Sam Altman Says Musk's Orbital Data Center Plans Don't Add Up
OpenAI CEO Sam Altman publicly dismissed Elon Musk's space-based data center ambitions as unrealistic on July 13, 2026, aligning with expert skepticism over SpaceX's orbital compute timeline.
OpenAI CEO Sam Altman and SpaceX founder Elon Musk exchanged public accusations over the viability of orbital data centers, with Altman accusing Musk of misleading investors on timelines for space-based AI compute infrastructure. The dispute, which surfaced on July 13, 2026, reflects a deepening divide between the two figures over how artificial intelligence's enormous energy and processing demands will actually be met. Most independent experts side with Altman.
The Economics Don't Work Yet
SpaceX has increasingly tied its roughly two-trillion-dollar valuation to a vision of orbital data centers performing AI inference tasks from space, operating as a so-called neocloud that bypasses terrestrial infrastructure constraints. The problem is that the numbers don't hold up under scrutiny. Engineers and analysts at companies including Google argue that without significantly cheaper launch costs and mass-produced, high-powered satellites, the model remains commercially unviable.
Musk has said SpaceX will begin flying data center satellites as early as next year, a timeline that depends heavily on the Starship rocket. Starship's 13th test flight is scheduled for July 16. Even if it succeeds, achieving the level of reusability needed to make space-based compute profitable is a separate and harder problem. Current roadmaps suggest full reusability remains out of reach in the near term, which keeps the cost of launching and replacing hardware too high for routine AI workloads.
Cooling GPUs in a Vacuum Is Not a Small Problem
Beyond launch economics, operating high-performance GPUs in space introduces engineering challenges that have no easy solutions. The power density required for AI training or large-scale inference is far higher than what Starlink communication satellites need. Thermal management in a vacuum, without convective cooling, demands entirely different approaches. Analysts generally accept that a proof-of-concept satellite could launch in late 2026, but a commercially viable fleet is unlikely before the 2030s at the earliest.
What This Means for Africa
The orbital data center debate is not abstract for Africa. Terrestrial data center capacity on the continent is concentrated in Lagos, Nairobi, and Cape Town, leaving large parts of Africa dependent on subsea cables that route traffic to Europe or North America with significant latency. In theory, low-orbit AI compute infrastructure could serve landlocked nations or regions with unreliable power grids that cannot support conventional server farms.
In practice, the skepticism from Altman and independent experts points in a different direction. African policymakers and founders should treat the neocloud concept as a long-term possibility, not a near-term plan. Diverting attention from terrestrial fiber expansion and sustainable energy investment toward an unproven orbital model would be a costly mistake. There is also a direct cost barrier: space-based compute would almost certainly be priced in US dollars, compounding the challenge for developers working in currencies that have depreciated sharply, including the Nigerian naira and the Egyptian pound. Sovereign, ground-level infrastructure remains the continent's most reliable path forward, regardless of what Musk and Altman say to each other online.
For now, the AI race runs on earthly power grids, and Africa's best bet is making sure more of those grids are on the continent.
Source: TechCrunch
Written by
StacyAI-assisted news curation. Every story is reviewed by our editors before publication.
