Stagnating AI revenue and Chinese models signal potential market correction

The AI industry faces a potential market correction as revenue growth for frontier labs slows and the market share of top-tier companies drops from over 90% to the 70% range. This stagnation follows a period of rapid growth earlier this year, driven by the rise of low-cost Chinese open-weight models.

Warnings are emerging that the growth engine of the artificial intelligence (AI) industry is losing its former momentum. Analysts suggest that the formula of injecting massive capital to enhance AI model performance is faltering, which could trigger a full-scale market correction. In particular, low-cost "open-weight" models from China are rapidly encroaching on market share, slowing the revenue growth of existing frontier labs.

AI revenue indicators stall as "growth formula breaks"

Analysis suggests that Annual Recurring Revenue (ARR), a key indicator of the AI industry's health, has entered a period of stagnation. According to a discussion video from The Man Who Explains the Economy (Kim Kwang-seok TV) on October 21, 2026, the ARR indicator for Entropy—a third-party company that measures AI model performance and calculates related metrics—showed a steep upward curve from January to June this year but has remained in a stagnant zone from July to the present.

Expert Kim Dante diagnosed this phenomenon, stating that "the revenue growth speed of the frontier labs that led the AI cycle is slowing down." He noted that while the market was dominated by expectations earlier this year that companies like Entropy and OpenAI would spend massive amounts on computing resources, "that flow has now flattened out." This suggests that the existing growth formula of continuing capital expenditure (CAPEX) to increase AI model performance may no longer be working.

This revenue stagnation could have a chain reaction across the entire AI ecosystem. Lee Hyo-seok explained that for the AI cycle to function, "Entropy (AI companies) must generate revenue, use that money to purchase computing resources from data centers, and then data centers, in turn, contract for things like memory semiconductors." In other words, if the revenue growth of the top-tier AI companies stops, the investment momentum flowing down to the lower value chain could weaken. Lee added that while a slowdown in growth is inevitable, the scale has become so large that "the 100-fold growth of Entropy's revenue from $100 million last year to $10 billion by the end of this year was largely due to a base effect."

Counterattack of low-cost Chinese models: "Winning with cost-effectiveness over intelligence"

The spread of "open-weight models" led by China is cited as one of the primary reasons for the stagnation in revenue growth. Open-weight models allow anyone to use them by making the model's weights public, a method Chinese companies are utilizing to distribute copies of Western models at low costs.

Kim Dante cited changes in actual tech startups as evidence. He said, "In the past, startups mainly used OpenAI products by paying high costs, but recently, they are building their own servers or installing Chinese models (GLM, Kimi, etc.) to use." He explained that a "routing" method is spreading, where users judge the difficulty of a question and assign easy tasks to cheap Chinese models while sending only high-difficulty questions to expensive pay-per-use models. Kim analyzed that this trend is being aided by the fact that many questions users ask ChatGPT or Claude involve tasks like summarization, translation, and research, which do not require extremely high intelligence.

Data from the website Open Router, which shows the status of AI API usage, supports this. The revenue market share of frontier labs, which exceeded 90% at the beginning of this year, has fallen to the 70% range, with Chinese open-weight models filling the void. Kim Dante analyzed that "while Chinese models were disappointing in terms of performance at the beginning of this year, they have now risen to a sufficiently usable level," adding that "the emergence of ultra-low-cost models is changing the market landscape."

Conflicting views on whether the AI correction has "begun or ended"

Opinions among experts are divided regarding which stage the current market correction is in. Using different "baseball game" metaphors, three experts diagnosed the current position regarding the market's direction.

Kim Dante defined the current situation as the "beginning of the correction (around the 3rd inning)." Based on the stagnation of AI companies' revenue and the expanding market share of Chinese models, he left open the possibility of further declines. Conversely, Lee Hyo-seok viewed it as being in the "middle of the correction (around the 5th inning)," noting that "since the problems have already been revealed, we are in the middle." He explained, "When a correction first occurs, it starts without people knowing what the problem is, but now we know what the problem is."

Professor Kim Kwang-seok judged the current state to be in the "middle-to-late stages of the correction (6th to 7th inning)" based on the KOSPI index. This view suggests that while there is room for further adjustment, the end of the correction is not far off. Although they presented different indicators and grounds, the experts reached a consensus that the growth engine of the AI industry is transitioning from a period of explosive expansion to a mature stage where efficiency and cost-effectiveness are increasingly prioritized.

Source: original video (YouTube)