Software

AI + ML

Softbank CEO says 'super AI' will arrive in 2035 and cost $9T

Oh, and it'll need the total current US power output


Video Masayoshi Son, the Japanese billionaire behind Softbank, has predicted that an artificial superintelligence 10,000 times smarter than a human brain will be created within about a decade - but at quite a cost.

Speaking at the Future Investment Initiative conference in Saudi Arabia, Son said that a scaled-up version of AI could add around 5 percent of global economic output by 2035, but it'll need a massive investment in chips and power.

For this reason Nvidia - currently the world's most valuable company, is undervalued - Son argued.

"I have predicted it would take 400GW of AI datacenter power, that's bigger than the total US electricity supply and it will require 200 million chips," he said, as you can see below. "The cumulative capex is $9 trillion - how do you recoup? It's too much investment for many people's view but I say it's still very reasonable capex. $9 trillion is not too big, maybe too small."

He opined that if the so-called super AI could produce around $9 trillion a year in benefits that initial investment would be "small money," and could be depreciated. He predicted four companies would make trillion-dollar profits, adding, as one might expected, that he wanted to be one of them.

While Son is all-in on AI, he's adjusting his timescale. In June he was predicting artificial superintelligence would be here in three to five years, saying that Arm - which Softbank owns a majority share of - would be one of the beneficiaries. But so will Nvidia, he predicted this week.

"I think Nvidia is undervalued because the future is much bigger," he suggested.

There are a few small problems along the way. Building a super AI would currently take colossal amounts of power and money - more than most governments could manage. And what if he's wrong? It has happened with Softbank before.

It's worth pointing out the investment house poured billions into office-sharing biz WeWork, for example, only to see the value of the investment plummet after a bonkers pre-IPO statement by co-founder Adam Neumann caused first the offering to be called off, and then led to his resignation. Full disclosure: The Register in San Francisco still has a WeWork office, and if there are more than a dozen people on the entire floor of the Embarcadero building, we'd be very surprised.

So is Softbank about to take another bath? Son thinks not, and the multi-billionaire says he's saving up to cover the costs of building the new super AI of the future. And when one of the richest people in the world with a net worth of around $16 billion tells you they're saving, it's a good indicator that the costs of such a scheme are going to be massive. ®

Send us news
21 Comments

AI firms and civil society groups plead for passage of federal AI law ASAP

Congress urged to act before year's end to support US competitiveness

Gary Marcus proposes generative AI boycott to push for regulation, tame Silicon Valley

'I am deeply concerned about how creative work is essentially being stolen at scale'

Sorry, but the ROI on enterprise AI is abysmal

Appen points to, among other problems, a lack of high-quality training data labeled by humans

OpenAI loses another senior figure, disperses safety research team he led

Artificial General Intelligence readiness advisor Miles Brundage bails, because nobody is ready

Polish radio station ditches DJs, journalists for AI-generated college kids

Station claims it's visionary, ex-employees claim it's cynical; reality appears way more fiscal

UK’s new Minister for Science and Technology comes to US touting Britain's AI benefits

$82B in investment shows we've still got it as a nation

Linus Torvalds: 90% of AI marketing is hype

Linux kernel creator says let's see which workloads use GenAI in five years

Voice-enabled AI agents can automate everything, even your phone scams

All for the low, low price of a mere dollar

UK gov report to propose special zones for datacenters, 'AI visas'

Vendors not keen on 'lengthy bureaucracy,' and cost when they try to hire skilled foreigners

Meta spruiks benefits of open sourcing Llama models – to its own bottom line

It's not like Zuck needs the coin despite increased infrastructure spend, headcount, losses on VR

No-Nvidias networking club convenes in search of open GPU interconnect

Ultra Accelerator Link consortium promises 200 gigabits per second per lane spec will debut in Q1 2025

The troublesome economics of CPU-only AI

At the end of the day, it all boils down to tokens per dollar