The Challenge of Pricing AI Services
Companies are struggling to set prices for AI services due to the unpredictable nature of token consumption
The development of Large Language Models (LLMs) has been a significant investment for firms like Microsoft, Google, and Anthropic, with hundreds of billions of dollars spent. As a result, these companies are now offering paid-for versions of their AI services with extra features. However, setting a price for these services is proving to be difficult.
## Tokenomics and Pricing The problem lies in the rapidly changing economics around tokens, the building blocks of LLMs and agentic AI. When a user asks an LLM to answer a question or generate software code, the prompt is broken down into mathematical chunks called tokens. The LLM's response also comes in the form of tokens, which are converted back into text, software code, or a set of commands. This process is not entirely predictable, making it challenging to determine the cost of using these services.
According to analysis by Goldman Sachs, the cost of individual tokens has plummeted in recent years, but the number of tokens consumed by businesses and consumers has skyrocketed. The bank forecasts that external token consumption will increase 24 times between 2026 and 2030 to 120 quadrillion tokens a month. This unpredictability is making it difficult for companies to manage their token costs.
## Managing Token Costs Companies are finding ways to work around this issue. Some smaller organisations are using flat fee personal accounts, which may not be sustainable in the long term. Others are thinking more carefully about what AI models to use and being more precise with their prompts. However, the situation can become difficult to control when companies build AI into a product that could be rolled out to thousands of users.
## Pricing Strategies Companies are still figuring out how to charge for AI services. Options include raising prices across the board, paying by results, or charging for bundles of incidents. However, whatever price structure is chosen could be upended if the large language model providers change their own pricing strategies. As Bill Peterson, senior director of product marketing at Sumo Logic, says, "Nobody's really figured it out." The company is previewing new security services based on agentic AI but is still in discussion with corporate customers about how to charge for them.
