This week, the LLM Token Expenditure Index, a measure of daily prices for large-language model tokens from Silicon Data, fell to 97 cents, marking its lowest point since its inception late last year and a decline of over 50% from earlier summer highs.
This drop suggests that users of AI models, such as OpenAI's ChatGPT and Google's Gemini, may face lower operational costs, but it poses challenges for the companies providing these models. As prices decrease, consumer expectations for lower access rates may diminish the pricing power of these providers.
The decline is partly attributed to the emergence of competitive open-source models, like Moonshot's Kimi K3, which offer lower pricing than established players. Additionally, OpenAI's recent price cuts for its GPT-5.6 models and the introduction of dynamic pricing by other labs further exacerbate the downward pressure on token prices.
Charles-Henry Monchau from Syz Group noted that this token deflation could compress revenues for companies like Anthropic and OpenAI, especially as they consider public offerings. The broader implications for investors include a potential reevaluation of expected returns on investments in AI, particularly as major tech firms like Nvidia and Microsoft continue to invest heavily in AI capabilities.
The overall market sentiment was reflected in a decline in technology stocks, with the Nasdaq Composite falling nearly 1% and the S&P 500 down 0.4%