SpaceX is moving data centres to space, signalling a new era for space-based computing

SpaceX is forging a new frontier in computing with its ambitious 'Starmind' initiative, aiming to deploy up to one million AI-focused orbital data centers. This isn't just about beaming internet; it's a strategic pivot to move intensive AI inference workloads directly into space, leveraging continuous solar power and the natural cold vacuum to meet the insatiable demand for artificial intelligence compute. The company recently filed an FCC application for this colossal constellation, signaling a full-scale assault on the terrestrial data center paradigm. The race to offload AI compute from Earth is heating up dramatically, driven by global power grid strain and mounting environmental concerns over traditional data centers. SpaceX, which merged with xAI in February 2026 and recently completed a historic IPO, is facing fierce competition and regulatory scrutiny. Rivals like Amazon's Project Kuiper (now Amazon Leo) and Blue Origin Project Sunrise have also filed plans for their own orbital data center constellations, with Amazon even formally objecting to SpaceX million-satellite proposal as 'speculative.' Meanwhile, Starcloud has already flown an NVIDIA H100 GPU in orbit, and Google is testing its own TPU hardware in space with Project Suncatcher. Looking ahead, SpaceX plans to launch prototype AI1 satellites by early 2027, with volume production expected by year-end at its new 'Gigasat' factory in Texas. While the promise of near-zero energy costs and reduced latency for specific applications is compelling, significant challenges remain, including radiation exposure, thermal management, in-orbit maintenance, and the immense task of manufacturing millions of satellites. The coming years will reveal whether these orbital ambitions can overcome the harsh realities of space and a crowded regulatory landscape, or if the dream of ubiquitous space-based computing remains just out of reach.