Citadel CEO Says AI Is "in the Middle of a Bubble," While CITIC Securities Says Long-Term Value of Chinese and US Models Remains Attractive

Ken Griffin, CEO of Citadel, said the artificial intelligence (AI) industry is now in a bubble. However, Chinese brokerage firm CITIC Securities remains optimistic about the long-term business value of large AI models from China and the United States, as well as their potential to increase in value.

Speaking at the "Future of Global Markets" conference in New York on the 20th, local time, Griffin said, "We are almost certainly in the middle of an AI bubble." He explained that major changes in the economy often happen alongside bubbles. He added, "If you insist on using the word bubble, such bubbles often occur alongside massive technological transformation." In an interview with Bloomberg, he said, "The ability of American companies to create value from AI models is highly uneven." He also said that every company needs to examine its bills from Anthropic and OpenAI carefully and think about how much more productive these AI tools can make it.

Griffin is now much more positive about AI than he was in the past. At the Davos forum, he called AI "garbage," and during a lecture at Stanford Graduate School of Business, he described its results as "depressing." He said, "The companies that survive on the other side of the bubble go on to lead the market for decades after it deflates." He added, "You had to hold on and not sell Amazon stock back then." He also predicted that AI technology would lead to the creation of many new startups in the United States.

Investors are increasingly worried about whether the huge amounts of money being spent on AI will lead to real profits. These concerns have raised doubts about whether the AI boom can last. The Wall Street Journal reported that Anthropic could spend $750 billion over the next few years just to secure computing power, ahead of what would be the largest initial public offering (IPO) ever. Griffin stressed that Anthropic's success depends on proving that AI models are essential across all industries.

In a report released on the same day, CITIC Securities said, "The market's attitude toward large models has become cautious, but we remain firmly optimistic about the long-term commercial value and valuation growth room of Chinese and US models." The brokerage said that the full process of reasoning, searching for information and combining it into useful answers (RSI) is improving faster, while AI models are also being updated more quickly. It added that the combined growth in annual recurring revenue (ARR) at OpenAI and Anthropic remains strong. The firm found that the ability to make money from AI models and the level of their use vary greatly. This is similar to the early stages of the mobile internet, when businesses were just beginning to make money from the technology. CITIC Securities believes there is still plenty of room for AI businesses to grow, even if growth comes only from more companies, uses and regions adopting these models.

CITIC Securities expects the fourth quarter of 2026 to be an important period for confirming whether overseas AI models are being widely adopted by businesses. It said Anthropic's quarter-on-quarter growth in annual recurring revenue slowed somewhat because of increased competition after GPT-6-Astra launched in August and September, as well as factors such as reverse distillation, also called anti-distillation. However, the firm expects Anthropic's ARR to start rising again after October. It pointed to the launch of Opus5.5 in late September and expectations for future models such as Fable 5.5 ahead of Anthropic's IPO. These developments could ease the market's excessive pessimism about the AI industry.

For Chinese AI models, CITIC Securities said that the shortage of data needed for further model training is becoming less severe as demand for data labeling grows rapidly. It believes Chinese-made models will eventually find ways to make progress in this stage of training as well. The brokerage said that the ARR of Chinese-made models is currently only at the level that overseas models reached 18 months ago, while their valuations are just one-tenth as high as those of overseas models. If Chinese models achieve a major improvement in performance, their low starting revenue, fast growth and attractive valuations could lead investors to value them much more highly across the market, the firm said.

The views of Griffin and CITIC Securities show that investors are still debating whether the money spent on AI can be earned back. Concerns about a possible bubble may cause greater price swings in the short term. Once AI models begin generating business results on a large scale, proof of their success is expected to play a key role in helping investors reassess their value.