K2 is preparing to send a two-metric-ton spacecraft called Gravitas into orbit, a purpose-built power platform designed to run energy-hungry sensors, radios, and on-orbit computers. The mission places the startup squarely in the middle of a fast-forming market for “space compute” — pushing processing and networking closer to where data is generated rather than beaming everything back to Earth.
Inside Gravitas: A 20 kW Testbed Purpose-Built for Orbit
Gravitas is built around one metric that matters for compute in space: sustained electrical power. K2 says the satellite will deliver 20 kW to hosted payloads, a figure that puts it in the same conversation as some of the most capable telecom buses ever flown. For context, large geostationary platforms like ViaSat-3 are reported above 25 kW, and industry estimates peg Starlink’s latest-generation satellites near 28 kW. Most modern satellites, particularly in low Earth orbit, operate with just a few kilowatts.
To feed that power budget, Gravitas carries a 40-meter solar array span when fully deployed. The platform is less a conventional satellite and more an orbital power plant and lab, making room for a dozen payload modules supplied by multiple customers, including the Department of Defense. Among them: a 20 kW electric thruster K2 expects to be the highest-power electric propulsion system yet flown.
The founders, brothers Karan and Neel Kunjur, previously worked at SpaceX and have pushed aggressive vertical integration — by the company’s count, about 85% of the spacecraft’s components are designed and built in-house. The aim is to control performance, cost, and supply-chain risk while iterating quickly on a high-power architecture.
Why Space Compute Needs More Power to Meet Demand
Orbital data is exploding. High-resolution imaging, wideband RF sensing, and multi-orbit communications generate torrents of information that are increasingly impractical to downlink raw. Shifting inference and pre-processing to orbit cuts bandwidth demand and shortens decision loops for users on the ground. European Space Agency experiments like PhiSat-1 have already shown how onboard AI can filter cloudy imagery and trim data volumes dramatically; operators routinely cite reductions on the order of tens of percent for certain workflows.
That shift, however, pulls the bottleneck into the satellite’s power system. Advanced processors, multi-beam phased-array radios, and optical crosslinks all compete for watts. More power extends link budgets — improving throughput and resilience against jamming — and enables continuous compute rather than duty-cycled bursts. It’s the same logic as terrestrial edge computing: put capability where it’s needed, and feed it with reliable power.
The strategic backdrop is clear. Massive constellations like Starlink and Amazon’s Kuiper are ramping; hyperscale cloud providers are actively evaluating orbital processing to reduce latency and backhaul; and the Pentagon is charting a proliferated architecture for missile warning, tracking, and communications that independent budget analyses have sized at about $185 billion across programs. In each case, power margins determine what missions are possible.
Execution Plan and Mission Milestones for Gravitas
K2’s first mission is structured around three gates:
- Deploy, stabilize, and demonstrate full power generation.
- Operate the hosted payloads and validate the high-power thruster.
- Use that propulsion to significantly raise orbit — a stress test for the platform’s thermal, power, and flight software stack.
New spacecraft rarely have smooth shakedowns, and K2 knows anomalies will be scrutinized. The company emphasizes data collection over perfection, positioning Gravitas as the first of many iterations. Leadership has outlined plans for a rapid cadence of demonstration and commercial flights in the near term and to move into customer production as the architecture matures.
Economics and the Starship Variable Shaping Launch Costs
Big satellites usually mean big launch bills. K2 argues the numbers still work. The company pegs Falcon 9 rideshare costs for a spacecraft in this class at roughly $7.2 million, and prices Gravitas-class platforms at about $15 million — a combination it says undercuts traditional high-power builds while delivering far more capability than equivalently priced smallsats.
The wild card is SpaceX’s Starship. If it achieves routine service and materially lowers cost-to-orbit, a scenario K2 models at around $600,000 for a similar mass, the company plans to scale up even further, turning today’s high-power demo into a modular backbone for orbital data centers. Until then, the bet is that paying more to launch fewer, more capable satellites can close real customer missions now — particularly in high-throughput communications and resilient sensing.
What to Watch as Gravitas Flies and Proves Its Thesis
Three indicators will signal whether K2’s thesis holds:
- Sustained 20 kW delivery to payloads without thermal or pointing penalties.
- High-power electric propulsion firing at the advertised levels.
- Contracted customers returning for larger constellations, testing price-performance against legacy GEO buses and mass-produced LEO platforms.
If Gravitas checks those boxes, it will mark a turning point for space compute — from proof-of-concept AI chips riding along to dedicated, power-rich nodes that look and operate more like edge data centers in orbit. That’s the kind of infrastructure shift that can redefine how, and how fast, space systems deliver value back on Earth.
