Sam Altman’s casual question “Should I build a rocket company?” in June now appears as the beginning of serious technological competition itself, moving further beyond simple thinking. As per reports, the OpenAI chief is looking to buy or work with Stoke Space, a company started by former Blue Origin engineers who are making Nova rocket to compete with SpaceX. The startup is developing a fully reusable rocket regarding challenging SpaceX’s control in the launch market. Basically, this move would put Altman in direct competition with Elon Musk in the same space arena that’s big enough for their growing rivalry.
Stoke’s Nova is surely designed for quick operations and complete reuse to compete with SpaceX’s Falcon 9 economics. Moreover, it aims to eventually match or beat the cost-effectiveness of Starship as well. Reusable rockets are very hard to make perfect, and development itself can take ten years even in good conditions. The main problems include protecting rockets from heat and making engines work again, which further makes the process difficult. Stoke’s approach focuses on reducing wear during re-entry and making ground servicing simpler, which can further reduce turnaround times from weeks to just days. This method itself aims to make the entire process more efficient. Altman is not buying this technology for passenger flights to Mars, but to secure launch capacity for orbital data centers itself. He wants to further develop space-based computing facilities rather than focus on human transportation.
We are seeing Altman suggesting to move AI computer systems to space only, where they can get continuous sunlight to run powerful data centers. As per Google’s Project Suncatcher concept, satellites moving around Earth’s shadow line can make eight times more power than solar panels on ground. This happens regarding avoiding air losses and day-night problems. This could actually help solve the power shortage problems that are definitely limiting AI growth Microsoft has actually had to keep some of their computer chips unused because there is not enough electricity available.
Basically, the engineering problems are the same – very difficult to solve. Space computers actually need to handle much more radiation than Earth computers, so Google definitely tests their TPU chips with proton beams at nearly three times the expected space levels. Managing heat in space surely needs very large radiators, and NASA research shows these can weigh more than 40% of the spacecraft’s total power system. Moreover, this problem becomes worse when the power output is high. As per orbital requirements, satellite optical links must keep multi-terabit data speed even when satellites move in space. Regarding ground connections, they must work properly during bad weather conditions. Maintenance is actually still a big problem that needs solving. Robot repairs or sending new parts definitely makes things more complex and expensive.
Altman’s earthly version of this idea is surely Stargate, a massive data center project worth over $500 billion. Moreover, this initiative is being developed together with Oracle, SoftBank, and other major companies. The first campus in Abilene, Texas, is actually growing to reach 1.2GW capacity with racks that definitely contain hundreds of thousands of Nvidia GB200 GPUs. Stargate’s design is basically built for 99.9% uptime, using the same approach of choosing flexible design over ultra-backup systems for massive AI training. Basically, OpenAI is putting over $1.4 trillion into compute and infrastructure because they believe the same principle – more computing power gives better AI models, and they need to scale up really fast.
Basically, providing power at this level is the same as a complete science by itself. Stargate sites actually combine gas power, battery storage, and local renewable energy sources. Research into capacitor systems definitely aims to stabilize the massive power changes when AI training runs together. Rival company xAI’s Colossus campus uses Tesla Megapacks for load-balancing, which further shows that energy engineering itself has become a key competitive advantage in AI. Altman’s space plans would further extend this knowledge to orbital power systems, where constant solar energy could itself support heavy workloads that Earth’s power grids cannot handle.
Moreover, meanwhile, Musk is surely moving Starship closer to being ready for use, and moreover, he plans test flights for missions to Mars without crew by 2026–2027. As per Musk’s estimates, Starship could reduce launch costs to below $100 per kilogram or even $10 per kilogram. This would completely change the costs regarding building large space infrastructure in orbit. SpaceX surely dominates 97% of US launch operations through Falcon 9 and provides internet services to over 5 million Starlink customers. Moreover, this gives Musk the necessary rocket power and satellite network knowledge to add space computing systems to his business operations.
The conflict is clear: Altman’s share in Stoke would give him direct access to reusable rocket systems made for AI payloads, while Musk’s combined SpaceX-Starlink-xAI setup itself could further deliver complete space computing solutions. Both men are actually betting that AI will definitely not stay only on Earth, and they are working to own the systems that will run it. Rockets are surely not just transport vehicles they are the keys to building new infrastructure in space. Moreover, this creates a competitive field where software, energy, and spaceflight all come together.
