AI-related stocks retreated in different markets this Monday (14), after executives from some of the sector's leading companies advocated a slowdown in the development of the most advanced models. The pressure was concentrated on chipmakers and suppliers of the infrastructure used to train and operate AI systems.
The move began after Dario Amodei, CEO of Anthropic, published on Saturday (12) an essay arguing that companies should reduce the pace of advancement of their models' capabilities to allow safety mechanisms to keep up with the evolution of technology. Sam Altman, of OpenAI, and Elon Musk, of xAI, expressed support for the proposal.
In Asia, the South Korea's Kospi closed down 3.3%, pressured by semiconductor manufacturers. SoftBank, one of OpenAI's main investors, lost more than 10% in Tokyo. In Europe, companies in the sector also retreated, with ASML among the biggest decliners.
In the United States, the Philadelphia semiconductor index fell close to 6% at the start of the session. Nvidia lost more than 3%, while AMD, Intel and Marvell posted declines near or above 5%. The Nasdaq also opened lower, although it recovered part of the losses over the course of the day.
The reaction shows the dependence created between AI growth expectations and investments in chips, servers, data centers and energy. Investors' fear is that a deliberate reduction in the pace of model development could moderate demand for the infrastructure that supported part of the valuation of technology companies in recent years.
Anthropic proposes permanent external oversight
Amodei said that slowing down does not mean interrupting model training or technical progress. The proposal calls for giving more time for testing, alignment and implementation of safeguards before new leaps in capability.
The plan has three stages. The first calls for independent teams with continuous access to companies' systems to verify safety practices and investigate incidents. Anthropic said it will adopt this measure even without an agreement with competitors. The remaining stages depend on coordination between companies and governments, including international negotiations on limits to the development of advanced systems.
Amodei justified the change by the recent acceleration in the ability of AI models to participate in the development of new generations of the technology itself. He also cited risks related to cyberattacks, misuse and loss of control over autonomous agents.
Altman said he agrees with the need to control the pace of the technological frontier and that OpenAI also intends to allow external evaluations. The executive also said that the company does not intend to hold its initial public offering this year, amid safety concerns.
Not all companies were affected in the same way. While hardware makers fell, shares of software companies such as Adobe, ServiceNow and Workday advanced, in a rotation based on the expectation that a slower evolution of AI could reduce some of the competitive pressure on their businesses.
The session was also pressured by factors outside the sector. Oil once again surpassed US$ 108 a barrel, while investors increased bets on an interest rate hike by the Federal Reserve this week, adding pressure to growth and technology stocks.
