The rapid emergence of highly capable, ultra-cheap Chinese open-weight artificial intelligence models has sent shockwaves through Wall Street, triggering fears that the AI sector is entering a deflationary death spiral. However, leading industry analysts argue that the plummeting cost of intelligence, driven by Chinese innovation, will actually benefit the global AI industry in the long run by supercharging demand, a classic manifestation of the Jevons Paradox.
According to the South China Morning Post, companies across the AI ecosystem have been forced to slash prices aggressively in recent weeks. Data from research firm Silicon Data’s LLM Token Expenditure Index reveals that large language model (LLM) inference prices per million tokens fell from above $2 at the start of June to just $1.2 this week. This index tracks pricing across both frontier proprietary providers and open-weight platforms, highlighting a dramatic, industry-wide contraction in the cost of computing power.
Under intense pressure from cheaper Chinese alternatives like DeepSeek and Moonshot AI, Silicon Valley heavyweights have initiated defensive price cuts to protect their market share. OpenAI, for instance, recently announced a massive 80 percent discount on developer pricing for its lightweight GPT-5.6 Luna model, alongside a 20 percent discount on its mid-tier GPT-5.6 Terra offering. These aggressive pricing adjustments contributed to a severe AI stock sell-off last month, as investors fretted that US hyperscalers were fundamentally overvalued in a race to the bottom.
The Jevons Paradox in Action
Despite the market anxiety, analysts point to the Jevons Paradox to explain why these price cuts are ultimately bullish for the sector. The paradox, an economic concept dating back to the 19th century, observes that as technological progress increases the efficiency with which a resource is used (reducing its cost), the overall rate of consumption of that resource actually rises, rather than falls. In the context of AI, cheaper tokens will unlock vast new use cases that were previously economically unviable.
“Competition is up and prices are down,” Silicon Data noted on the social media platform X. “This is good for consumer and enterprise users of AI (agents) and promotes much wider and faster AI adoption.” When the cost of querying an LLM drops by 80 percent, developers can afford to run more complex, multi-step reasoning tasks, deploy autonomous agents that make hundreds of API calls per minute, and integrate AI into low-margin consumer applications.
This dynamic is already visible in the Chinese domestic market. The fierce price war among Chinese AI developers has resulted in some models being offered essentially for free, driving an explosion in token consumption and experimental application development. By forcing Western companies to match this pricing trajectory, Chinese models are effectively accelerating the global transition from experimental AI pilots to ubiquitous, embedded AI infrastructure.
The Threat to Proprietary Moats
While the Jevons Paradox suggests the overall pie will grow, the distribution of that pie is shifting rapidly. The commoditization of the model layer poses a severe threat to companies that built their business models on charging premium access fees for proprietary intelligence. As open-weight models from Chinese labs achieve parity with closed US systems, the willingness of enterprises to pay a premium for standard inference tasks evaporates.
This forces companies like OpenAI and Anthropic to continuously push the frontier of capabilities just to justify their pricing, while the “good enough” middle of the market is entirely captured by open-weight alternatives. The rapid rise of models like Kimi K3 demonstrates that Chinese labs are not only competing on price but are also capable of matching or exceeding Western performance on key benchmarks, particularly in coding and reasoning tasks.
For Silicon Valley hardware providers and cloud hyperscalers, however, the trend is overwhelmingly positive. If cheaper models drive a massive increase in total token consumption, the underlying demand for compute infrastructure, data centers, networking gear, and GPUs, will continue to soar. In this sense, China’s aggressive open-weight strategy is acting as a massive subsidy for global AI adoption, accelerating the timeline for AI to become a foundational utility akin to electricity or broadband internet.
What the Jevons Paradox Means for China’s AI Strategy
For Beijing, the Jevons Paradox is not an unintended side effect of Chinese AI development, it is the intended outcome. China’s open-weight AI strategy is designed to commoditize intelligence, making AI services so cheap and widely available that adoption becomes universal and irreversible. By releasing powerful models for free, Chinese labs are not sacrificing revenue; they are building the global infrastructure dependency that will eventually generate enormous strategic and economic leverage.
This strategy mirrors China’s approach in the solar panel and lithium battery industries, where massive state-backed investment drove down costs to the point where the entire world became dependent on Chinese supply chains. The difference is that AI models, being digital public goods, can be distributed globally at near-zero marginal cost. Once a developer in Kenya, a factory in Vietnam, or a hospital in Brazil integrates a Chinese open-weight model into their operations, switching costs become significant, creating durable lock-in.
For US policymakers, the Jevons Paradox presents a particularly challenging dilemma. Restricting access to Chinese models would harm American developers who benefit from the intense price competition. Allowing unrestricted access accelerates the global adoption of Chinese AI infrastructure.
There is no clean policy solution that simultaneously protects US commercial interests and prevents the global spread of Chinese AI. The market, guided by the Jevons Paradox, is making the decision for them: cheap, capable intelligence will be adopted wherever it is available, regardless of its country of origin.