Space Compute Satellites: Physically Possible, Economics Still Unproven In vac… · @ryanwang · TwKit

@ryanwang

Space Compute Satellites: Physically Possible, Economics Still Unproven In vacuum, the only way to get rid of heat is through thermal radiation. According to the Stefan-Boltzmann law, the power radiated per unit area goes up with the fourth power of temperature. Raising the radiator temperature from 40°C to 80°C can increase heat rejection by roughly 60%. This makes high-power computing in orbit physically feasible. @elonmusk and @SpaceX have publicly pointed to a design with about 250 kW per satellite, using an optimized Nvidia Vera Rubin NVL72 system, flying in a dawn-dusk sun-synchronous orbit, and preferring to run the chips at higher temperatures around 370 K. Under these conditions, a 250 kW thermal load needs roughly 200 to 300 square meters of double-sided radiator area, and the solar arrays are in a similar range. These sizes are large but not impossible to build. In practice, though, you have to subtract the heat coming in from outside. In low Earth orbit, Earth infrared radiation is around 180 to 240 W per square meter. Once you add albedo contributions, the net heat rejection at 60°C drops noticeably from the ideal number. There is also a 20 to 40°C temperature drop between the chips and the radiators, and the working fluid and system pressure limit how hot you can actually run. All of this makes the radiator area and system complexity higher than the most optimistic early estimates. The main engineering challenge is specific mass, measured in kilograms per kilowatt. Public estimates put a complete satellite, including radiators, solar panels, compute modules, and structure, in the range of 20 to 32 kg per kW. Elon’s stated target of around 10 kg per kW would require very aggressive lightweighting. In microgravity the rack structure can be simplified a lot, but the mass of cold plates, fluid systems, and components that must survive launch loads is much harder to cut. Very low orbits also face atmospheric drag. If the large panels cannot stay edge-on to the flow, the yearly propellant use rises significantly. On the economics side there is still a lot of uncertainty. Starship’s public goal is 100 to 200 dollars per kilogram once it is fully reusable. At that price, launch cost per kilowatt is a few thousand dollars, which remains a meaningful part of the total when you amortize over five years. You also have to compare against the facility and electricity costs of a ground data center over the same period. Even more important are the cost of space-grade solar arrays, the manufacturing cost of the satellites themselves, and the hardware lifetime and failure rate when nothing can be repaired in orbit. None of these have enough public evidence yet to show clear competitiveness. It is worth noting that SpaceX has by far the deepest experience in both launch and satellite operations. Their understanding of how low the specific mass can actually go, real propellant consumption, reliability design, and cost curves is almost certainly more detailed than what outsiders can estimate from public information. External discussion can only conclude that the concept is physically possible but economically unproven. That does not mean SpaceX does not already have clearer internal paths to solving these problems. In short, space-based compute is physically feasible. The heat and energy balance can work. Reaching economic competitiveness still requires low launch costs, low specific mass, acceptable manufacturing costs, and sufficient system lifetime all at the same time. These conditions have not been fully demonstrated in public information, but given SpaceX’s experience, it would be a mistake to underestimate how far along they may already be internally. $SPCX

曝光 238 · 评论 0 · 点赞 1 · 书签 0