China's daily AI token usage has surged 1,000-fold to over 140 trillion, and Shanghai is now designing futures contracts tied directly to those tokens rather than hardware rental rates.
The Shanghai Futures Exchange is laying the groundwork for AI token futures contracts, according to sources who spoke with Reuters. Regulatory approval remains unresolved and no launch date has been set, but the strategic intent is clear: Beijing wants to build the financial plumbing for artificial intelligence before American exchanges lock in their own standards.
Token consumption inside China has exploded from roughly 100 billion tokens per day in early 2024 to more than 140 trillion by the end of March. For context, a single Chinese character costs about 0.7 tokens to generate. A 15-second video requires roughly 300,000 tokens.
That breakneck growth has already broken things. Compute shortages forced DeepSeek, ByteDance's Doubao, and Zhipu AI to ration user access and suspend API services. DeepSeek suffered a nearly 12-hour outage in March. Alibaba Cloud halted sales of its baseline coding package in April, steering customers toward a premium option that frequently sells out.
Companies along the AI supply chain need a hedge against compute costs. Airlines hedge fuel. Food producers hedge wheat. The American approach, led by CME Group and Intercontinental Exchange, focuses on GPU rental rates. Shanghai is choosing a different underlying asset: the token itself.
HashKey Group CEO Xiao Feng calls tokens the "digital fuel" powering AI models. BlackRock CEO Larry Fink told a conference this month that surging token demand could spawn an entirely new asset class in compute futures. Zhang Yunquan, a researcher at the Chinese Academy of Sciences, formally proposed compute futures to China's parliament in March.
The technical distinction matters. A GPU futures contract hedges against the cost of hardware access. A token futures contract hedges against the cost of consumption. For an AI application provider spending millions on API calls, the second instrument is far more precise.
Infrastructure already exists
China published 16 compute supply indices in December covering national, regional, and hub-level data. Those indices, released by the official commodity index company, could become benchmark references for any future futures contract. The underlying data infrastructure is further along than many observers realize.
Baocheng Futures estimated in a research note this month that China could launch compute futures in three to five years, though the firm noted that the current fragmented market remains an obstacle. The U.S. has deeper financial markets and a head start on product design. ICE and CME could launch GPU futures as soon as regulators sign off.
Yet China holds structural advantages of its own. IDC data cited by China Daily shows token pricing in China runs one-sixth to one-tenth of overseas rates. Lower green energy prices and aggressive domestic chip development create that cost edge. While American exchanges build around GPUs that remain in chronic shortage, Shanghai is building around a unit of measure China already mass-produces at scale.
Yilei Shao, dean of the Shanghai AI-Finance School at East China Normal University, told Reuters that China should move faster. "The United States and China are the only two nations capable of mass-producing artificial intelligence," she said.
Who sets the pricing standard
The deeper play is about market infrastructure, not just hedging. The first exchange to establish liquid, trusted derivatives for compute could define pricing benchmarks that the rest of the world follows. London set the gold price. Chicago set the grain standard. The AI economy's equivalent is still up for grabs.
Token futures won't solve China's hardware bottleneck. They won't make a single additional GPU appear. But they would give AI companies a tool to manage the volatility of a resource they cannot live without. As AI agents and multimodal models send token consumption even higher, that tool will shift from nice-to-have to essential.
No formal filing has reached Chinese regulators yet. The timeline remains uncertain. But the signal is unmistakable: the financialization of compute has begun, and two different models are racing to become the global standard.
", "excerpt": "China's daily AI token usage has surged 1,000-fold to over 140 trillion, forcing compute rationing. Now the Shanghai Futures Exchange is designing token futures while U.S. exchanges focus on GPU rental contracts.
Also read: BYD unveils China's first 4nm self-driving chip • Europe's €120 Billion Chip Pivot: No More Blank Checks for Empty Fabs • Has the hunt for AI compute uncovered the next Cerebras?
