Masayoshi Son does not agree with Elon Musk's space data centers: launching rockets and maintenance costs are not cost-effective, and there are communication delays.

SoftBank founder Masayoshi Son publicly rejected Elon Musk’s proposal to build space-based data centers at a shareholders’ meeting, saying that electricity costs account for only a small part of a data center’s overall costs. He added that launch and maintenance expenses, along with unavoidable communication latency, make the overall benefits unattractive.
(Background: Son warns of AI “black ships coming”: SoftBank teams up with OpenAI to roll out a new enterprise cybersecurity protection service)
(Additional context: SoftBank announces an AI battery venture—converting Osaka Sharp’s LCD plant into a zinc halide plant, partnering with South Korea’s Cosmos Lab to challenge China’s lithium-cobalt supply chain)

This Tuesday (the 23rd), at a shareholders’ meeting for the telecommunications business under SoftBank Japan, Masayoshi Son directly pushed back against Musk’s claim that data centers should be built in space. His logic starts with the cost structure. The advantage of space-based data centers most often cited is “saving on electricity”—no need for terrestrial power grids, and in theory they could be powered continuously using solar energy.

But Son pointed out that electricity costs only account for a small portion of a data center’s total operating costs; the real bulk is hardware spending such as AI chips. Even if you compare the electricity savings with the transportation costs of launching into space, the costs of maintaining systems while in orbit, and the unavoidable communication latency caused by signals traveling to and from Earth, there is not enough incentive.

Son described Musk as an “outstanding change-maker,” but he believes this path is not worth taking now.

Why “now” is the key word

Son’s full argument is not just that “space is not good,” but that “timing determines everything.” He said:

“In the AI race, what happens in the coming few years matters far more than what might happen more than ten years from now.”

Even if space-based data centers are technically feasible, the timeline is more than 10 years. Meanwhile, the competition for AI computing power is being decided right now, right here. The core logic Son repeatedly emphasizes is that “the early mover wins.”

On the other hand, how big is SoftBank’s footprint on Earth? The company previously pledged to invest about $65 billion into OpenAI’s Stargate project; in May 2026, it announced that it would invest as much as 750 billion euros in France. The first phase would build a 3.1GW capacity with 450 billion euros, expected to be completed before 2031. The construction locations include Dunkirk, Bouscayrol, and Bouchain in northern France. In addition, there is a plan for a super AI data center park in Ohio with $500 billion, targeting a capacity of 10GW.

Perhaps these figures are also part of the reason Son dismisses the space route, after all, all of his resources are already committed to Earth, with no room to further scatter them.

Another hurdle for Earth’s computing power

But at the same time that Son was calling for “Earth first,” another counterforce on Earth was taking shape.

Mayors from 40 cities reached a consensus to work together to curb the rapid expansion of data centers and mitigate their impacts on power grids, water supply systems, and local communities, with London and Phoenix among them.

Melbourne Mayor Sally Capp provided the most straightforward data: Melbourne currently has around 50 large data centers in operation, and it is expected that by 2030 they will account for about 10% of local electricity demand, rising to 20% by 2040. Her analogy is even more striking: “Data centers are the biggest source of impact on power grids since air conditioning became widespread in the 1950s. It took decades for air conditioning to become common, but this happened in just a few years.”

This is a problem Son and Musk both face, yet one that has barely been discussed publicly. No matter where the computing power is built, it still requires energy, cooling, and infrastructure—and these resources on Earth are not unlimited. Supporters of the space route might say that this is precisely the reason to leave Earth. But Son’s answer is already very clear: that’s for the next race, not this one.

Win first in the present, then talk about the future.