AI token prices have dropped sharply, raising questions about the strength of an investment boom premised on seemingly unlimited demand for computing power. No single token price measures the entire market, but several public indexes indicate comparable declines.
Silicon Data's LLM Token Expenditure Index recently fell below $1 per million tokens. The index has declined 54% from a late May peak of $2.07 per million tokens. The volume-weighted benchmark incorporates a changing model mix that increasingly favors cheaper models. Ramp, a corporate spending management company, produces another index using actual business spending on OpenAI and Anthropic. Ramp's index has fallen 48% from a late February peak of $1.22 per million tokens.
The price decline reflects corporate customers adopting cheaper, open-weight models and price reductions by OpenAI and Anthropic. Ramp lead economist Ara Kharazian said customer spending does not remain firmly attached to individual providers. Kharazian said a new model can shift billions of dollars in business spending between providers from one month to the next. Silicon Data research head Steve Hou said part of the decline comes from smarter routing that directs requests away from frontier models toward open models. Hou also said AI companies have introduced successive generations with greater capabilities and intelligence at lower prices.
OpenRouter co-founder and chief operating officer Chris Clark said growing use of agents for longer AI interactions also contributes to lower token prices. OpenRouter provides software that allows developers to access multiple AI models. Clark said large portions of longer interactions are cached instead of processed through the models again. Clark said tokens for cached inputs generally cost much less than other tokens.
Lower prices could lead to a reassessment of one of the biggest investment expansions in U.S. history. Goldman Sachs says the largest U.S. hyperscalers are on course for $800 billion in capital expenditures this year. Goldman Sachs forecasts another $1.1 trillion in spending by those hyperscalers in 2027.
The price decline has generally attracted little concern, with frequent references to the Jevons Paradox. That economic phenomenon describes technological improvements lowering prices and thereby encouraging additional economic activity. AI usage measured in tokens continues to grow very rapidly. However, Ramp's accounting shows business token spending remained flat for two consecutive months even as businesses continued using computing resources. Kharazian said Ramp was observing rising token usage alongside unchanged spending for the first time, with that pattern persisting for two straight months. Business spending on tokens is closely connected to the revenue AI model developers receive.
AI Token Prices Drop Sharply Amid Shift to Cheaper Models
FLASH AIShare
Summary
AI token prices have fallen sharply, with altcoins to watch seeing major declines. Silicon Data's LLM Token Expenditure Index dropped 54% from a late May peak to under $1 per million tokens. Ramp's index also fell 48% from a late February high. Cheaper open-weight models and price cuts from OpenAI and Anthropic are key factors. The fear and greed index shows growing caution as token costs fall. Usage is still rising, but business spending has been flat for two months.
Source:Show original
Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information.
Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.