May-June 2026 Connect WiT Data Centres in Space Alexander Gerasimov, N4A Analytics Sergey Maltsev, Independent Space Technology Expert “Life should not be simple; it should be driven by passion.” — S.P. Korolev This article examines the idea of deploying data centers in space. While the concept may initially sound like a publicity stunt or another ambitious project promoted by technology giants, it may also reflect a genuine future requirement of the digital economy. The future economy is increasingly tied to the integration of artificial intelligence into the physical world. AI is expected to automate industrial and business processes, not only executing them but also continuously improving them. The greatest economic impact of digitalization is likely to occur in transportation and logistics, where AI can optimize the management of moving assets at every level. Our previous research into large-scale digitalization showed two important trends. First, the economic benefits of digitalization are distributed unevenly across industries, with transportation and logistics receiving the largest share of value creation. Second, digitalization is extremely expensive. In many sectors, the expected benefits barely exceed the required investments, making cost reduction a critical objective. Digitalization costs consist of labor and ICT infrastructure. Artificial intelligence agents may significantly reduce labor costs by automating process redesign and software development. Infrastructure costs, however, remain a major challenge. The United States alone has announced more than $800 billion in planned investments in AI-focused data centers. Operators hope that larger facilities will reduce the cost of computing. However, utilization rates remain a critical factor. Even the largest AI clusters often operate far below optimal capacity, dramatically increasing the cost of computation. An ideal ICT architecture would combine core and edge computing resources connected through an intelligent orchestration system capable of dynamically balancing workloads. Such an approach would reduce network requirements and move latency-sensitive processing closer to the controlled object. Edge computing is commonly associated with 5G MEC deployments. However, global AI-driven transportation and logistics require broadband connectivity across the entire planet, including oceans. In practice, low-Earth-orbit satellite constellations appear to be the only realistic solution capable of providing such coverage. If satellites become the primary communication platform, edge computing must also move into space. Real-time AI applications often require end-to-end latency below 10 milliseconds. This is difficult to achieve using traditional relay satellites alone. A more practical approach is to distribute AI agents between onboard systems and orbital edge-computing platforms. In this architecture, latency-sensitive processing takes place on the vehicle itself and on nearby orbital computing nodes, while less time-critical data is transmitted to terrestrial hyperscale data centers for model training and long-term analytics. From a technical perspective, adding computing satellites to existing low-Earth-orbit constellations is feasible. Modern satellites already use optical inter-satellite links that could connect communication satellites with dedicated computing spacecraft. These orbital computing nodes would be larger than conventional relay satellites but far smaller than the “mega data centers” often described in public discussions. Several major engineering challenges remain. The most significant are thermal management and radiation protection. Space offers abundant solar energy, but heat dissipation in a vacuum is difficult and requires large radiator surfaces. Radiation can also cause memory errors and hardware failures, making protective technologies essential. Orbital computing remains significantly more expensive than terrestrial infrastructure. Estimates suggest that one gigawatt of computing capacity in orbit may cost several times more than an equivalent deployment on Earth. To close this gap, launch costs and spacecraft manufacturing costs must decline dramatically through mass production and reusable launch systems. Despite these disadvantages, the economics of the entire system—not just a single component—must be considered. Orbital edge-computing nodes could reduce requirements for onboard hardware, lower network bandwidth consumption, simplify system architecture, and allow hyperscale data centers to be built in locations with abundant and inexpensive energy. The effectiveness of end-to-end orchestration may become even more important than the cost of the satellites themselves. A unified architecture operated by a single provider could simplify workload management across communications, edge computing, and core data centers. Several companies have already announced ambitious plans in this area. SpaceX has reportedly proposed a distributed orbital AI infrastructure with a target capacity of 100 GW, potentially consisting of up to one million spacecraft connected through laser communication links. Google, Planet Labs, Starcloud, Blue Origin, and Chinese space technology companies have also announced various orbital computing initiatives. Many of these projects are likely to be scaled back, delayed, or canceled. Nevertheless, the underlying demand for orbital edge-computing systems appears real. Their primary role may not be the training of large language models but rather the real-time control of physical systems through AI agents. Looking ahead, a global distributed computing infrastructure with AI-enabled edge nodes in space begins to resemble concepts once found only in science fiction. The necessary investment, technological momentum, and business incentives already exist. For that reason, technological progress must be accompanied by equally significant progress in social and economic institutions. Advanced technologies should serve humanity rather than threaten it. The challenge is not only to build powerful systems, but also to ensure that they are used responsibly.