{
“title”: “AI Tokens Enter Commodities Trading”,
“text1”: “
The raw material of modern AI, the token, is about to get its own derivatives market. China’s Shanghai Futures Exchange is designing a derivatives market for AI tokens, according to TechCrunch AI, which cites a Reuters report. If it ships, businesses and investors could trade contracts tied to the cost of AI compute the same way they trade gold, oil, or wheat.
\n\n
This is significant because tokens are the unit everything in AI gets priced on, and until now there’s been no way to hedge their cost.
\n\n
What’s actually happening
\n\n
- \n
- Shanghai Futures Exchange is building a derivatives product tied to AI tokens, the building blocks AI models read and generate.
- CME Group and the Intercontinental Exchange (the company that owns the NYSE) are each working on futures contracts for renting GPUs, the chips that run those models.
\n
\n
\n\n
TechCrunch AI reports these are running on parallel tracks. One targets the chips. The other targets the output those chips produce.
\n\n
Why tokens, and why now
\n\n
Almost every major AI company prices its service in tokens. OpenAI charges $5 per million input tokens and $30 per million output tokens for its latest GPT-5.5 model through the API. Amazon’s Bedrock lets cloud customers pay per token too. So a token is already a real, dollar-denominated commodity. It just hasn’t had a market built around it.
\n\n
GPUs are further along. There’s a working spot market for renting them by the hour. Data from AI Mining Co., which tracks daily pricing across 28 marketplaces and cloud providers, shows median Nvidia H100 rates running from $1.40 to $4.27 an hour. H200 chips averaged between $2.34 and $5. Over the past week alone, average H100 prices swung from $2.79 to $3.33. That kind of volatility is exactly what futures markets exist to tame.
\n\n
What stands out here
\n\n
The move lands in the middle of the biggest infrastructure buildout the tech industry has ever seen. Cloud providers, private equity, and infrastructure players have poured hundreds of billions into data centers, all betting demand for compute keeps climbing. A new wave of \”neocloud\” companies is fighting for the same business, some specializing in inference, others going head to head with Oracle, AWS, and Google Cloud.
\n\n
When money moves at that scale, price risk becomes a problem worth solving. A token-based derivative ties directly to how AI companies price their services. That gives three groups a tool they didn’t have before:
\n\n
- \n
- Businesses can lock in compute costs instead of riding price swings.
- Data center operators can hedge the revenue side of their buildout.
- Investors get a clean way to bet on the direction of AI demand.
\n
\n
\n
\n\n
It’s the same logic that turned oil and grain into traded commodities. Once something is essential and its price moves, somebody builds a market to manage the risk.
\n\n
What to watch
\n\n
A few things will tell you whether this becomes real infrastructure or stays a headline.
\n\n
- \n
- Standardization. A token from GPT-5.5 isn’t the same as a token from a smaller model. Markets need a clean, agreed-upon unit to trade. How the exchanges define that contract is the whole ballgame.
- Which launches first. GPU rental futures from CME or ICE have a more mature spot market underneath them, so they may arrive before token derivatives.
- Adoption. A futures product only works if enough buyers and sellers show up. Watch whether the big AI labs and cloud providers actually participate.
\n
\n
\n
\n\n
For anyone running real workloads on AI, the practical takeaway is simple. Compute is becoming a tradeable commodity, and the cost of running your models may soon be something you can budget against with a contract instead of crossing your fingers each month.
\n\n
The pieces are still being assembled, and no contract has launched yet. But the direction is clear: AI compute is graduating into the world of commodities trading. For the full details, see the original report at TechCrunch AI.
”
}