The startup pulled its demonstration forward by roughly a year from an originally announced 2028 target, but the binding problem is heat: in vacuum there is no air for fans, and waste heat from dense AI processors has to be radiated away.
In vacuum, a computer chip has only two ways to shed waste heat: radiate it or conduct it into something else. There is no convection, and fans are useless without air. That physical fact is the hidden reason Sophia Space selected Apex's Nova satellite platform for its first on-orbit AI demonstration in 2027: not chip supply, not launch cadence, and not bus price.
SpaceNews reported on June 22 that Sophia, a Pasadena-area startup founded in 2023, will fly its compute module aboard an Apex Nova bus next year, pulling the demonstration forward from an originally announced 2028 target. The company frames the selection as a vote of confidence from a strategic bus supplier. The physics tell a more interesting story.
Sophia's payload is called TILE, for Thermal Integrated LEO Edge. The company says the module is designed to run dense AI inference in low-Earth orbit and to reject the resulting heat without active cooling. Apex co-founder and CEO Ian Cinnamon said Nova is built to host real-time orbital computing missions, according to a June 22 press release. The partnership is therefore not a routine procurement: it is a thermal coupling. Sophia's compute node has to plug into a chassis that can move substantial waste heat from a vacuum-rated processor into a radiator, and the bus has to be designed to accept that thermal load without buckling its own subsystems.
That coupling matters because of who Sophia is pitching. The company says its pipeline includes civil, commercial and defense customers, and it has named the Pentagon's Golden Dome missile-defense initiative as a target. Sophia has argued that turning raw sensor data into decisions on orbit, rather than downlinking it to a ground station, is the kind of capability defense planners have been told they want. Heat rejection in vacuum is the chokepoint that determines whether that capability is real or a slide.
The trade publication Aerospace America profiled Sophia's fanless approach earlier this year, and Sophia's own engineering blog has published on the cooling problem. None of those sources publish independent benchmark numbers for TILE, and Sophia has not released thermal throughput or orbital lifetime specifications backed by third-party testing. For a 2027 demo, that is a non-trivial gap.
The capital side is moving faster than the engineering disclosures. Sophia announced a $10 million seed round in February 2026 and is now finalizing a $7 million Simple Agreement for Future Equity, the convertible instrument commonly called a SAFE, according to the same press release. The new round would bring total funding to $22 million, with SparkLabs Group and Evergreen, described in the release as a network of NVIDIA alumni, among the SAFE participants. Sophia CEO Rob DeMillo has framed the SAFE as the bridge to a larger priced round.
Apex, headquartered in Pasadena, sells Nova as a configurable satellite platform rather than a one-off spacecraft. Its product page positions Nova around payload-agnostic hosting, which is the architectural shape an orbital-compute customer wants. Apex has not publicly disclosed Nova's pricing, payload thermal envelope, or compute-payload capacity as confirmed specifications, so the 2027 demo will also be the first public test of whether the bus can do what Sophia is betting it can.
Two things are worth watching in the next twelve months. First, whether Sophia closes the SAFE and what the eventual priced round looks like, because $22 million is a small budget for a thermal and compute program that has to deliver a flight article in 2027. Second, whether Sophia or Apex publishes independent thermal or compute throughput data on TILE before launch. The story is not really about who is supplying the bus. It is about whether a fanless, vacuum-rated compute module can actually keep a dense processor cool enough to be useful, and whether the bus underneath it can take the waste heat off the table.