Interview with Stephen Hunter, who manages the International Space Station's computer resources, on the tech they use, issues they face, AI, and more
Steve Ranger / ZDNet : Thanks: @steveranger
Context & Ripple Effects
This 2018 ZDNet interview with Stephen Hunter sits at the start of an arc the related coverage only completes years later: Marc Andreessen's case that software eats the physical world eventually reaches orbit, and by 2026 Jensen Huang is talking about data centers in space as a real Nvidia business line. Hunter's account of running computer resources under launch-weight, radiation, and no-onsite-repair constraints is the baseline against which any orbital compute ambition has to be judged.
The interview also lands inside a government thread: Michael Kratsios' framing of federal AI policy shows Washington treating AI capability as strategic infrastructure, and the ISS is one of the few places where that infrastructure physically cannot be patched on demand.
First-order effects
- NASA's ISS operations team, led on this front by Hunter, has to plan compute upgrades around launch schedules and certification cycles rather than vendor release calendars — the interview documents how far behind terrestrial IT that leaves station systems.
- Vendors supplying the station face a customer whose requirements (radiation tolerance, reliability over repairability) invert normal product priorities.
Second-order effects
- If Huang's space data-center ambitions move from talk to hardware, suppliers of hardened, low-power compute gain a second buyer class beyond government stations — commercial orbital operators inheriting the same constraints Hunter describes.
- Terrestrial cloud and edge-computing vendors watch orbital latency-sensitive workloads as a potential adjacent market, with the ISS serving as the reference deployment for what survives up there.
Third-order effects
- If the pattern holds, computing in orbit shifts from bespoke, government-certified hardware toward commercial designs adapted for space — the same commoditization path terrestrial data centers took, with certification bodies rather than hyperscalers setting the pace.
- AI policy frameworks like the one Kratsios outlined would need to extend to assets operating outside national jurisdiction, making orbital compute a regulatory question as much as an engineering one.
The trend: Computing is edging off Earth — from the ISS's constrained onboard resources toward commercially operated orbital data centers — with government-certified hardware slowly giving way to adapted commercial stacks.