A closely watched gauge measuring the price of AI tokens has this week sunk to an unprecedented low, marking the latest signal of weakening prices within an increasingly competitive sector.

The Large Language Model Token Expenditure Index, published by research firm Silicon Data, dropped to $0.97 on Monday. This represents the lowest reading since the index was launched late last year, reflecting a decline of more than 50% from its peak earlier this summer.

Silicon Data's index tracks the average market price for a single LLM token. This steep price decrease indicates that users querying mainstream chatbots like OpenAI's ChatGPT, Anthropic's Claude, and Google's Gemini will face lower fees. However, for companies developing foundation models, this trend is far from positive. Falling index prices cultivate user expectations for service price cuts, thereby weakening providers' pricing power.

At the end of July, OpenAI reduced prices for two of its GPT-5.6 models. Additionally, Charles-Henri Moncho, Chief Investment Officer at Syz Group, noted that other frontier labs are also launching products with "dynamic pricing" features, where API call rates can fluctuate based on market demand. These developments are adding further downward pressure on the token market.

"Foundation model vendors are bearing the brunt of this impact," Moncho wrote. "Token deflation is squeezing revenue while computing procurement costs remain rigid. A strategic pivot is already visible: the moat must transition from raw model performance—where the gap between open-weight and closed frontier models is now measured in mere months—to distribution capabilities, memory features, and context capacity."

Moncho added that the declining comprehensive cost of generating a single token within the industry is also pushing the index lower. The sustained decline in the LLM Token Expenditure Index could place fresh profitability pressures on AI leaders like Anthropic and OpenAI, both of which are evaluating the timing and viability of public listings. This summer, both companies confidentially submitted IPO filings to regulators. As token prices fall, investors may need to recalibrate their expectations regarding returns on capital invested in the AI sector.

Major technology conglomerates, including Nvidia and Microsoft, have already poured tens of billions of dollars into expanding AI computing capacity. Steve Hou, Head of Research at Silicon Data, commented that the recent price drop demonstrates that current supply, combined with both cutting-edge closed-source models and cost-effective alternatives, might already be sufficient to "meet the performance requirements of the vast majority of tasks."