Japanese conglomerate SoftBank would invest $3 billion to renovate an electric vehicle plant in Lordstown, Ohio. Following renovations, the plant is expected to include equipment for OpenAI data centres, including those in Milam County, Texas, among other locations. The Japanese company is going all in on the startup at the heart of the AI boom.
SoftBank Sold Nvidia Stake to Fund AI Push
SoftBank sold its $5.8 billion stake in leading chipmaker Nvidia to pay for CEO Masayoshi Son’s big AI push centred around ChatGPT creator OpenAI. SoftBank bought the Lordstown site in August for $375 million. In September, OpenAI, Oracle and SoftBank said they plan to build five US AI data centres tied to the $500 billion Stargate project. The goal is to build out a nationwide advanced AI network.
This investment is part of a larger joint venture announced at the White House back in January. SoftBank had promised $18 billion at the time, according to The Information. Ohio will host a factory building modular data centres, portable, pre-assembled units that enable faster setup and scalable capacity on site. The Information reported this. The facility will host a small working data centre as a demonstration model, it said.
Factory Will Start Making Units Early Next Year
Manufacturing of modular data centre units is expected to start early next year, the report said. SoftBank and OpenAI did not immediately respond to Reuters requests for comment. In an October livestream, Sam Altman, chief executive of OpenAI, said his company wants to build 30 gigawatts of computing capacity estimated at $1.4 trillion, eventually scaling to add about 1 gigawatt a week. Currently, costs run above $40 billion per gigawatt.
Unlike its rivals, Meta and Google, OpenAI does not have an advertising or cloud services business to offset the expense of building out these data centres. In fact, it needs lots and lots of cash to sustain its growth. The company burns through cash building and running the AI systems in service of its core products. That is because ChatGPT is an expensive way to handle queries that require powerful computers.
Big Bets on AI Infrastructure
SoftBank is making a huge bet that AI will be the next big thing. Masayoshi Son has made huge bets before. Sometimes they pay off big. Other times, they do not. He believes AI will transform everything. That’s why he’s putting billions into OpenAI and related projects. Selling the Nvidia stake was a big move. NVIDIA’s stock has gone up a lot. But Son needed cash now to fund his AI vision.
The Lordstown factory has quite an interesting history. It is used to manufacture electric vehicles. Well, that did not quite pan out. Now it will have a second life, making equipment for AI data centres. The modular approach is smart. Instead of building each data centre from scratch on site, companies can make parts in a factory. Then they ship them where needed and assemble them quickly. This saves time and money.
The $500 billion Stargate initiative is massive; it entails the procurement of diverse bulk equipment for five data centres located in the US. The factory in Ohio will be a kind of local supplier for the necessities, but $3 billion is only the initial part of the story. The total amount of money spent on the data centres will increase a lot. Actually, this hardware is very vital to OpenAI, which has to keep upgrading ChatGPT and developing new AI products. If there were a lack of computing power, the company would not be able to train better models and make them available to all its users.
Sam Altman’s ambition of 30 gigawatts sounds hardly reachable. That is an extraordinary quantity of computing power. The price tag of $1.4 trillion is staggering. Several people question whether OpenAI could actually raise that much money. But Altman pulled off amazing things before. He convinced investors to put billions into OpenAI in the first place. The company went from unknown to famous in only a couple of years.
Catch more technology and investment news at Inspirepreneure magazine, as major companies build the future in artificial intelligence infrastructure.
