AI Token Costs Hit Record Low as OpenAI Price War Intensifies

The cost of artificial intelligence is falling. This could potentially benefit businesses and developers while creating a new challenge for companies spending billions of dollars to build frontier AI models.

Silicon Data’s LLM Token Expenditure Index fell to a record-low 97 cents on Monday. The benchmark measures the effective market price of large language model tokens, and has now dropped by more than half from its peak earlier this summer.

(Source: CNBC)

The decline forms part of an aggressive AI price war. Cheaper open-source and open-weight models, particularly from Chinese developers, are giving businesses alternatives to premium offerings from OpenAI, Anthropic and Google. Moonshot AI’s Kimi models are among the competitors helping to push prices lower.

OpenAI also responded by cutting prices across parts of its GPT-5.6 lineup, which added even more downward pressure to the market.

Lower token prices are good news for developers and businesses deploying AI at scale. Cheaper inference could make AI agents, coding assistants, customer service systems and enterprise automation a lot more economical. For frontier AI labs, however, the economics are a bit more complicated.

Companies including OpenAI and Anthropic commit huge amounts of capital to computing infrastructure while the amount they can charge per token is falling. That creates a potential margin squeeze if growing AI usage cannot offset declining unit prices.

Anthropic, for example, has continued expanding its long-term compute capacity through major infrastructure agreements to compete with OpenAI, Google and other model providers.

The pressure may also change where AI companies look for competitive advantages. If models capable of handling most everyday tasks become interchangeable, simply offering a powerful model may no longer be enough. Instead, companies could compete through distribution, enterprise integrations, proprietary data, persistent memory, AI agents and broader software ecosystems.

The trend could also have consequences beyond the model developers themselves. Nvidia, Microsoft and other technology giants have invested heavily in data centers, chips and cloud infrastructure on expectations of enormous long-term AI demand.

That raises one of the biggest questions facing the AI industry: whether soaring demand for artificial intelligence can grow quickly enough to compensate for the falling price of intelligence itself.