The market price of AI model tokens continues to decline. The LLM Token Expenditure Index, tracked by Sil token issuance data, fell to 97 cents on Monday, reaching its lowest level since the index's launch and marking a significant drop from its summer high.
This decline reflects intensifying competition in the industry. Cheaper open-source models and price reductions by multiple leading labs are pulling down market expectations for token pricing.
Price decline compresses revenue
For model providers, a decline in token prices compresses revenue potential while computing power investments continue to require ongoing costs. Charles-Henry Monchau, Investment Director at Syz Group, noted that foundational model labs are the most directly affected.
In late July, OpenAI reduced the prices of two GPT-5.6 models. Monchau also noted that some labs have begun offering "dynamic pricing," where access fees fluctuate with demand, further increasing market price pressure.
Competition is shifting toward distribution capabilities.
Monchau believes that as the capability gap between open-source and closed-source models narrows, the focus of competition is shifting from the models themselves to distribution, memory, and contextual capabilities. In other words, the room for maintaining high prices based solely on model capability is shrinking.
This development has also prompted the market to reassess the returns on AI infrastructure investments. Both Anthropic and OpenAI secretly filed IPO documents with regulators this summer, and the downward pressure on token prices may make investors more cautious about their future profit pathways.
Tech stocks decline in tandem
AI pricing pressure has also weighed on tech stocks. On Tuesday, the Nasdaq fell nearly 1%, while the S&P 500 dropped 0.4%. Giants like NVIDIA and Microsoft, which have invested billions of dollars to expand their AI capabilities, are facing a reassessment of expected returns.