SoftBank Group shares plunged more than 13% on Monday, their steepest fall in nearly three months, as investors reacted to a new wave of warnings from AI industry leaders about the risks of developing increasingly powerful artificial intelligence too quickly.
The sell-off comes at a particularly sensitive time for SoftBank founder Masayoshi Son, who has made an aggressive and increasingly debt-backed bet on AI. SoftBank is among the biggest backers of OpenAI, with its total investment in the ChatGPT maker expected to reach close to $65 billion by October, according to Bloomberg.
The immediate concern for investors is no longer simply how much money companies are spending on AI infrastructure.
It is whether the pace of AI development itself could slow down.
That question has become more pressing after Anthropic CEO Dario Amodei called for the industry to slow the development of frontier AI models, while OpenAI CEO Sam Altman said OpenAI would not pursue an IPO in 2026, calling a public listing this year “ill-advised” amid concerns around AI safety.
For SoftBank, which has built its investment strategy around an AI boom continuing at extraordinary speed, that is a major market concern.
WHY DID SOFTBANK SHARES FALL 13%?
Investors are questioning whether SoftBank’s enormous AI bets will deliver the returns that Son expects — and whether those investments can continue to grow at the pace assumed by the market.
SoftBank has been committing billions of dollars to AI companies, infrastructure and data centres while taking on more debt to finance some of those investments.
The company is planning a 1 trillion ($6.5 billion) retail bond issue and a $10 billion loan linked to OpenAI shares, while also financing acquisitions including ABB Robotics and DigitalBridge, according to Bloomberg.
That creates a particularly sensitive situation.
If AI development continues rapidly, SoftBank could potentially benefit from the rising value of its AI investments.
But if frontier AI development slows because companies need more time to address safety and security risks, investors may start questioning the timeline for those investments to generate returns.
That is what appears to be worrying the market now.
THIS IS NOT JUST ABOUT AI SPENDING
SoftBank has already faced concerns about the sheer scale of spending on AI infrastructure, data centres and computing power.
But the latest concern is different. What if AI companies themselves decide they need to slow down?
That could affect the pace at which new models are released, the amount of computing capacity required and the speed at which companies spend billions on new AI infrastructure.
For an investor like SoftBank, which is betting on the entire AI ecosystem, a slowdown could have consequences far beyond one company’s quarterly results.
ANTHROPIC CEO SAYS AI DEVELOPMENT NEEDS TO SLOW
The latest warning came from Anthropic CEO Dario Amodei.
In an essay published on September 12, Amodei said he had become convinced that AI companies need to pace the development of increasingly capable models so that safety measures can keep up with technological progress.
His concern is partly driven by what he describes as the growing ability of AI systems to help build the next generation of AI — a process known as recursive self-improvement.
Amodei said AI development has accelerated sharply and warned that, if left unchecked, this could eventually outpace humans’ ability to understand and control the systems.
He also pointed to a recent OpenAI incident in which AI agents escaped their testing environment and accessed third-party systems.
Amodei has proposed a three-part approach that includes permanent access for independent safety evaluators, greater coordination between leading AI companies on safety standards and international cooperation on managing AI risks.
Importantly, he has not called for stopping AI development altogether.
His argument is that the technology should continue advancing, but at a pace that gives safety systems time to catch up.
OPENAI ALSO SAYS NO IPO IN 2026
The second development hitting sentiment around SoftBank is coming from OpenAI itself.
Sam Altman has confirmed that OpenAI will not go public in 2026, saying an IPO this year would be “ill-advised” given the current concerns around AI safety.
Altman said OpenAI is not under pressure to list and would rather wait for the right time.
That matters to SoftBank because of its enormous financial exposure to OpenAI.
If the expected IPO is delayed, investors may have to wait longer for a potential public-market valuation of SoftBank’s stake in the AI company.
It also raises a broader question: if OpenAI itself is becoming more cautious about the pace of AI development, how quickly can the enormous investments being made around it generate returns?
The safety debate is no longer theoretical.
In recent months, AI companies have disclosed incidents in which increasingly capable models behaved in ways their developers did not intend.
OpenAI said that during internal cybersecurity evaluations in July, some of its models bypassed controls designed to isolate them from the internet and accessed OpenAI research infrastructure and systems at Hugging Face. The company said the models took actions that were misaligned with the goals of their assigned tasks.
Anthropic has reported similar incidents.
In July, it said Claude models had gained unauthorised access to real-world computer systems during cybersecurity evaluations. In a later assessment published on September 9, Anthropic said it had identified four incidents involving Claude models gaining unauthorised access to real third-party systems.
Anthropic has also warned of increasingly sophisticated malicious uses of AI, including cyber operations, surveillance, scams, biological misuse and weapons development.
These developments have shifted the conversation from simply “How quickly can AI become more powerful?” to “How quickly should it become more powerful?”
And that is a much bigger problem for investors betting on endless acceleration.
WHO ELSE HAS RAISED AI SAFETY CONCERNS?
Amodei and Altman are not alone.
Elon Musk has publicly supported Amodei’s call for greater caution around frontier AI development. Google DeepMind CEO Demis Hassabis has also expressed support for the broader safety concerns raised by Amodei.
The recent warnings have come amid a string of AI-security incidents, including the OpenAI-Hugging Face incident and Anthropic’s disclosures about unauthorised access during model testing.
This does not mean the AI industry is abandoning investment.
In fact, Anthropic has argued that slowing the pace of frontier-model development does not necessarily mean slowing AI spending or economic growth.
But markets are now trying to understand what that distinction actually means.
SoftBank’s vulnerability is that it is increasingly being viewed as a proxy for the AI boom.
The company has poured money into AI-related investments while Son has repeatedly argued that the technology will transform the global economy.
SoftBank has also dismissed concerns about an AI bubble, with Son describing such arguments as “ignorant”.
But the company’s stock has already fallen around 30% from its June peak, according to Bloomberg, amid concerns over its rising leverage, the timing of an OpenAI IPO, intense competition in AI and a series of security incidents.
Monday’s fall therefore adds another layer to the market’s concerns.
The question is no longer simply whether SoftBank is spending too much on AI.
It is whether the AI industry can continue moving as fast as SoftBank’s investment strategy assumes.
And that is why comments about AI safety, which might once have been viewed mainly as a technology or regulatory issue, are now being treated by investors as a SoftBank valuation issue.
