China is developing a futures market for AI tokens at the Shanghai Futures Exchange to allow companies to hedge costs of AI computation, according to multiple sources [1, 2, 3]. The AI tokens represent the smallest unit of information processed by AI models and are used to price AI services. Xiao Feng, chairman and CEO of HashKey Group, described tokens as the "digital fuel" powering AI models [1].
The design of the AI token futures product is currently preliminary and subject to change, with no clear timeline for approval or launch announced yet [1, 3]. The goal is to build a state-capitalist AI compute market that integrates enterprises, exchanges, policies, and the AI supply chain, helping stabilize enterprise AI costs and attract investment in China’s AI sector [3].
China’s daily AI token usage has surged 1000-fold since early 2024, reaching over 140 trillion tokens by March 2026 [1, 3]. This growth reflects the country’s expanding AI activities amid ongoing supply constraints for AI compute resources. The futures market aims to address these constraints and give China control over AI cost pricing and relevant financial markets [3].
Meanwhile, US exchanges like CME Group and Intercontinental Exchange are developing futures contracts based on GPU compute rental prices rather than AI tokens, signaling an alternative derivative market approach [1, 2, 3]. This marks divergent strategies in the AI derivatives space, as China focuses on AI token usage units while US markets target hardware rental pricing [1, 2, 3].
Industry observers note that AI token futures could be a new arena for US-China competition in technology, alongside current struggles over chips and data centers [3]. China may debut compute futures markets within 3 to 5 years, although a fragmented market remains a challenge [1].
Reports first surfaced on May 28, 2026 confirming Shanghai’s plans amid intensifying US-China rivalry in AI development [1, 2, 3]. The coming years will show how these financial products shape the broader AI ecosystem in China.
