SoftBank founder Masayoshi Son has dismissed warnings of an artificial intelligence investment bubble, comparing sceptics with people who resisted cars and aircraft and arguing that opposition to the technology amounted to rejecting human progress.

Speaking at the SoftBank World conference in Tokyo on Tuesday, 14 July, Son said AI would become more capable than humans across a widening range of intellectual and physical tasks, reshaping the global economy over the next 15 years.

“Those who dislike AI have essentially refused their own evolution,” Son said. “Those who condemn AI are themselves spitting upwards.”

Son predicted that AI-related industries could account for about 20% of global gross domestic product (GDP) by 2040.

Based on his estimate of the size of the world economy at that point, that would represent roughly $46 trillion in annual output and could generate profits of about $23 trillion.

“In 2040, approximately 20% of the world’s GDP will be replaced by AI-related industries, the world of superintelligence,” he said, according to the Associated Press.

Supporting an AI economy on that scale would require almost $5 trillion in global infrastructure investment each year by 2040, Son said. The spending would cover data centers, semiconductor factories, electricity generation, communications networks, and robotics.

“Five trillion dollars a year, you might think that is unbelievable, but I firmly believe this is the cost required,” he said.

Son argued that such spending would be commercially sustainable if AI eventually generated one-fifth of global output.

“The business model will be viable because by 2040, if AI revenue makes up 20% of global GDP, spending 800 trillion yen a year is a rounding error,” Reuters quoted him as saying.

100 trillion AI agents

The scale of the forecast goes far beyond today’s AI market. Son expects about 100 trillion autonomous AI agents to be operating by 2040, according to NHK World-Japan. Such agents are designed to carry out multistep tasks, use software and make limited decisions with less direct human supervision than conventional chatbots.

He also predicted that humanoid robots would move to the center of physical work. “For the first time, humanoid robots will take the main role in physical labor, replacing humans,” NHK quoted him as saying.

That would extend Son’s vision from software and office work into factories, warehouses, transport, construction, and other labor-intensive industries. Some reports said he envisaged as many as 1 billion humanoid robots alongside the vastly larger population of digital agents.

Powering the system would present an equally formidable challenge. Son estimated that AI data centers could eventually require about 3 terawatts of electricity, around 1.8 times current global electricity consumption, according to Reuters.

Gas-fired generation could provide part of the near-term supply, he said, while nuclear fusion might eventually become a longer-term source of power for AI infrastructure.

Fusion, however, has yet to become commercially viable, while large gas and data center projects face high costs, lengthy construction periods, grid constraints and political opposition.

Son also warned that businesses unwilling to adopt AI risked becoming irrelevant. His comments were aimed particularly at Japanese executives, whom he has repeatedly criticized for moving too slowly in embracing technological change.

He said SoftBank could not confine its ambitions to Japan and would need to expand overseas, particularly in the US, where much of the world’s AI model development, computing infrastructure and investment is concentrated.

Realigning AI investments

Son’s enthusiasm is hardly disinterested. SoftBank has reorganized much of its strategy around AI through its controlling stake in chip designer Arm, investments in OpenAI, data center projects, semiconductor ventures, robotics and energy infrastructure.

SoftBank announced in April 2025 that it had agreed to invest up to $40 billion in OpenAI, subject to conditions. Reuters reported that its cumulative investment in the ChatGPT developer could reach about $60 billion by the end of 2026.

The group has also sold assets, including its Nvidia holding, to fund its AI expansion. SoftBank reported a profit of about ¥5 trillion, or $32 billion, for the financial year through March, helped by gains in the value of its technology and AI investments.

The strategy carries substantial risks. Investors have become more concerned about SoftBank’s leverage, the financing demands of its projects and possible delays to an initial public offering by OpenAI, which could affect the timing and value of any eventual return.

More broadly, critics question whether AI companies can generate enough durable revenue to justify unprecedented spending on chips, data centers and power. The industry also faces electricity bottlenecks, semiconductor constraints, rapid equipment depreciation and uncertainty over how quickly experimental systems can be converted into profitable products.

Son’s answer is that those concerns focus on the wrong time horizon. He sees current expenditure not as the peak of a speculative cycle, but as the first stage in constructing an economy in which autonomous software and intelligent machines carry out a large share of human work.

That vision is consistent with a career built around enormous, concentrated bets on technological shifts. Son invested early in Alibaba and Arm, but also backed companies such as WeWork that produced severe losses. His latest wager, however, is larger than either precedent.