Time and deployments will test the assertion.

Broader industry moves lend credence to the moment. Bloom Energy expanded deals with Oracle to 2.8 gigawatts of solid oxide capacity earlier in 2026. American Electric Power committed to up to one gigawatt. Utilities and tech giants alike bet on these systems to bridge the gap until new nuclear or transmission lines materialize later in the decade.

Department of Energy initiatives reflect the urgency. The agency leases federal sites for self-powered AI campuses, some backed by gigawatts of on-site generation that starts with natural gas before transitioning toward advanced nuclear. The military pursues its own microreactor program, awarding billions for deployments across installations.

Petra operates in this ferment. Its fuel cells won’t replace the entire grid. They don’t pretend to. Instead they offer a targeted tool for operators who cannot wait. Compact. Efficient. Fuel-flexible. Deployable where traditional plants cannot go.

Success hinges on execution. First deployments in 2028 will prove whether the theoretical efficiency translates to field performance. Defense testing outcomes will shape military adoption. Scaling production by 2029 demands capital and manufacturing know-how that a 15-person outfit must rapidly acquire.

Yet the demand signal grows louder each quarter. AI compute keeps rising. Efficiency gains in chips cannot fully offset the explosion in usage. Data center power demand may grow at a 24% compound annual rate through 2030, McKinsey analysts project. Supply will likely lag. On-site generation fills part of that void.

Petra Power bets its ceramic devices can claim a slice. Lower bills. Fewer emissions. Smaller physical presence. In an industry racing against power constraints, those attributes command attention. The coming years will reveal if the bet pays off.