China has begun designing a futures market for AI tokens. The effort, still in its earliest phases, marks a distinct approach from American exchanges racing to list contracts on GPU compute power. Sources close to the planning told Reuters the Shanghai Futures Exchange is studying contracts tied to the smallest units of information that AI models process.
Tokens serve as the basic building block. They measure consumption inside large language models. One source described the research as preliminary. Another noted the work draws partial motivation from the intensifying competition with the United States.
But the contrast stands out. CME Group and Intercontinental Exchange have moved toward futures linked to the cost of renting computing hardware. Shanghai’s version would connect instead to how companies price AI services themselves. Both products aim to give participants across the supply chain a way to protect against swings in the expense of powering intelligence at scale.
The timing feels urgent. China’s daily token usage jumped one thousand times from the start of 2024. It reached more than 140 trillion by the end of March. Official data backs the figure. Liu Liehong, administrator of the National Data Administration, called tokens “the settlement unit linking technological supply with commercial demand.” China Daily reported the remarks.
Xiao Feng, chairman and CEO of HashKey Group, offered a vivid comparison. Tokens function as the “digital fuel” or “raw material” that powers AI models, he said in comments relayed by Reuters. BlackRock CEO Larry Fink went further at a conference this month. Surging demand for tokens could spawn an entirely new asset class, he suggested, particularly through futures on compute.
Shortages have already bitten. Several Chinese AI models rationed access in recent months because computing power ran tight. The pressure highlights why hedging matters. Companies burning through trillions of tokens daily face real volatility in costs. A standardized futures contract could bring predictability. It could also create price signals that guide investment in data centers, chips and energy.
Yet plenty remains unclear. No launch date has been set. Regulatory approval from the China Securities Regulatory Commission lies ahead. The Shanghai Futures Exchange itself, long a center for commodity and metals trading, has not commented publicly. Neither has the CSRC. Plans could still shift.
Brokerage Baocheng Futures sketched a longer horizon in a note issued earlier this month. China might debut compute futures in three to five years. Fragmented markets present one obstacle. Building liquidity in a brand new asset class will take time. Still, the intent looks serious.
Beijing treats artificial intelligence as a strategic sector. Growth depends on it. Officials have accelerated construction of a spot market for computing power. Data center operators, model developers and heavy users all take part. The token economy now sits at the center.
Fortune captured the shift in April. China coined an official term for the concept: ciyuan. It translates roughly as “word currency.” The National Data Administration unveiled it at a State Council press conference. Daily processing hit 140 trillion tokens. That compared with 100 billion at the beginning of 2024. Chinese models even surpassed American ones on OpenRouter, a marketplace that ranks AI offerings by performance and price. Fortune laid out the numbers.
Price wars add fuel. Xiaomi slashed API costs by as much as 99 percent for its MiMo-V2.5 series late last month. DeepSeek made permanent steep discounts on its V4-Pro model. Chinese inference now runs at fractions of Western rates. Jefferies analysts estimated local models cost on average one-sixth as much per token as offerings from OpenAI or Anthropic. The Reuters piece on token obsession from mid-April flagged the dynamic. Reuters Breakingviews analyzed the risks.
TechCrunch drew the commodity parallel one day after the original report. Just as traders buy and sell contracts on gold or oil, participants may soon hedge exposure to AI tokens. The Shanghai exchange’s product would tie directly to service pricing. Businesses, investors and data center owners could lock in costs. The piece quoted the Reuters exclusive while expanding on infrastructure implications. TechCrunch explored the comparison.
China Daily returned to the subject this week. Token consumption climbed from 100 billion daily at the start of 2024 to 100 trillion by the end of 2025. It now exceeds 140 trillion. The newspaper quoted Liu again. The surge shows China’s AI development has entered rapid growth. It also signals maturing commercialization. The opinion column highlighted potential.
Conversations on X reflect the excitement. One widely shared post from Crypto Banter noted the explosion to 140 trillion tokens by March. It framed the move as financialization of AI output itself. “Think oil, electricity, or bandwidth… now tokens,” the account wrote. Others called it a clear signal that AI is being treated as critical infrastructure.
Skeptics exist. Reuters Breakingviews warned that tokens are only as good as the models behind them. China risks falling behind in foundational capabilities even as it floods the market with cheap access. Model quality, data advantages and chip restrictions all factor in. The token rush began with viral digital assistants. Whether it sustains depends on real business adoption beyond entertainment and experimentation.
Still, the trajectory looks steep. Global users have discovered Chinese models through platforms like OpenClaw. Bloomberg documented the export angle in April. Affordable services draw overseas demand. Token consumption becomes both a domestic metric and a potential tradeable good. Bloomberg described the surge in international usage.
The Shanghai Futures Exchange holds a unique position. Its experience in metals and commodities gives it tools to design contracts that reflect physical-like delivery of compute. Token futures would not deliver literal bits. They would instead settle against benchmarks of AI service costs or consumption indices. Details have yet to emerge.
American efforts focus on the hardware layer. GPU futures from CME and ICE target the rental rates for the silicon that makes inference possible. Both paths address the same problem: extreme price sensitivity in an industry where demand grows faster than supply. One bets on the shovel. The other bets on the gold.
Either way, the financial industry senses opportunity. Larry Fink’s comments suggest Wall Street already imagines structured products, ETFs and derivatives built on top of AI consumption data. BlackRock’s interest adds weight. If tokens become an asset class, the firms that standardize measurement and create benchmarks will hold influence.
China’s fragmented computing market complicates progress. Different regions and operators run incompatible systems. A national spot market seeks to change that. Futures could accelerate standardization. Liquidity would follow if enough participants see value in hedging.
Recent price cuts illustrate the pressure. Xiaomi’s 99 percent reduction on certain models aims to capture share. DeepSeek’s permanent discounts lock in lower rates. The strategy boosts volume. It also trains models on more data and widens the user base. Tokens keep flowing.
Analysts debate long-term effects. Cheaper access may spur innovation. It may also compress margins for developers. Export of tokens, or rather the services that consume them, could become a new category in trade statistics. Beijing has begun touting the possibility.
For now the futures project remains quiet. Early stage research. No formal application. Regulators will weigh risks of speculation, manipulation and systemic exposure to technology volatility. Commodity futures succeeded because underlying markets were deep. AI tokens represent something newer. The data exists in abundance. The question is whether it can be packaged into transparent, tradable instruments.
Industry insiders watch closely. Data center operators want cost certainty. Model companies seek stable revenue signals. Investors look for exposure without owning chips or renting racks. A successful token futures contract could pull all three together.
The Reuters report landed on May 28. Coverage spread quickly across crypto news sites and social platforms. TechCrunch, TradingView and CoinDesk-adjacent accounts amplified the story within hours. Discussions on X highlighted the shift from hardware to output. One post called it “turning AI usage itself into a tradable asset class.”
China Daily’s follow-up three days later reinforced the narrative. Tokens anchor value in the AI era. They quantify business models. The newspaper positioned China as poised to lead token economics thanks to its massive internet population and policy support.
Challenges remain. Power shortages. Semiconductor limits. Quality gaps in frontier models. None of those erase the consumption numbers. One thousand times growth in two years commands attention. Financial markets respond to scale.
So the work continues at the Shanghai Futures Exchange. Quietly. Methodically. A new contract taking shape that could, if launched, change how companies large and small manage exposure to the intelligence economy. The U.S. builds its own derivatives. China charts a parallel but different course. The race has more than one track.