Tech Explained: Space-based Data Centres: The Next Frontier of AI Infrastructure in Simple Terms

The most valuable private company in the world ever is in the making. Elon Musk’s SpaceX is set to take over his artificial intelligence (AI) start-up, xAI, best known for its Grok chatbot. Reports estimate xAI’s valuation at at least around $125 billion and SpaceX’s at $1 trillion.

The deal is not just about consolidating Musk’s sprawling business empire, it also reflects a strategic, forward-looking vision for technology development. As per Musk, this combination would form an ‘innovation engine’ placing AI, rockets, space-based internet and media under one roof. He feels that in the long term, space-based AI is the only way to scale and wants to make space-based data centres a reality.

An AI data centre is a facility designed to house the specialised information technology-associated infrastructure required to train, deploy and run AI applications and services. It features advanced computing, networking and storage architectures, along with high-capacity power and cooling systems to support intensive AI workloads.

Unlike their traditional counterparts, these data centres are purpose-built for cloud, AI and machine learning tasks, relying on high-performance graphics processing units (GPUs) rather than central processing units. Deploying GPUs requires significantly more physical space.

Organisations often rely on either hyperscale or colocation data centres for AI workloads, with the former frequently housing over 5,000 servers across at least 10,000 square feet. Colocation involves companies leasing space, servers and bandwidth within hyperscale facilities, allowing businesses to access AI-ready infrastructure without the upfront investment.

There is a realisation that various data centres are energy guzzlers. Large data centres are already consuming approximately 1.5% of global power. This figure is expected to rise sharply as models such as GPT-6 and Llama 5 are trained, which will require multi-gigawatt clusters. The pace and scale with which AI is developing is expected to put a tremendous amount of pressure on an already-strained global energy grid.

In addition, there are issues like the requirement of cooling AI data centres on earth, which are both energy and water-intensive. A major amount of freshwater is consumed during cooling processes. Also, hyperscale facilities demand major investments in land acquisition. Furthermore, there is the major issue of carbon emissions. A space-based data centre could reduce carbon emissions by up to tenfold compared with a traditional land-based facility.

Hence, there is a view that future data centres could be located in space.

There are some companies that have already started working on such ideas. Generally speaking, a space-based data centre need not necessarily look like a building or a complex. It is expected that such centres could be established as a modular constellation of satellites. These networked satellites with solar arrays would operate together in low earth orbit. In sun-synchronous orbit, they could draw uninterrupted energy from enormous solar arrays. Their GPU-heavy computing hardware would need to be housed in shielded, radiation-hardened modules.

Several major tech companies, private space firms and national agencies are racing to develop AI data centres in orbit to overcome the earth’s growing constraints around energy, cooling and land. The core appeal lies in constant access to solar power and natural, water-free cooling. These efforts aim to support energy-hungry AI workloads, while reducing reliance on terrestrial infrastructure.

As per reports, SpaceX has proposed leveraging its Starlink network to deploy large-scale orbital computing, while Google’s Project Suncatcher and Blue Origin’s long-term plans focus on sun-synchronous constellations and gigawatt-scale space infrastructure. The Nvidia-backed Starcloud and Axiom Space are already testing on-orbit AI computing. A few startups are also trying their luck in this field.

Apart from the US and EU, China also has some major plans in this arena. China’s state-owned space company, the China Aerospace Science and Technology Corporation, plans to develop space-based data centres as part of a broader five-year strategy. This initiative will focus on building an integrated space-system architecture that combines cloud, edge and terminal technologies, enabling computing power, data storage and transmission directly from orbit.

However, there are different opinions in the scientific community about the feasibility of establishing such data centres in space. Orbital data centres will only be economically viable if launch costs fall significantly. But some associated activities could offset energy savings. For instance, chips in space-based data centres may need replacement every five to six years, requiring on-orbit servicing missions. These operations, along with the added risk of space debris, would introduce additional technical and environmental challenges.

Most global projects focused on developing space-based data centres are expected to reach completion within the next three to four years, with some projections indicating that orbital data centres could become economically viable by the mid-2030s. It is expected that by the mid-2030s, various options would be available globally for undertaking launches into low earth orbit, thus significantly reducing launch costs.

Interestingly, not much debate is found happening with respect to the legal issues associated with putting data centres in space. There could be issues related to space and data sovereignty. At present, the legal framework for AI data centres in space is largely undefined. Various agencies involved in developing such structures should remain ready to navigate a mixture of international space laws, emerging technology and data protection regulations.

India is keen to establish itself as a key player on the global AI stage and is poised to become one of the fastest-growing leaders in AI. But despite hosting almost 20% of the world’s data, India has only 3% of global data centre capacity. India needs to develop major compute infrastructure if it has to remain relevant in the game of data and AI.

ISRO recognises that computing and storage are becoming part of space strategy. It is known to be perusing the feasibility of placing ‘physical data centres’ in orbit so some satellite and communications data can be processed and stored in space. Some India space startups are also known to have interest in this field. However, it appears that India’s key focus is more towards establishing land-based data centres.

Finance minister Nirmala Sitharaman during her 2026-27 budget speech announced that India will be providing a tax holiday till 2047 to any foreign companies that provide cloud services to their customers globally by using data centre services from India. Such a provision offers certainty to foreign companies that are in the business of providing cloud services and procurement services from data centres in India. This also is expected to boost investments in data centres and could create new jobs.

On the other hand, a long tax holiday for data centres could come with notable trade-offs, including the loss of substantial long-term tax revenue. More importantly, such data centres would increase the demand for electricity and freshwater, giving rise to environmental concerns. There could also be added pressure on local infrastructure and a rise in litigation stemming from land acquisition and land-use disputes.

India has pledged to achieve net-zero carbon emissions by 2070 at the COP26 Summit in 2021, and it must ensure that the expansion of large AI data centres does not undermine this commitment. In this context, a balanced approach – combining AI data centres on land with some operating from space – could offer a viable solution. To achieve this, ISRO, along with India’s space industry and startups, should develop a coordinated program to build and operate data centres in space.

Ajey Lele is a researcher and is the author of the book Institutions That Shaped Modern India: ISRO.

This piece was first published on The India Cable – a premium newsletter from The Wire – and has been updated and republished here. To subscribe to The India Cable, click here.

This article went live on February tenth, two thousand twenty six, at twenty minutes past eleven in the morning.

The Wire is now on WhatsApp. Follow our channel for sharp analysis and opinions on the latest developments.