Strategic Intelligence: Tokens and the Economics of AI

Strategic Intelligence: Tokens and the Economics of AI

Summary

AI token costs are falling, but this is being offset by accelerated token generation by AI agents and reasoning models. It is becoming clear that instead of scaling like classical software, Jevons paradox applies here: cheaper tokens make demand rise exponentially, keeping total costs high. Task complexity will also continue to increase, which means more tokens for given workloads.

Key Highlights

• US AI infrastructure capex levels will remain elevated through 2027. US hyperscalers can sustain $1 trillion in spending through 2027, supported by massive order backlogs, mixed data center ownership, and debt financing. However, pressure will start to build in 2028, potentially forcing them to cut capex.
• Low-priced, high-performance LLM models from China are giving users more choice. Evidence of the pressure China’s open-weight models are putting on the US frontier models is the significant price cuts announced by OpenAI in the summer of 2026.

Scope

• This report investigates the economics of token consumption, the impact of the surge in token demand on the overall prospects of the large language model (LLM) business model, and potential future capex investments.
• It takes a closer look at China’s hyperscalers and the role of the country’s telecom companies, which have already begun offering token plans similar to mobile data plans.
• Finally, the report will address what enterprises must do to manage costs in such a dynamic environment.

Reasons to Buy

• Enterprises must adopt best practices to manage token costs efficiently. Enterprises will need to maintain a portfolio of AI models to manage costs and protect corporate data. Small and medium-sized enterprises (SMEs) are already exploring Chinese open-weight models, while large enterprises are looking at US- and EU-based open-weight models, including those from companies such as Cohere and Astral.
• This report will help readers make sense of the latest developments in the AI industry and adapt their AI strategy in response.