Code-named Izanagi, the project marks the billionaire’s next big endeavor as SoftBank sharply curtails startup investments.
Son envisions creating a company that can complement chip design unit Arm Holdings Plc and allow the billionaire to build an AI chip powerhouse, said the people, who requested anonymity because the discussions remain private.
In one scenario being considered, SoftBank would provide $30 billion, with $70 billion possibly coming from institutions in the Middle East, one of the people said.
If he succeeds, the chip project will mark one of the largest investments in the AI arena since the advent of ChatGPT, dwarfing Microsoft Corp.’s recent bet on OpenAI of more than $10 billion.
Son named the project after Izanagi, the Japanese god of creation and life, partly because it includes the initials for artificial general intelligence, the people said. Son has for years foretold the coming of AGI in his presentations, saying that a world filled with machines that are smarter than humans will be safer, healthier and happier.
Discover the stories of your interest
Details of how the project will be funded or where the money will be spent have yet to be sorted out, and the project may evolve further, the people said. Son is continually toying with multiple investment ideas and strategies to strengthen Arm’s reach into the AI market and is always exploring different types of next-generation chips, they said. It’s unclear which company or companies will play a central role in developing the technology needed to challenge Nvidia, by far the leader in high-end AI accelerators.
Representatives of SoftBank and Arm declined to comment.
After a series of brutal setbacks in his startup investments, Son is focusing his efforts on Arm. The entrepreneur sees an opportunity to create a monster of a company in the same league as the Magnificent Seven stocks, the people said.
SoftBank had ¥6.2 trillion ($41 billion) in cash and cash equivalents as of December 31, thanks to a rebound in global equity markets. Its balance sheet got a boost from a windfall in T-Mobile US Inc. shares, worth almost $8 billion, as well as from the company’s 90% stake in Arm, which added roughly $50 billion in market value over the past week alone.
While Son and OpenAI’s Sam Altman have held talks about joining forces and raising money in semiconductor manufacturing, Izanagi in its current form is separate from Altman’s ambitions, the people said.
Son at one point sought to invest in another company developing a foundational AI model and asked the leaders of that business to help with the chip venture, but they declined, one person said.
Alongside Son’s pursuit of AI-related investments, SoftBank has been exploring ways to use Arm’s chip designs. As a member of the SoftBank board and a technology expert, Arm chief executive officer Rene Haas is advising Son on the project, the people said.
In a recent interview on Bloomberg TV, Haas was asked about this role in helping Son achieve his AI aspirations. “When you think about artificial general intelligence and what is required to make that happen in terms of compute, power efficiency, energy — those are all great areas for us to be involved in and focused on,” Haas said, noting that he was speaking as Arm CEO, rather than as a SoftBank director.
Son is spearheading Project Izanagi directly. The billionaire previously orchestrated the creation of SoftBank’s Vision Fund, which was at first a $100 billion project with Middle East backing that represented one of the world’s largest pools for tech investment. But in the past 18 months, investment from the unit has slowed to a fraction of its former scale.
Son is known for abruptly changing his mind, and many names and technologies have been thrown around in meetings with him, people who have worked with the entrepreneur have said.
Son, however, has been unwavering in his enthusiasm for AGI. In October, he told a large group of Japanese enterprise clients to adopt AI or get left behind.
“AGI is what every AI expert is after,” Son said. “But when you ask them about a detailed definition, a number, the timing, how much computing power, how much smarter AGI is than the human intelligence, most of them don’t have an answer.”
“I have my own answer: I’m convinced AGI will be real in 10 years,” he said.
