NVIDIA CEO Jensen Huang acknowledged the long-term potential of orbital computing during an appearance on the All-In podcast on Thursday. Huang stated that initial efforts should naturally begin on Earth, but emphasized that preparing for the construction of space infrastructure remains critically important. This concept is gaining increased attention as AI workloads surge, prompting technology firms to explore alternatives beyond traditional, high-energy-consumption data centers.

Huang identified cooling systems as one of the most significant challenges in building data centers in orbit. On Earth, systems rely on conduction and convection to dissipate heat. However, these methods are ineffective in the vacuum of space. "You can only use radiation," Huang explained, adding that this approach requires "very large radiating surface areas" to release heat, which would make the systems complex and costly. While not entirely impossible, he indicated that solving this problem could take several years.

Although space offers abundant solar energy and vast, open areas, the cost of launching hardware and building infrastructure remains a major obstacle. Despite these challenges, Huang noted that NVIDIA has already taken initial steps in space computing. "We are already there," he said, pointing out that the company has deployed CUDA-based systems on satellites for imaging and AI processing tasks. NVIDIA's CUDA system enables developers to utilize NVIDIA GPUs for more than just graphics processing, allowing them to handle complex computational tasks faster than standard processors.

Huang added that processing data directly in space, rather than transmitting it back to Earth, is the logical next step. "That kind of work should be done in space," he remarked. "It will take years," Huang concluded. "That's fine; I have time."

Previously, during NVIDIA's Q4 earnings call, Huang elaborated on his vision for space-based data centers. "The economics of space data centers are not favorable today, but they will improve over time," he stated. While cautious about the short-term outlook, Huang highlighted that space provides "abundant energy" and "ample space" for solar-powered AI satellites. At a recent GPU technology conference, he emphasized the potential of orbital data centers, coinciding with NVIDIA's release of its Space-1 Vera Rubin module.

Huang also mentioned that NVIDIA's THOR chip has received "radiation certification" and noted that the company is already using satellites for image processing.