Financial organizations are racing to build new infrastructure for LLM tokens, which could become the most significant market in the future.
According to Reuters, China’s Shanghai Futures Exchange is presently creating a derivatives market for AI tokens. The announcement coincides with separate announcements from the NYSE’s owner, the Intercontinental Exchange, and major derivatives exchange CME Group that they are preparing to introduce futures contracts for GPU rentals.
Although GPU markets are still developing, there is a sizable market for spot rates on GPU rental, which are usually priced by the hour, due to the variety of businesses that use, sell, and rent GPUs. The median prices for Nvidia H100 GPUs ranged from $1.40 to $4.27 per hour across 13 marketplaces, while the average price for H200 GPUs was between $2.34 and $5 per hour across 10 marketplaces, according to data from AI Mining Co., which tracks daily GPU rental pricing across 28 marketplaces and cloud providers. Additionally, the average H100 price during the last seven days has varied between $2.79 and $3.33.
However, there is less infrastructure surrounding tokens, which are the essential components of modern AI models, even though there are developed marketplaces for GPUs. Major AI companies often charge $5 per million for input tokens and $30 per million for output tokens if you want to utilize the API for their most recent GPT-5.5 model. As with Amazon’s Bedrock system, cloud companies are increasingly providing the option to charge per token.
The endeavor takes place in the midst of an unparalleled expansion of AI infrastructure. In anticipation of the continued growth in demand for GPUs and computation, cloud service providers, private equity firms, and infrastructure players have invested hundreds of billions in the construction of data centers. A growing number of international neocloud businesses are also competing for a share of this market. While some of these new players are vying with cloud behemoths like Oracle, AWS, and Google Cloud to provide their services to AI firms, others are specializing in inference.
By focusing on AI tokens, the Shanghai exchange’s derivative product would be linked to the pricing of AI firms’ services, providing investors, enterprises, and data center operators with a means of protecting themselves from compute costs.