AI bubble risk is heating up: how China’s low-cost competition is reshaping valuation logic for tech stocks

Economist Peter Schiff recently issued a warning: the AI stock bubble may already have burst. He said Kimi K3 from Dark Side of the Moon is intensifying low-cost AI competition in China, putting pressure on the valuations of U.S. AI-related companies. AI-related stocks have been weak; SPCX is down more than 46% from its historical high. Schiff also specifically emphasized a key distinction: “AI isn’t a bubble, but AI stocks are.”

Against this backdrop, China’s AI models are reshaping the global competitive landscape with astonishing cost efficiency. The weighted average cost to run DeepSeek V4 Flash for standardized intelligent tasks is just 2 cents, while Anthropic’s comparable model is as high as $2.75. Multiple U.S. firms have started switching vendors—after Lindy AI moved from Anthropic to DeepSeek, its costs fell to one-tenth of the previous level. OpenRouter data shows that U.S. companies’ share of Token usage for China-based AI models is the highest, reaching 46%. Even Microsoft is evaluating adding DeepSeek to Copilot as a low-cost option.

The shock that low-cost competition brings to tech stock valuations is structural. In recent years, the high valuations of U.S. AI giants were built on three core assumptions: compute spending is highly correlated with model capability; closed-source models can maintain technical moats and high gross margins; and demand for AI infrastructure will keep exploding. But China’s model rise is challenging these assumptions one by one—DeepSeek trains powerful models with far less compute than OpenAI, while open-source models threaten proprietary “moats” and their ability to command premium pricing. As one investor put it: “If China’s open models are not only cheaper, but also start competing head-on in the high-end market with U.S. closed-source models, how long can the AI valuations built in the past few years hold up?”

However, there is a deep split in the market. The cautious camp says typical bubble-like features have already appeared locally, with some segments’ valuations entirely entering the momentum-and-sentiment-driven stage. The optimistic camp believes this round of the AI rally is backed by real industry demand and orders—industry supply growth is far below demand growth, and order fulfillment is only 30% to 40%. Goldman Sachs also asserted that Chinese AI stocks are “nowhere near bubble territory.”

The deeper issue is this: although the future picture of AI looks impressive, applications that can create sustainable revenue are still limited. The market is shifting from “stacking Tokens” to requiring AI to prove its commercial value at lower cost. China’s rise in low-cost AI may be accelerating this value reset—not denying the revolutionary potential of AI technology, but puncturing the capital-market valuation frenzy surrounding AI themes. #夏日创作营