On September 14, shares of AI-related companies fell following calls to slow down the development of the technology. According to Yahoo Finance, SoftBank shares, one of OpenAI’s investors, dropped by as much as 13.2% at the start of trading on the Asian market.
The decline also affected semiconductor manufacturers: Kioxia shares fell by 9.8%, SK Hynix by 5.3%, Samsung Electronics and Tokyo Electron by 3.7%, and TSMC by 1.2%.
GLM model developer Z.ai lost 10.5%. In addition to the general sell-off in the sector, the decline was preceded by a discounted new share placement.
On September 13, the company announced raising $5 billion: about $2 billion from share sales and another $3 billion from convertible bonds. The shares were placed at 714 Hong Kong dollars, 10% below the closing price on September 11.
Approximately 60% of Z.ai’s net proceeds will be directed towards developing the next generation of models and their fully automated training system. Another 15% is allocated for business expansion.
Calls to slow AI development have led investors to question whether investments in its supporting infrastructure will pay off.
Swissquote analyst Ipek Ozkardeskaya explained that if the demand for computing grows more slowly, data center revenues may not meet expectations. However, operators will still need to repay debts, pay rent, and fulfill electricity purchase obligations.
She noted that companies that built data centers with borrowed funds expecting rapid demand growth, as well as lenders to such projects, are particularly vulnerable.
What AI Company Leaders Proposed
The reassessment of expectations was prompted by an essay by Anthropic CEO Dario Amodei. His call was supported by Sam Altman of OpenAI and Elon Musk of xAI.
I agree with Dario that we need to pace the frontier. This has been a primary topic of discussions we’ve had at OpenAI in recent weeks.
Committing to having independent evaluators with employee-like access is a great idea, and we will do the same. We’ll have more to share soon. https://t.co/1YhhIybZX7
— Sam Altman (@sama) September 12, 2026
According to Amodei, the capabilities of models are growing faster than the means to ensure their safety. He cited the increasing involvement of AI in creating the next generation of systems as one reason.
Another concern was an incident involving OpenAI and Hugging Face. As Amodei described, AI agents attacked external targets and attempted to hack the system evaluating their performance.
He warned that more powerful neural networks with similar behavior could cause significantly greater harm.
To mitigate risks, the Anthropic CEO proposed three areas of work:
- Independent checks within laboratories. External specialists should have continuous access to development processes, verify compliance with safety measures, and report incidents.
- Common rules for companies. Developers should agree on safety standards and limits on model development pace with government support.
- International agreements. Countries need to discuss joint measures considering the complexity of verifying their implementation.
Anthropic has already committed to implementing the first point. The others require agreements with other companies and governments.
Amodei emphasized that the proposal does not mean halting model training. Its goal is to give developers and independent evaluators more time to assess risks.
He considers a full international pause unlikely in the near future.
U.S. President Donald Trump did not support calls to slow AI development. Speaking to reporters in Ireland, he stated that maintaining American leadership over China remains a priority.
Trump allowed for the introduction of protective measures but called critics’ concerns exaggerated. He said they discuss scenarios that will not materialize.
Earlier, on September 13, Altman ruled out an OpenAI IPO in 2026, citing AI safety concerns. He said the company also discussed potential pauses in model development upon reaching new capability levels.
