SoftBank founder and chief executive Masayoshi Son set out one of the technology industry’s most aggressive timelines for artificial superintelligence, arguing that systems vastly beyond human ability could arrive within ten years. Speaking at the company’s annual meeting in Tokyo, Son said AI might become one to ten times smarter than humans by 2030 and as much as 10,000 times smarter by 2035. The forecast was not a technical consensus, but a statement of strategic intent from an investor with the capital and influence to shape which projects receive support.
Son Draws a Sharp Line Between AGI and ASI
Son distinguished artificial general intelligence from artificial superintelligence by degree. In his framing, AGI could match a human genius and reach roughly ten times the capability of an average person. ASI would belong to a different category, exceeding human potential by a factor of 10,000. Such numerical comparisons are rhetorically powerful but difficult to define. Intelligence is not a single measurable quantity, and performance in reasoning, memory, creativity or physical interaction may advance at different rates.
The distinction nevertheless matters strategically. AGI is commonly discussed as a system able to perform a broad range of intellectual tasks at human level, while ASI implies a decisive and potentially uncontrollable advantage over people. Moving the conversation directly to ASI raises the stakes for governance, safety and economic planning. If the goal is not merely a flexible assistant but an intelligence far beyond its creators, ordinary product testing and incremental regulation may be insufficient.
A Personal Vision Becomes SoftBank’s Corporate Direction
Son described the pursuit of ASI in deeply personal terms, saying he believed realising it was the purpose for which he and SoftBank existed. That language suggested a potential shift from treating AI as one investment theme among many to making superintelligence a central organising ambition. SoftBank’s history includes large, concentrated technology bets, so the declaration had implications beyond speculation. Capital allocation, semiconductor assets and partnerships could all be directed toward infrastructure and companies aligned with this objective.
Visionary conviction can mobilise resources, but it can also compress uncertainty. Forecasts about AGI and ASI vary widely, and present systems still struggle with reliability, grounded reasoning and tasks outside their training patterns. A ten-year deadline may encourage investment and experimentation, yet it can create incentives to interpret every model improvement as confirmation. Sound strategy requires separating a founder’s confidence from evidence about which technical obstacles remain and whether scaling current methods will solve them.
Safe Superintelligence Moves Into the Industry Mainstream
Son’s comments arrived as former OpenAI chief scientist Ilya Sutskever launched Safe Superintelligence Inc., a company dedicated to developing superintelligent systems while keeping safety ahead of capability. The proximity of the announcements showed that ASI was becoming an explicit institutional goal rather than a distant philosophical topic. One approach emphasised speed and transformative potential; the other made safety and capability a combined engineering problem. Both assumed that superintelligence was important enough to organise a company around.
The safety challenge grows with the capability being claimed. Systems that outperform humans across most domains could affect labour markets, security, scientific research and political power. They might generate enormous benefits, but failures or misuse could also scale beyond existing oversight. It is therefore not enough to promise that safety will be considered later. Evaluation methods, access controls, alignment research and decision-making structures need to develop before capabilities make deployment pressure overwhelming.
Ambition Must Be Matched by Evidence and Governance
The scientific community had not reached agreement that AGI or ASI was achievable on Son’s schedule—or even on how either threshold should be tested. Current AI was impressive within specific domains but remained far from dependable human-level reasoning across every context. That uncertainty should not lead policymakers to ignore the prospect, because preparations take time. It should instead encourage scenario planning that covers rapid progress without treating a single forecast as fact.
Son’s prediction was most significant as a signal of where influential capital might flow. When major investors believe superintelligence is close, they fund chips, data centres, model companies and research teams accordingly. Those investments can make the forecast partly self-reinforcing by accelerating development. The necessary counterweight is governance that scales with ambition: transparent milestones, independent testing and serious attention to who controls the resulting systems. Whether ASI arrives in ten years or much later, the race described by Son will shape technology markets well before the destination is reached.
Boards and investors should therefore evaluate the thesis through more than a dramatic end-state. Intermediate milestones could include improvements in planning, reliability, scientific reasoning and the ability to operate safely over long tasks. Each gain brings commercial opportunities and risks even if ASI never arrives. Clear milestones would allow SoftBank to distinguish progress from hype and decide when infrastructure spending is justified. They would also give policymakers a more concrete basis for oversight. Preparing for extreme capability is prudent, but responsible preparation means tracking observable evidence rather than allowing an undefined future intelligence to explain every present investment. It also requires contingency plans for slower progress, so that expensive infrastructure and corporate strategy continue to create value if the forecast proves premature. That discipline would make the vision more credible to stakeholders who do not share Son’s confidence and timetable.
The forecast should remain a hypothesis to test, not a deadline that replaces critical judgement.
