Elon Musk and the SpaceX team have been thinking about how to use the power of artificial intelligence (AI) in outer space for years. They take their first cue from Iain Banks’ brilliant sci-fi Culture series. The vision discusses deploying AI satellites that might transform large-scale data processing and storage. The economic reality of these kinds of joint public-private ventures poses daunting challenges. At a lifetime of something like five years for satellites, investors are under tremendous pressure to realize returns on their investments quickly.
Even as the demand for satellite data continues to grow, the logistical challenges of eventually launching hundreds of thousands of satellites are still very daunting. Matt Gorman, CEO of Amazon Web Services, emphasizes the current limitations, stating, “There are not enough rockets to launch a million satellites yet, so we’re pretty far from that.” This limitation presents an obstacle for bold companies such as Starcloud and Google. They are fully motivated to roll out complex satellite constellations, but encounter major roadblocks in the process.
The Cost of Satellite Deployment
The economics of launching satellites into orbit are further muddled by expensive costs to launch as well as technological constraints. Satellite manufacturing is the largest part of those costs. Yet, if companies can produce high-powered satellites at about half the cost of existing Starlink satellites, it might alter the financial calculations for orbital data centers.
NASA’s Andrew McCalip has created a Space vs. Terrestrial Data Center Cost Calculator that helps communities understand the potential costs of terrestrial vs. space-based data centers. He notes that “A FLOP is a FLOP, it doesn’t matter where it lives,” highlighting that the performance metrics remain consistent regardless of location. Recognizing this perspective reveals an important opportunity. In short, if we are able to greatly reduce the cost of doing business in space, the return on that investment would be much greater.
Google’s Project Suncatcher hopes to allay these economic fears by deploying prototype vehicles in 2027. In particular, the STEP initiative aims to understand the benefits and challenges of leveraging space as an AI processing platform. The following year, in 2025, Project Suncatcher published a white paper deeply illuminating the comparative power costs of terrestrial versus space based data centers. Those savings largely reflect annual expenses for ground data centers, ranging from $570 to $3,000 per kilowatt of power, depending on local costs and the efficiency of the systems used.
Technical Challenges in Space
While the promise of cost savings and efficiency is enticing, a number of technical hurdles need to be overcome. Philip Johnston, CEO of Starcloud, points out that “Training is not the ideal thing to do in space.” He posits that inference tasks can be handled fairly easily with dozens of GPUs (even on just one satellite). Yet there are increasingly insurmountable challenges to training AI models due to environmental and energy constraints.
Johnston further states, “After five or six years, the dollars per kilowatt hour doesn’t produce a return, and that’s because they’re not state-of-the-art.” This prevalent feeling among the public emphasizes the need for continual improvements in technology, to make sure that investments continue to pay off years down the line.
Mike Safyan elaborates on another critical concern: “You’re relying on very large radiators to just be able to dissipate that heat into the blackness of space, and so that’s a lot of surface area and mass that you have to manage.” As a result, efficient thermal management development, testing and implementation becomes crucial in maintaining stable, reliable and lasting satellites that will be deployed into orbit.
Future Prospects and Predictions
Even with these challenges, industry leaders are confident about where orbital AI will take us. Danny Field expresses enthusiasm for the potential evolution of this sector: “I’m excited to see how some of these companies get to a point where the economics make sense and the business case closes.” This optimism reflects a growing belief that with innovation and investment, the economics of orbital AI could become more favorable.
Elon Musk has been vocal about his confidence in this future, asserting, “By far the cheapest place to put AI will be space in 36 months or less.” This claim suggests a significant shift in perspective regarding the viability of orbital AI as technological advancements continue to unfold.
xAI’s head of compute predicts an ambitious future where “1% of global compute will be in orbit by 2028.” If accurate, this projection could signal a paradigm shift in how data processing is managed across industries, making space-based solutions increasingly relevant.