Anthropic

OpenAI ruled out going public this year, citing concerns about the security of its models. Analysts demand proof of profitability, while possible restrictions and expensive infrastructure threaten to delay projected returns into the future. S. Securities and Exchange Commission (SEC) in June. The company expects to launch the offering by mid-October at the earliest, after delaying the publication of its prospectus until late September, in a deal seeking to raise up to 100 billion dollars with a total valuation of about 2 trillion, according to the Wall Street Journal (WSJ) in August.

According to Axios, Anthropic plans to list on Nasdaq, with a possible debut in October. "The question now is not whether we are facing a bubble, but how much value AI will be able to create," says Kim Posnett of Goldman Sachs, who defines the moment as an "investment supercycle" with possible "speculative excesses." For its part, OpenAI has ruled out going public this year, its CEO, Sam Altman, announced last Saturday, linking the postponement to recent concerns about AI security, pointing to 2027 as a possible new horizon. Beyond concerns about the capabilities that the most advanced models may achieve, their development raises a financial question, as it requires millionaire investments in data centers, chips, and electricity and increases doubts about whether this enormous expenditure will ultimately generate profits.

Joachim Klement, an analyst at Panmure Liberum, points out that the declines recorded this week by Nvidia, one of the main protagonists of the accelerated expansion of AI, and other manufacturers are "a small sample" of what would happen if the current AI spending cycle ended. S. regulation unlikely in the short term, but expects greater intervention from states, especially regarding data centers. S. Nevertheless, appetite for AI-related companies has been high in the markets in recent months, albeit with volatility. 55 billion dollars in its IPO last May at 185 dollars per share, opened at 350 dollars and surpassed 385, although it now trades again around 185 dollars.

Meta CEO Mark Zuckerberg has argued that each company must take "responsibility" for ensuring its AI models are safe, thus distancing himself from proposals to coordinate a slowdown in the development of this technology. Zuckerberg directly referred to the "great debate" about the advisability of slowing down the advancement of AI capabilities until alignment standards are met, as proposed by Dario Amodei, CEO of Anthropic, and other tech leaders such as Sam Altman of OpenAI, and Elon Musk of SpaceX and xAI. He also highlighted that Meta delayed the launch of its AI agent, Muse, for several months to focus on safety and protection.

Zuckerberg's comments follow Musk's proposal earlier this week advocating for major AI developers to submit their models to security reviews and collaborate with each other to test them. In that regard, Meta's director said that MLS, its artificial intelligence laboratory, already has independent evaluators and advisors.