Nvidia is warming to a futuristic idea that has long sounded like science fiction: putting data centers in space. During the company’s latest earnings call on February 25, CEO Jensen Huang said space-based data centers are technically feasible today. The problem isn’t whether they can be built—it’s whether they make financial sense. Right now, he says, they don’t.
Huang’s comments highlight a clear split between engineering capability and real-world economics. In other words, the technology needed to deploy computing infrastructure beyond Earth’s atmosphere is not the main barrier. The bigger challenge is the cost of making it happen at a scale that competes with traditional ground-based data centers.
Still, Nvidia’s outlook isn’t purely skeptical. Huang indicated he expects the equation to shift over time as engineering approaches evolve. That’s an important detail, because it suggests the company sees space data centers not as a dead-end concept, but as something that may become more practical as new designs, better launch economics, improved power generation, and more efficient cooling and hardware strategies mature.
The interest in space-based computing is tied to a broader trend: global demand for AI and high-performance computing is accelerating, and companies are exploring every possible way to expand compute capacity. Data centers are increasingly limited by land, permitting, grid access, and energy supply. Space, at least in theory, offers unique advantages—such as access to abundant solar energy and potentially different thermal management options. But those benefits must outweigh enormous logistical hurdles, including launch costs, maintenance challenges, radiation exposure, and the complexity of operating critical infrastructure in a harsh environment.
For now, Nvidia’s message is straightforward: space data centers can be done, but the cost-benefit picture isn’t there yet. Even so, the door is open. If engineering innovations continue to reduce cost and complexity, the concept could move from “possible” to “practical” in the years ahead—especially as demand for AI compute keeps rising and the world searches for new ways to power the next wave of data center growth.
