Security Architects & CTOs - Orbital AI Compute Claims

𝗔 𝗺𝘂𝗻𝗶𝗰𝗶𝗽𝗮𝗹 𝗱𝗶𝘀𝘁𝗿𝗶𝗰𝘁 𝗷𝘂𝘀𝘁 𝗱𝗲𝘀𝗶𝗴𝗻𝗮𝘁𝗲𝗱 𝗮 𝘀𝗽𝗮𝗰𝗲-𝗰𝗼𝗺𝗽𝘂𝘁𝗶𝗻𝗴 𝗵𝘂𝗯 𝗶𝘁 𝗱𝗶𝗱 𝗻𝗼𝘁 𝗻𝗲𝗲𝗱 𝘁𝗼 𝗯𝘂𝗶𝗹𝗱. 𝗙𝗼𝗿 𝘀𝗲𝗰𝘂𝗿𝗶𝘁𝘆 𝗮𝗿𝗰𝗵𝗶𝘁𝗲𝗰𝘁𝘀 𝗮𝗻𝗱 𝗖𝗧𝗢𝘀, an orbital satellite's headline power figure has moved twice in six weeks, and its most quoted performance claim belongs to a different product entirely. On 31 August, Shanghai's Songjiang district opened a space-computing hub. It did not need to attract tenants. A state-backed satellite constellation operator was already resident there, alongside two satellite manufacturers and a precision-machinery research institute. The announcement formalised a cluster that already existed. Four days later, an orbital AI compute satellite built by a launch provider and a chip maker was still working out what it actually is. Since the partnership's announcement on 24 August, its stated power draw has moved from 120 and 150 kilowatts, to 160 and 250, to 175 and 250, none of the earlier figures retracted. A widely repeated claim that it delivers twenty-five times the AI compute of a previous chip generation actually belongs to a different product the same chip maker announced in March, for six customers that do not include this launch partner. Neither side's technology is shown to be superior here. What differs is what can be checked. Songjiang's tenants are verifiable entities that existed before this week's designation; the American partnership's own supporting detail has moved while I was watching it. Three questions do most of the work when a claim like this reaches your desk: where did the number first appear, has it changed since, and has it been attached to more than one product. Both this week's figures failed one of them. An openly inspectable ground segment is available to any operator willing to build on it instead of around it. GNU Radio handles ground-station signal processing across the industry, and F Prime, the Jet Propulsion Laboratory's own open flight software framework, has flown on small-satellite missions since 2017. Vertical integration is a choice about who controls the stack, not a technical requirement of building one. A New Zealand launch provider put a Japanese Earth-observation satellite into orbit this same week: real launch-site participation, but not tasking authority over what gets observed, and not the AI-enabled analysis that turns a radar return into something a board can act on. Owning the antenna and owning the analysis are separate capabilities. Where has your organisation taken an orbital-compute claim at face value, only to find the number had moved, or belonged to a different product, once you went looking for the source? This is one of seven weekly series in The Hamberger Report. Subscribe on LinkedIn and the next one arrives in your feed. #CyberSecurity #AISovereignty #HambergerReport