Parallax Raises $117M for AI Data Center Turbines

A data center does not earn anything from a turbine promised for the 2030s. Parallax emerged from stealth on September 24, 2026 with a $117M fundraise to build around a different priority: manufacturing throughput.

Founder and CEO Carl Schoeller says Parallax is developing an approximately 10 MW gas turbine for onsite AI data-center power. Founders Fund, Eclipse, Lux Capital, Diffusion, Greylock, and General Catalyst backed the financing, but Parallax has not disclosed a round label, valuation, ownership terms, or individual check sizes.

The company’s thesis is a deliberate engineering trade-off. Parallax plans to use 3-D printing, fewer parts, lower combustion temperatures, and more common materials to manufacture turbines faster, even though that approach gives up some fuel efficiency. The $117M is financing the attempt to prove that an available turbine can be more valuable to a data-center operator than a more efficient machine stuck in a multiyear production queue.

What Parallax Announced

Parallax describes itself as “The AI Energy Company,” and the company is beginning with smaller gas-fired turbines designed specifically for data centers. The Wall Street Journal reported that the California startup was founded in 2026 and intends to make units capable of generating about 10 MW, small enough to move in shipping containers and deploy in multiples at sites requiring hundreds or thousands of megawatts.

The financing is best described as a $117M fundraise. Schoeller’s first-party announcement names all six firms as leaders, while independent reporting and investor confirmation corroborate the amount and participation. No source reviewed established whether the capital represents one round or multiple closings, so the event should not be labeled Seed, Series A, or another stage.

Schoeller previously co-founded Theseus, a defense-technology company that developed GPS-denied drone navigation. TechCrunch reported that Theseus raised a $4.3M Seed round in 2025 after a 3-D-printed hackathon prototype drew attention from drone manufacturers and U.S. Special Forces. That history establishes Schoeller’s prior hardware experience; it does not remove the much larger technical and commercial burden attached to a gas turbine.

The Machine Parallax Is Choosing

Gas turbines built for utilities and aircraft have been optimized for efficiency, long operating lives, and demanding performance. That architecture can require exotic alloys, extreme combustion temperatures, specialized manufacturing, and a core assembled from more than 2,000 parts.

Parallax is choosing a different set of constraints. The company says a 3-D-printed core can reduce those thousands of parts to a handful, while lower combustion temperatures permit the use of more common materials and broader aerospace supply chains. Schoeller’s launch thread says the design uses materials with a global supply five orders of magnitude greater than those found in current turbine blades and reduces core part count by three orders of magnitude. Those figures are company claims and have not been independently validated.

The trade-off is explicit. Lower combustion temperatures reduce fuel efficiency, which can increase operating costs over the machine’s life. Parallax is betting that faster manufacturing, easier sourcing, modular transport, and a shorter path to delivered capacity will compensate for that penalty in a market where the first problem may be getting any suitable turbine at all.

Why AI Data Centers Care About Delivery Time

AI infrastructure has tied compute procurement to energy procurement. A developer can secure chips, buildings, and capital while waiting years for transmission upgrades, grid interconnection, and generation equipment. The power system’s schedule can therefore decide when the compute asset begins earning revenue.

The demand backdrop is substantial. The International Energy Agency’s current outlook estimates global data-center electricity consumption could roughly double from 485 TWh in 2025 to about 950 TWh in 2030. It also estimates that approximately 15 GW to 27 GW of onsite natural-gas generation could serve data centers by 2030, mostly in the United States.

U.S. estimates carry a wide range because hardware shipments, utilization, and cooling choices remain uncertain. The Department of Energy and Lawrence Berkeley National Laboratory estimated that American data centers could consume 325 TWh to 580 TWh in 2028, equal to 6.7% to 12% of total U.S. electricity consumption. That range explains why capital is moving into several approaches at once, including DevCuration’s coverage of TAR’s off-grid AI power strategy and Bluecore Energy’s floating nuclear plan.

What the $117M Still Has to Prove

Parallax is targeting a first prototype by the end of 2026, testing in 2027, and customer deliveries in 2028. Those dates describe a development plan, not a completed machine or contracted operating fleet.

The technical proof will include sustained output, thermal performance, uptime, maintenance intervals, component life, manufacturing yield, and repeatability across units. The commercial proof is broader: customer orders, permitting, fuel supply, emissions compliance, site integration, service capacity, and the lifetime cost of accepting lower efficiency in exchange for earlier delivery.

None of those outcomes is public yet. Parallax has not disclosed customers, orders, revenue, valuation, a production location, independent performance data, or a current CTO. The absence is not unusual for a company emerging from stealth, but it marks the distance between a fundable thesis and infrastructure a data-center operator can place on a construction schedule.

A New Capital Market Around the Power Date

Parallax’s financing shows how the AI buildout is changing the unit of competition. The question is no longer only which chip, model, or data-center design performs best. Investors and operators are also underwriting which combination of generation equipment, grid access, materials, manufacturing, permits, and service can produce usable power on the date the compute needs it.

That makes Parallax’s apparent weakness part of its proposition. A turbine that burns more fuel for each megawatt-hour can still create value if it is manufactured sooner, installed predictably, and maintained reliably while an alternative remains unavailable. The economics will depend on how much that schedule advantage is worth after fuel, emissions, maintenance, and site costs are counted.

The $117M gives Parallax room to move from a supply-chain argument to a working industrial system. The company’s next chapter will be written at the test stand and the first data-center site, where thousands of consolidated parts, common materials, and an aggressive delivery promise must survive heat, rotation, regulation, and the customer’s uptime requirement at the same time.