OpenAI’s next model is being designed by AI, SoftBank’s Son says

SoftBank’s Son Says AI Is Designing OpenAI’s Next Model

SoftBank CEO Masayoshi Son told CNBC that OpenAI’s upcoming model is being designed by another AI model rather than human engineers, a development he characterized as an early indicator that artificial superintelligence is arriving faster than expected.cnbc

“No Longer Human Designing”

In an interview aired this week, Son said he learned directly from OpenAI CEO Sam Altman and the company’s top engineers that “already their next model no longer human designing. Their model is designing next model”. He described it as a turning point: “Once that happens, model generate next model and next model and it’s going to be exponentially smarter than all of us. That’s a super intelligence”.youtube

Son also revised his timeline for what he calls artificial superintelligence — AI that is 10,000 times more intelligent than humans. In 2024, he predicted it would arrive within a decade. He now says that estimate was deliberately conservative. “When I said 10 years back then, I was trying to be conservative. Cuz people get shocked. But in my mind, I thought it was coming in 4 years instead of 10 years. Now I say it’s coming in next 2 years,” Son told CNBC’s Arjun Kharpal.youtube

Son’s Stake in the AI Race

Son’s bullish posture comes as SoftBank has committed more than $64 billion to OpenAI, making it the company’s largest investor. In the first quarter of 2026, SoftBank recorded a $25 billion gain on its OpenAI stake, according to The Wall Street Journal. The Japanese conglomerate’s shares have surged roughly 70% this year on investor enthusiasm over AI, though analysts have raised concerns about the firm’s concentrated exposure and leverage.wsj

Son has also positioned SoftBank across the AI supply chain, from its chip-design subsidiary Arm to massive data center buildouts, including a $500 billion AI campus in Ohio. He told CNBC the AI revolution would be “50 times bigger than dotcom”.facebook

A “Zebra” Moment for Intelligence

Son said he uses ChatGPT two to three hours daily and believes current models already surpass human expertise in more than half of all subjects — what he called a “zebra AGI” moment, where capabilities exceed humans in some domains but lag in others. He predicted that within two years, AI would be superior in 70 to 80 percent of subjects discussed, and in those areas, “10 times smarter than average people”.youtube

CNBC noted that the practice of using AI to develop AI is not unique to OpenAI, reporting that Anthropic has similarly used its own coding tools in the development of Claude.youtube