Google’s Project Suncatcher to put ML infrastructure in space
Google is launching a prototype satellite to test whether its Tensor Processing Units can survive space radiation and launch stress for future orbital AI infrastructure.
summary is enough The blog post is a high-level announcement, but the comment thread features detailed thermodynamic breakdowns of why space data centers are impractical.
summary
Google launched Project Suncatcher to explore placing machine learning infrastructure in low Earth orbit to exploit continuous solar power. An upcoming prototype mission will test how Trillium TPUs handle launch vibrations and space radiation. However, engineers note that orbital data centers face severe hurdles with vacuum heat dissipation and massive radiator requirements. The project remains a preliminary research moonshot rather than an imminent product.
key points
- The initial test launch evaluates TPU resilience against launch vibrations and proton radiation doses simulating a five-year mission.
- Cooling in a vacuum requires massive radiator panels because airflow is non-existent, making thermal management a primary bottleneck.
- Scaling requires clusters of satellites communicating via high-bandwidth, short-distance laser links.
who should read it
Engineers and researchers tracking aerospace hardware limitations, thermal engineering, and unconventional infrastructure scaling.
discussion skeptical
consensus
Commenters are heavily skeptical about the economics and physics of space data centers, arguing that terrestrial cooling and solar farms are vastly cheaper.
pushback
A minority points out that space infrastructure avoids terrestrial NIMBYism, permitting delays, and local geopolitical energy bottlenecks.
For real. It's an enormous problem solved only with 1) sheer scale, and 2) Science Fiction.
Both of those are expensive as hell, by the way.
Cooling via radiation follows Stefan–Boltzmann: P = εσAT⁴. Let's assume a good surface (emissivity ~0.9) at 300 K (27 °C) at 400 W per square meter per side. A flat panel radiating from both faces into deep space gets 800 W/m, not including the losses from, say, the Sun, or from IR coming off the Earth. Now, input power. Sunlight in orbit ~1,360 W/m², assume ~22% cell efficiency, we got 300 W/m². So each 1 MW compute, 3,300 m² of solar panel and …
It applies the Stefan-Boltzmann law to mathematically demonstrate the immense radiator and solar array sizes required to cool even a tiny 1 megawatt data center in space.
Generated by an LLM from the article and thread. Check the sources before relying on it.