AI companies have historically offered users vast token limits, almost as if tokens—the basic units AI processes—were endless. However, rising operational costs, global chip shortages, helium supply issues, and data center constraints have forced major AI providers to start limiting token access more strictly. This shift signals the end of the “all-you-can-eat” AI model and sparks a contest to see who can sustain demand subsidies the longest and potentially claim market dominance. Recent examples include Meta removing internal AI usage leaderboards, OpenAI moving to token-based pricing, and Anthropic restricting subscription usage in favor of APIs due to surging demand. These changes reflect the growing financial pressure in the sector, as serving AI models represents a major cost. Meanwhile, China’s AI players like Zhipu AI are raising token prices substantially, while Alibaba is offering free access to attract long-term users and developers. The scarcity of compute resources points to an ongoing balancing act between pricing, performance, and strategic market positioning.
Back
