Somewhere right now, a data center that could deliver artificial intelligence to entire industries is sitting half-dark. It’s not waiting on code or chips, but on power that’s years away from the grid.
Megawatts have overtaken models as the biggest constraint on AI, and the gap is widening. Hyperscalers have requested more than one terawatt of interconnection requests, while the U.S. grid is projected to add just 40 gigawatts of new capacity by 2030.
What AI needs is specific: Firm, 24/7 electricity, dense enough to sit next to a hyperscale campus. We’ll ultimately need every source of energy, but gas turbines are the fastest deployable source of firm, dispatchable power. Unfortunately, you can’t get one on the timeline the AI buildout is actually moving.
Fifty years of chasing single-digit efficiency gains has locked incumbents into an exotic supply base that physically cannot scale: Single-crystal blade castings from two suppliers in the world, rhenium superalloy (about 60 tons of global supply a year, most already spoken for), rotor disks from fewer than 20 qualified U.S. shops. You cannot build turbines an order of magnitude faster on a foundation like that.
So Parallax is doing what the incumbents can't: Completely rethinking the engine from the ground up. By running a lower turbine inlet temperature, they design the two worst chokepoints (single-crystal blades and rhenium) out of the machine entirely, then rebuild the parts around additive manufacturing with a widely available superalloy. The result is a clean-sheet, ~10 MW gas generator with no exotic materials in it, and a supply chain that can actually scale. It's a deliberate trade: Gen-1 gives up some efficiency against the incumbents to buy that scalability, and a clear materials roadmap narrows the gap over time.
Parallax is emerging from stealth today with $117 million in funding to date, and Eclipse is proud to have co-led the $42M seed with Lux Capital, Founders Fund and Valor. Greylock and General Catalyst led the company’s $75M Series A.
