Ethos Space Resources, a two-person startup in Y Combinator's Summer 2026 batch, says it will extract silicon from lunar dust to build terawatt-scale data centers in space. The Los Angeles firm claims $1 billion in letters of intent and a 2028 Moon mission backed by a completed NASA contract, according to its Y Combinator profile, though these claims have not been independently verified in government awards databases or third-party press.
The pitch is straightforward, if audacious. Ross Centers, CEO and a lunar resource geologist, and Brogan BamBrogan, CTO and SpaceX employee #23 who later led engineering at Hyperloop One, argue that artificial intelligence infrastructure at scale cannot be built on Earth. The grid won't support it. "We make silicon on the Moon, to power everything in space," the company's Y Combinator page states. Their process: melt lunar regolith in hard vacuum, pass current through it, and liberate silicon, aluminum, and oxygen.
Whether that translates to revenue by decade's end is another matter entirely.
The timing reflects real pressure on terrestrial energy systems. The International Energy Agency projected in 2024 that global electricity consumption by data centers, AI workloads, and cryptocurrency would reach 620 to 1,050 terawatt-hours in 2026, up from 460 TWh in 2022. Bloomberg New Energy Finance reported in July that U.S. data center capacity would hit 118 gigawatts by 2030 but flagged a 63-GW shortfall against AI chip shipments by 2033 due to energy constraints. Power is the bottleneck.
"The moon has infinite power, infinite cooling, and infinite mineral resources to manufacture these data centers," Centers said in a July 2025 interview. The argument has a certain logic. Lunar regolith contains 40 to 49 percent silicon dioxide by weight, equivalent to roughly 19 to 23 percent elemental silicon, NASA data shows. The lunar environment offers vacuum and temperature extremes that certain manufacturing processes require anyway. No atmosphere to cool, no gravity to fight.
Space-based compute proposals have accelerated in 2026. Orbital, a California startup, raised $5 million in a pre-seed round in June and filed with the FCC in July for up to 100,000 satellites carrying 10 GW of compute capacity. Axiom Space announced orbital data center nodes in April 2025. Sophia Space partnered with Kepler Communications in 2026 to demo NVIDIA-powered edge compute on an optical relay starting in the fourth quarter. The U.S. Government Accountability Office published a Science & Tech Spotlight in May outlining feasibility and constraints: power, radiators, spectrum allocation, and orbital debris rules all limit near-term scale.
What Ethos describes is molten regolith electrolysis, a process NASA, Blue Origin, and Houston-based Lunar Resources have demonstrated at lab and pilot scale since 2024. Ethos did not disclose funding amount or investors. The company said it completed a NASA contract but did not specify the program or value.
NASA's Kennedy Space Center produced molecular oxygen and metals in December 2024 by running 25 kilograms of lunar simulant through an LR-1 reactor from Lunar Resources in a vacuum chamber. "This is the first time NASA has produced molecular oxygen using this process," Dr. Annie Meier, the Molten Regolith Electrolysis project manager, said in a May 2025 article updated in June 2026. Evan Bell, mechanical structures and mechatronics lead at KSC, added that the process "can produce the oxidizer, which is half of the propellant mix, and it can create vital metals used in the production of solar panels that in turn could power entire lunar base stations."
Blue Origin's Blue Alchemist program passed a critical design review in September 2025. The integrated system claims to produce silicon solar cells, aluminum wire, oxygen, iron, and slag from regolith simulant. "Blue Alchemist changes everything about how we approach space. It is the foundation for a sustainable robotic and human presence across the solar system," the company wrote. Blue Origin targeted an autonomous demonstration in a simulated lunar environment in 2026, though public verification of that milestone has not yet appeared. NASA's TechPort lists the work under an ISRU Power Tipping Point award.
Metals emerge in sequence as the melt breaks down: iron first, then silicon, then aluminum, NASA project notes said. Ethos published an internal calculation in July 2026 estimating 75 to 90 kilowatt-hours per kilogram to purify lunar silicon to 6N grade. Fraunhofer ISE reported in 2025 that terrestrial solar-grade polysilicon production required 60 to 71 kWh per kilogram in efficient plants. Older Siemens routes consumed more.
The energy budgets are comparable, at least on paper.
Manufacturing at the Edge of Darkness
If silicon production proves viable at scale, it unlocks photovoltaic manufacturing on the Moon without launch-mass penalties. Lunar poles offer peaks of near-eternal light where elevated solar arrays could generate continuous multi-megawatt power, recent modeling in npj Space Exploration suggested. A 2026 paper proposed solar-thermal collection with regolith thermal storage to bridge the 14-day lunar night, reducing battery mass for smaller installations.
Parallel efforts target other pieces of lunar infrastructure. ICON holds a $57.2 million NASA SBIR Phase III contract running through 2028 to mature regolith-based construction for landing pads and habitat shells under its Project Olympus. Astroport Space Technologies received NASA STTR Phase II funding in 2025 for a "Brickbot" that melts and sinters regolith tiles for pads and roads. OffWorld adapted terrestrial mining robots for lunar excavation in a 2026 NASA TechPort entry. Starpath Robotics won a 2024 NASA "Break the Ice" challenge with a compact robot for icy-regolith excavation.
NASA's carbothermal reactor program, developed with Sierra Space, demonstrated integrated solar collection and chemical reduction of simulant in 2026 prototype tests at Johnson, Glenn, and Kennedy centers. The system uses concentrated sunlight to extract carbon monoxide and oxygen. NASA lists molten salt electrolysis, carbothermal reduction, and MRE as competing pathways in a 2026 review in the journal Space: Science & Technology.
The field is crowded, in other words. And not all approaches will survive contact with lunar dust and thermal cycling.
The Legal Landscape
"US law is very clear that if you make something on the moon, that's yours," Centers said in July 2025, adding that "international law agrees." Title 51, Chapter 513 of U.S. Code, passed in 2015, recognizes rights of U.S. citizens to recover space resources, consistent with international obligations. Luxembourg enacted a space resources law in 2017. Japan followed in 2021. The Artemis Accords, signed by 71 nations as of August 2026 according to NASA, support resource utilization within Outer Space Treaty constraints, which bar territorial claims but leave extraction rights debated.
Some space lawyers dispute that "international law agrees" without qualification. The Outer Space Treaty prohibits national appropriation of celestial bodies. Whether commercial extraction constitutes de facto appropriation remains contested in UN Committee on the Peaceful Uses of Outer Space proceedings. The Artemis Accords are voluntary norms, not treaty law, and non-signatories—including Russia and China—do not recognize their framework.
NASA's Artemis program has shifted timelines repeatedly, with Artemis II's crewed lunar flyby slipping from its original 2024 target to 2026, and subsequent missions adjusting accordingly. In 2026, the agency restructured Artemis III as a low-Earth-orbit systems demonstration, deferring the first lunar landing to Artemis IV in 2028. CSIS Aerospace Security wrote in May that schedule risks persist in human lander development and life-support systems.
The regulatory scaffolding is still under construction. That creates both opportunity and risk for early movers.
The Core Question
Ethos's $1 billion in letters of intent and 2028 Moon mission claim sits at the aggressive end of space-hardware projections. Those figures appear on its Y Combinator page and company site but have not been corroborated in government awards databases or third-party press. Blue Origin's Blue Alchemist has a documented CDR milestone and NASA backing. Ethos, with two people and undisclosed funding, has not shown hardware prototypes or named LOI partners publicly.
The technical path is clearer than the commercial one. MRE produces oxygen and metals. Purifying silicon to photovoltaic grade requires energy budgets near terrestrial norms if Ethos and Blue Origin calculations hold. NASA's December 2024 vacuum test and integrated carbothermal prototypes in 2026 show the chemistry works. TRL advancement to flight hardware demands years of thermal cycling, dust mitigation, and autonomous operations testing. NASA's 2026 budget documents flag dust accumulation on vertical solar arrays and power distribution as pacing items for Artemis surface operations.
The market case rests on whether space data centers materialize at scale. GAO's May spotlight noted that a satellite with a few hundred kilowatts of compute would need a 10,000-square-foot solar array and matching radiators. Mass and volume collide with launch economics and orbital debris rules. The FCC's 2022 five-year deorbit mandate and tighter ITU coordination requirements from the 2023 World Radiocommunication Conference constrain mega-constellation designs. Early orbital compute pilots target inference workloads for defense and latency-sensitive applications, not training at terawatt scale.
If energy constraints on Earth persist and launch costs fall further, manufacturing solar panels and compute infrastructure from lunar silicon shifts from science fiction to engineering problem. Ethos frames its process as a "silicon-first flywheel" where early ISRU output powers larger ISRU plants, eventually feeding mass drivers that launch finished satellites from the Moon. The concept dates to 1970s studies but remains unproven at any scale. The company's July 2026 "Dyson Supply Chain" note pegs the total addressable market of a Dyson swarm at "two billion times larger than terrestrial GDP," a figure that underscores the gulf between vision and near-term milestones.
Centers and BamBrogan bring pedigree: Colorado School of Mines, SpaceX's Kestrel engine and Dragon heat shield, Hyperloop One. Whether a two-person team with undisclosed funding and a 2028 launch target can leapfrog incumbents with flight heritage and NASA partnerships is the question. Lunar Resources, with its LR-1 reactor already tested in KSC vacuum chambers, and Blue Origin, targeting a 2026 autonomous demo, have narrower near-term goals and more visible traction.
Investors in space resources and cleantech should note that LOIs are not revenue. 2028 is twenty-six months away in an industry where schedules slip by years. The IEA's 2024 projection of 620 to 1,050 TWh in data center demand and BNEF's 63-GW U.S. capacity gap signal real pull for off-grid energy solutions. Whether that pull translates to hundred-thousand-satellite constellations powered by lunar silicon, or to terrestrial nuclear and renewables with modest orbital edge compute, remains the decade's defining infrastructure question.
Ethos is betting the Moon gets there first. The odds are long, but perhaps not as long as they were five years ago.
