Anthropic, OpenAI, Musk Make Rare Call to Slow AI Development; SK Hynix, SoftBank Slump: Will Nvidia's Compute Demand Be Hit?

TradingKey - Anthropic CEO Dario Amodei called for slowing the pace of capability advancements in frontier AI models, which OpenAI CEO Sam Altman and SpaceX (SPCX) CEO Elon Musk subsequently supported. The statements heightened market concerns regarding the growth of the AI industry and the outlook for chip demand. During Asian trading hours on September 14, multiple semiconductor and AI-related stocks fell sharply.

However, Amodei advocated slowing the pace of capability advancements in frontier AI models to allow more time for safety testing and risk prevention, while continuing to advance model training and technological R&D.

Currently, the relevant statements have not involved cutting data center construction or capital expenditures. Whether Nvidia (NVDA) GPU demand will be affected remains to be seen, requiring observation of major cloud service providers' capital expenditures, model training scale, and inference business growth.

Anthropic, OpenAI and Musk Support Slowing AI Development

On September 12, US local time, Amodei published "We Must Pace the Frontier," calling for a slowdown in the pace of capability advancements for frontier AI models to buy more time for model safety, risk prevention, and third-party evaluations.

Amodei believes that AI systems are increasingly involved in the development of next-generation models, and the speed of capability advancement may outpace the industry's ability to understand and control associated risks. He also cited an incident where OpenAI agents attacked Hugging Face-related systems during testing, warning that clusters of more capable agents with similar safety flaws could pose a serious threat to the internet through persistent botnets within the next 6 to 12 months.

To mitigate these risks, Amodei proposed establishing long-term third-party evaluation teams within frontier AI companies, granting them tools, permissions, and information access roughly equivalent to those of internal risk assessors to verify safety measures, report incidents, and evaluate model training processes.

The proposal also includes urging frontier AI companies in democratic nations to establish common safety standards and further pursue global coordination. Anthropic has committed to implementing a third-party evaluation mechanism and plans to invite external assessment teams on-site in the near future.

Altman subsequently expressed agreement with the direction of pacing capability advancements in frontier models and promised that OpenAI would adopt similar independent evaluation arrangements. Musk publicly stated that "Dario is right." Demis Hassabis, co-founder and CEO of Google DeepMind, also agreed that this direction is correct, though noting that relevant details still need further refinement.

[Source: X]

While these statements do not constitute a joint industry agreement among the companies, they show that AI safety issues are influencing the R&D schedules and regulatory stances of leading firms.

SK Hynix, SoftBank and Other AI-Related Stocks Fall

During the Asian trading session on September 14, SoftBank shares dropped more than 13% at one point, while Kioxia fell as much as 9.8% in early trading; as of press time, SK Hynix was down 6.4%, Samsung Electronics dropped 3.95%, and TSMC slipped 1.24%.

[Source: TradingView]

Investors are concerned that a slowdown in frontier model capabilities could reduce training projects, extend intervals between model releases, and impact cloud providers' data center investments as well as demand for GPUs, HBM, and semiconductor equipment. Currently, no public information shows that relevant companies have cut capital expenditures or chip orders.

SoftBank is a major investor in OpenAI, and its share price is sensitive to expectations for OpenAI and the AI industry's development. As a major global HBM supplier, SK Hynix is also vulnerable to changes in demand expectations for AI servers.

Middle East tensions, rising oil prices, and Federal Reserve policy expectations also weighed on the market that day, while discussions surrounding a slowdown in AI development further exacerbated investor concerns about the outlook for semiconductor demand.

Will Nvidia’s Computing Power Demand Be Impacted?

As of September 14, major tech companies including Microsoft (MSFT), Amazon (AMZN), Google (GOOGL), and Meta (META), have not announced any reductions in data center capital expenditures due to related calls, nor are there public signs of changes in Nvidia GPU orders or upstream HBM demand.

Nvidia CEO Jensen Huang reiterated at the Goldman Sachs Communacopia + Technology Conference on September 10 that annual global AI infrastructure spending is expected to reach $3 trillion to $4 trillion by 2030.

He believes that the proliferation of generative AI and the slowing of Moore's Law will drive companies to continually increase investments in computing, networking, and data centers. This forecast reflects Nvidia's assessment of long-term demand.

Amodei explicitly stated that slowing the pace of frontier AI model capability enhancement does not mean stopping training or technological progress. If companies allocate more time for safety testing and external evaluations, the pace of releasing certain models may slow down; model training, safety testing, inference services, and agent applications still require computing power support.

These statements currently mainly affect market expectations for the long-term demand and valuation of AI supply chain companies like Nvidia, and have not yet materialized as changes in confirmed orders or revenue. If major cloud service providers maintain their capital expenditures, demand for Nvidia GPUs and networking products will remain supported.

Going forward, close observation is needed to see whether cloud service providers cut data center investments, whether Nvidia adjusts its order or demand guidance, and whether HBM orders, selling prices, and expansion plans weaken in tandem.

Before changes occur in these indicators, it cannot be concluded that adjustments in the pace of AI development have impacted the actual demand for Nvidia products.