THE TEARDOWN
The biggest cost in AI has no hedge. In three days, the CFTC decides whether it gets one.
The Structural Signal
CME Group and Silicon Data filed two compute futures on August 11, 2026. They track the hourly cost of renting Nvidia GPUs.
The target launch was October 5 on NYMEX. The CFTC paused it.
On September 21 the regulator extended its review by 49 days, to November 9. It cited novel or complex issues.
The broader comment period on compute derivatives closed October 20. The contracts still have no launch date.
CME is not hiding its interest. On the third-quarter call on October 21, CEO Terry Duffy framed the product as risk management for participants.
Derek Sammann reported strong demand from neoclouds, data centre firms, banks and funds. CME posted record quarterly revenue of $1.7 billion.
A new asset class would add a line to that. This is a product launch wearing a public service argument.
The Mechanical Breakdown
Each contract covers 730 GPU-hours. That is one month of rental for a single chip.
Two contracts sit in the filing. GPU1 tracks the Nvidia H100, and GPU2 tracks the newer Blackwell B200.
Both settle in cash. Nobody ships a GPU, and the trade is a bet on rental price.
Now the index, which is where the product actually lives. Silicon Data publishes a daily price from hourly on-demand rental rates.
The sample is the critical choice. It pulls from neoclouds, the smaller GPU-focused hosts.
CoreWeave, Lambda and Crusoe are in it. AWS, Azure and Google Cloud are not.
Read that decision carefully. The neocloud tier is where prices swing most and where data is thinnest.
That is the segment with a visible spot price, so it is the only one you can build an index from. It is also the least representative of where most compute is actually bought.
The index read $2.88 per GPU-hour on October 24. That is up 44% from $2.00 in December 2025.
It holds IOSCO compliance, the global benchmark standard. Data partner dxFeed put it on Refinitiv feeds.
So Wall Street can see the price today. It cannot trade it.
Now the ownership question. Carmen Li, a former Bloomberg data executive, founded Silicon Data in 2024.
The company raised $30.5 million this year from CME Ventures, DRW, Samsung Next and VanEck. The exchange invested in the benchmark its own contract would settle against.
That cuts both ways and both should be stated. Alignment gets the product built, and it also means the venue and the price source are not independent.
The CFTC’s concerns are not hard to infer. Rental prices are scattered across dozens of small providers.
GPU-hours cannot be stored. An unused hour expires, which kills the cost-of-carry link that anchors every other commodity curve.
A thin market with few price sources is a manipulation surface. The regulator wants proof the benchmark holds under pressure.
There is a prior question underneath it. Does compute meet the legal test for a commodity when every rack differs in speed, location and software?
Forget Cash... BoA Just Said It's Over
On January 24th, 2022, Bank of America told Bloomberg something that should terrify every American with a savings account…
The digital dollar is inevitable. Implementation is expected between 2025 and 2030.
Most Americans missed it completely.
This is your bank talking.
The bank where your neighbors have mortgages. Where your family keeps savings. Maybe where you have your retirement money.
And they just admitted they're about to trap your wealth in a digital cage.
Think about what this means.
When the digital dollar goes live, every dollar you saved will suddenly be trackable at the transaction level. Subject to new digital withdrawal limits. Able to be frozen remotely without court approval. Programmable with spending restrictions.
Bank of America knows this. They're helping build it. And they're telling you it's inevitable.
But here's what Bank of America didn't tell Bloomberg.
There's still a way to legally protect your wealth before this digital prison locks.
The escape route?
Become your own bank before they flip the switch.
I'm an ex-Wall Street investment banker and CFA Charterholder who witnessed this plan firsthand.
And I'm not the only one moving.
The largest institutions on earth are already exiting the system they helped build. BlackRock launched its tokenized treasury fund directly on-chain. JPMorgan processes over $1 billion daily through its own blockchain. Visa is settling stablecoin transactions outside the traditional banking system entirely.
They're not waiting for the digital dollar. They're building the exit.
While everyone else gets trapped in Bank of America's digital system, my research team has been positioning members outside that reach… completely under their own control.
Our research methodology was independently audited by Conquest Investment Advisory AG, a German firm regulated by BaFin.
Across 571 research signals, 86.34% rose more than 20%, with an average historical return of 416.88% to all-time high.
No government tracking. No withdrawal limits. No remote freezing.
Financial sovereignty.
I just released an emergency video showing exactly how to become your own bank before the digital dollar launches.
This information won't be available once the system goes live.
Legacy vs. Autonomous
The legacy market is private deals. AI labs sign long leases with cloud hosts or neoclouds.
Prices are not public. There is no forward curve, so nobody knows what six months out should cost.
Without futures, the only hedge is a lease. That locks volume, not price.
A listed contract changes the shape. A lab could hedge compute the way an airline hedges jet fuel.
Market makers supply liquidity. Price discovery moves from phone calls to a screen.
Now where it fails, and the failure is structural rather than teething.
Basis risk is the first. A cash-settled index contract hedges the index, not the cost of one specific cluster.
The gap between a neocloud spot rate and a hyperscaler reserved contract can run for months. A lab hedging with GPU1 is hedging a correlated thing, not its own bill.
Storability is the second. With no carry, the forward price is pure expectation and nothing arbitrages it back.
Concentration is the third. If a handful of providers set most of the input prices, the benchmark is easy to lean on.
The AI angle here is honest and worth stating plainly. No model runs this contract.
The settlement logic is an index and a formula. What AI created is the commodity itself, and AI-driven trading firms will set the pace once volume exists.
Do you remember the subprime crisis of 2008?
Wall Street built a gigantic "pyramid of risk" on subprime loans in the housing market.
And then it all came tumbling down, along with almost the entire financial system.
Dr. Martin Weiss says a new $3 trillion "subprime" crisis is underway.
And it's already starting to spin out of control.
Capital Flow Implications
Four pools move.
The first is CME’s. Clearing fees, data sales and a forward curve it owns.
The GPU rental market ran about $52 billion in 2026, per Mordor Intelligence. Futures on a slice of that is a meaningful new fee line.
The second is brokerage. Three commodity brokers have lined up to take orders, per Bloomberg on October 20.
ADM Investor Services, BGC Group and Nanhua USA. BGC went earlier still, launching an over-the-counter compute desk in June under Zach Espinosa and Marc Kuber.
The third is the benchmark franchise, and it is the durable one. Silicon Data becomes the reference price for compute.
That is the role price reporting agencies play in oil. Index licensing outlives the products built on it.
The fourth pool shrinks. Private compute brokers earn on information asymmetry.
A public screen kills that spread. The buyer who had to call three people now checks a quote.
ICE is the competitive overhang. It announced rival futures in May with index provider Ornn.
If CME clears first, ICE loses the opening in a new asset class. Benchmarks are winner-take-most.
Verdict
The review expires November 9. If it clears, AI’s core input becomes a listed commodity for the first time.
What compresses is the private compute brokerage spread and the information advantage of whoever had the best phone list. What expands is exchange revenue and a benchmark licence nobody owned a year ago.
Watch open interest, not the launch. A contract with a ticker and no volume is a press release.
Watch the basis between the index and hyperscaler contract rates. If that gap stays wide, the hedge does not work for the buyers who need it most.
Territory: +machines on the asset, +incumbents on the venue.
A public price for compute kills the spread on every private GPU deal, and the listing, clearing and index fees land with an exchange and a benchmark it part-owns.

