SoftBank is preparing to move into one of the hottest corners of the tech economy: renting out computing power for artificial intelligence.
According to The Japan Times, the Japanese conglomerate plans to offer AI-tailored cloud services in the United States starting next fiscal year. The idea is to provide the specialized infrastructure that companies need to build and run AI systems, delivered over the cloud rather than requiring customers to own the hardware themselves.
The Register frames the move as SoftBank entering the "rent-a-GPU race" — a reference to the graphics processing units that have become the essential engines for training AI models. Demand for that kind of horsepower has surged, and The Register notes that America is looking for additional support for AI training capacity.
By positioning itself as a supplier of that capacity in the US market, SoftBank would be competing to serve the AI developers and businesses that need heavy computing resources but want to avoid the cost and complexity of building their own data centers.
Why it matters: AI progress is increasingly bottlenecked by access to computing power, and SoftBank's planned US entry signals how central the business of renting out that power has become to the technology's future.