Space AI Elon Musk Wants 1 Million Satellites, While Sam Altman Calls It "Ridiculous."
The Space Race for Intelligence: Elon Musk’s Orbital AI Data Centers vs. Sam Altman’s Skepticism
The rivalry between Elon Musk and Sam Altman has reached a new frontier: Outer Space. Following SpaceX ambitious filing to launch one million satellites to support AI processing in orbit, the tech world is buzzing with the potential and the practicality of moving the world’s compute power into the stars.
Musk’s Vision: Harvesting the Sun
Elon Musk’s plan involves a strategic merger of resources between SpaceX and xAI. The goal is to build a massive, solar-powered data center network in space to meet the skyrocketing energy demands of AI. Musk has frequently warned that Earth’s current power grids are insufficient for the "AI era," and moving data centers to space offers 24/7 access to unfiltered solar energy without the environmental constraints of terrestrial land.
Altman’s Reality Check: "Ridiculous" for Today
During his recent visit to the India AI Impact Summit, OpenAI CEO Sam Altman didn't hold back when asked about Musk’s orbital ambitions. Altman labeled the idea of space-based data centers as "ridiculous" within the current landscape.
While he didn’t entirely dismiss the possibility of space-based compute in the distant future, Altman emphasized that the current barriers including exorbitant launch costs, the impossibility of on-site hardware repairs, and the complexity of data latency make it unfeasible for at least the next decade.
Google’s Silent Stake: Project Suncatcher
Musk isn't the only one looking upward. Last year, Google unveiled Project Suncatcher, a research initiative exploring the possibility of deploying Tensor Processing Units (TPUs) in space. However, unlike SpaceX, which has the proprietary rocket technology to make mass deployment possible, Google’s project remains in the early conceptual and testing phases.
- In space, there is no air to cool the fan (Convection). Cooling must use radiation (Radiation) only. This is very difficult for AI chips to overheat. Space has an extremely cold background temperature near absolute zero. If the Thermal Management system is successfully designed, it will be a paradise for high-performance computing.
- Sending data from Earth to be processed in space and then sent back down (round-trip latency) continues to hinder applications that require millisecond speeds, such as self-driving cars. or remote surgery, so AI in space may be suitable for this kind of work "Asynchronous Training" or training large models that do not require immediate results.
- Musk's comments about Earth's lack of energy are well-founded. According to the IEA, by 2026 data center energy consumption around the world could double (equivalent to the electricity consumption of the entire country of Japan). Using solar power in space that is eight times stronger than on Earth is a very tempting solution.
- Launching 1 million satellites raises concerns about Kessler Syndrome, which Sam Altman might consider a legal and ethical risk. It's "not worth it" compared to building nuclear fusion plants on Earth to power AI instead.
Source: Business Insider

Comments
Post a Comment