AI frontier slowdown could give second tier leg up

Intel

14/09/2026 13:00

The author is a Reuters Breakingviews columnist. The opinions expressed are her own. Refiles to spell out Dario Amodei's three-step framework in paragraph two.
By Robyn Mak
HONG KONG, Sept 14 (Reuters Breakingviews) - The world should be worried when the planet's three most influential figures in artificial intelligence all agree it's time to slow down. Over the weekend, Anthropic CEO Dario Amodei published a blog post calling on companies to moderate the pace at which they advance model capabilities, drawing support from arch-rivals Sam Altman at OpenAI and Elon Musk, who owns xAI. But while AI doomsayers have fresh reason to fret about mankind's future, markets and investors could have, for now, less to fear.
Amodei's lengthy missive lays out a three-step framework for boosting safety including embedding third-party evaluators into companies, greater coordination among frontier firms and ensuring international cooperation to handle emerging risks. It comes just days after a researcher resigned from his company, warning in a now-viral post that AI "could kill us all by the end of the decade". Humans losing control over bots is a well-grounded fear: in July, a group of 700 OpenAI agents autonomously hacked open-source platform Hugging Face, the latest incident of AI going rogue. Amodei references that attack and admits that in six to 12 months, "such a swarm will be capable of taking over the entire internet".
Details on what guardrails to erect and how exactly they will be implemented will matter. But broadly speaking more oversight would help accelerate the industry's shift from developing and training advanced models to using what already exists at scale, in what's referred to as inference. Currently, most of Anthropic and OpenAI's computing capacity combined is allocated to training, with just 37% going toward answering prompts or executing tasks, estimates research firm SemiAnalysis. That's up from 29% in early 2024.
The good news is that demand for agents to code, manage e-mails and calendars and other tasks is taking off, which is already driving demand for chips and infrastructure. By 2030, inference will account for 43% of data-centre demand, one and a half times the share for AI training, reckons McKinsey. Any slowdown in model advancements, then, would probably divert a larger share of the forecast $1 trillion in global AI spending this year to enabling the use of AI. 
That should insulate so-called hyperscalers like Amazon AMZN.O, Alphabet GOOGL.O and Meta META.O, which make up the bulk of capital spending to some degree. Yet current AI winners like $5.3 trillion Nvidia NVDA.O, whose graphics processing units are vital for training models, and South Korea's SK Hynix 000660.KS, which supplies high-bandwidth memory for such chips, look worse off.
Intel INTC.O, for example, which specialises in another type of semiconductor called central processing units, may be a major beneficiary. Data centres catered to training advanced models generally require 1 CPU for every 8 GPUs; boss Lip Bu Tan has flagged that for iterative tasks carried out by agents, the ratio is one to one. Nvidia challengers that focus on inference, like Cerebras Systems, also stand to gain, as well as Samsung Electronics 005930.KS, whose conventional memory products are more suited for CPUs.
With less emphasis on frontier AI capabilities, model developers can also better compete on efficiency and costs. That will reshape competition and business models. So investors need not panic about a sudden market downturn. Whether that's enough to save the world from an impending machine-driven apocalypse is another matter.
Follow Robyn Mak on X.
     
CONTEXT NEWS
Anthropic CEO Dario Amodei on September 12 outlined a three-step framework intended to pace development and create more time to manage its risks. The proposal immediately received endorsements from several other executives at major AI companies, including Elon Musk, CEO of Grok-owner SpaceX, and OpenAI boss Sam Altman.
The plan calls for embedded independent evaluators with employee-like access to verify safety practices, coordination among frontier AI firms to set safety standards and limit unchecked AI development, and international cooperation to manage AI risks.
Separately, Altman said in an interview with Forbes published the same day that OpenAI will not go public in 2026, citing safety concerns over artificial intelligence.

(Editing by Antony Currie; Production by Ujjaini Dutta)
((For previous columns by the author, Reuters customers can click on MAK/  robyn.mak@thomsonreuters.com))