SoftBank's Son joins Altman in doubting Musk's orbital data centres

The News

Masayoshi Son, the chief executive of SoftBank, has become the most prominent investor to publicly question Elon Musk's pitch for data centres in orbit. Speaking at a shareholder meeting on 27 June 2026, Son argued that putting compute in space will not meaningfully lower costs and will take too long to matter, insisting that "the next few years will be far more important" than anything that arrives a decade from now.

He is not alone. OpenAI's Sam Altman has voiced similar doubts about the timeline, as TechCrunch's Anthony Ha reported. The scepticism is notable because both men are spending enormous sums on the opposite approach. SoftBank has committed up to 75 billion euros to a data-centre programme in France, while the wider industry continues to pour money into terrestrial compute, including Groq's recent 650 million dollar funding round.

The case for orbital data centres rests on a genuine compute shortage and the idea that space sidesteps the planning fights and community objections that slow earthbound builds. Critics counter that the economics do not add up. Satellites in such a constellation would need replacing every few years, a cycle that mostly guarantees repeat business for SpaceX, which already controls an estimated 80 to 90 per cent of the global launch market.

Why It Matters

The disagreement is really a fight over where the next trillion dollars of AI infrastructure should go, and when. Son's argument is that the AI war is being decided now, on the ground, not in some orbital future. When the people building the biggest data centres on Earth start dismissing the flashiest alternative, it is worth asking whether the alternative is a serious plan or a way of keeping launch demand high.

There is precedent for infrastructure visions that outrun their economics. The fibre-optic glut of 1999 to 2001 saw operators lay vast capacity on the promise of bottomless demand, only for much of it to sit dark for years once the bubble deflated. Every executive in this debate is, as the reporting puts it, talking their own book. Son sells terrestrial capacity, Musk sells launches, and each forecast conveniently favours the speaker's balance sheet.

Indian Angle

For India, the terrestrial-versus-orbital quarrel is not abstract. The country is one of the fastest-growing data-centre markets in the world, with Reliance, AdaniConneX, CtrlS and Yotta racing to add hyperscale capacity, much of it earmarked for AI workloads under the government's IndiaAI Mission and its push to procure tens of thousands of GPUs. Son's verdict, that the contest is won in the next few years on the ground, is effectively a vote of confidence in exactly the kind of build-out Indian operators are betting on.

It also sharpens a hard local problem. Indian data centres are constrained by power availability, land prices in hubs like Mumbai and Chennai, and water for cooling. Orbital compute, if it ever worked, would bypass all three, which is why Indian planners cannot dismiss the idea entirely even as the cost maths looks forbidding today.

Finally there is sovereignty. MeitY's data-localisation expectations and the new Digital Personal Data Protection framework assume servers sit in a jurisdiction that regulators can reach. A constellation orbiting above national borders would scramble that assumption, raising questions Indian policymakers have not yet had to answer.

FAQ

What did SoftBank's CEO actually say?

Masayoshi Son told shareholders on 27 June 2026 that orbital data centres will not significantly cut costs and will take too long to build, and that the AI race will be decided in the next few years rather than a decade out. He is backing terrestrial capacity instead, including a programme worth up to 75 billion euros in France.

Why does this matter for Indian companies?

India's data-centre operators, from Reliance to Yotta, are committing billions to ground-based AI capacity. A leading global investor publicly favouring that approach validates their strategy, while the orbital debate highlights the power, land and water constraints Indian builders must still solve.

Is anyone actually building data centres in space yet?

Not at scale. The concept remains largely a pitch tied to SpaceX's launch business, which holds an estimated 80 to 90 per cent of the global launch market. Critics note the satellites would need frequent replacement, benefiting launch providers more than compute economics.

Where can I read the original report?

The full reporting, by Anthony Ha, is available on TechCrunch via the link in the attribution below.

This story was reported by TechCrunch. Read the full original coverage at TechCrunch.