Forward contracts versus futures for GPU systems, and why Rillor is a forward.

Two instruments can both let you commit today to a price for a GPU system you take delivery of later. They sound interchangeable in a procurement meeting, and they are not. One is a forward contract. The other is a future. The difference is not pedantic. It decides whether you post variation margin every night, whether your counterparty is an anonymous clearing house or a named OEM, and whether the contract ends with a wire and a price difference or with a 48U liquid-cooled rack arriving on your dock.

A Rillor contract is a forward, and it is built that way on purpose. This piece draws the line between the two structures, explains why standardizing six contract fields does not turn a forward into a future, and shows why a Rillor forward sits squarely inside the CFTC's forward-contract exclusion. If you run a procurement or treasury function and you are evaluating which instrument to use to lock compute capacity, the distinction below is the one that should drive the decision.

What a forward contract actually is

A forward is the older and simpler of the two instruments. Two parties agree privately to transact a specific asset at a specific price on a specific future date. The contract is customized to those parties, it trades over the counter rather than on an exchange, and it settles once, at maturity, using the original contracted price. There is no central clearing house standing between the two sides, and no daily exchange of money to track how the price has moved in the interim.

A Rillor contract is a forward in exactly this sense, with one structural refinement. It is not a two-party deal. It is a tri-party delivery obligation among a verified buyer, an OEM, and Rillor as the operator. The buyer commits to take physical delivery of a named SKU at a named facility in a named delivery month, at a fixed price. The OEM commits to ship. Rillor sits in the middle: it runs KYC on both sides, captures the end-customer-of-record for the OEM's NVIDIA business unit, holds the deposit with an independent escrow agent, requires a seller performance bond, and orchestrates the handoff. We walk that structure clause by clause in Anatomy of a Rillor forward contract.

The economic intent is the load-bearing detail. A Rillor contract exists because a buyer wants the hardware and a seller wants to ship it. Price certainty is a benefit of the structure, not the purpose of it. That single fact, intent to physically settle, is what separates the entire instrument class from a future.

What a futures contract is, and why it differs

A futures contract is a standardized legal agreement to buy or sell a defined quantity of an asset at a predetermined price for delivery at a specified time, between parties who are not yet known to each other when they trade. It is engineered for an exchange. The CFTC describes the mechanics plainly: futures are typically traded on organized exchanges that set standardized terms, are cleared through a clearing house that acts as the buyer to every seller and the seller to every buyer, and are marked to the market daily, with traders required to post margin that is typically between two and ten percent of contract value.

Unpack each of those and you have the four properties that define a future, none of which a Rillor forward has.

The clearing house is the structural heart of a future. When you trade a listed future, the exchange's clearing house steps between you and the other side and becomes your counterparty. You no longer face the person who took the other end of the trade. That novation is what lets futures trade anonymously and at scale, and it is the thing that requires daily mark-to-market and margin: the clearing house has to keep both sides collateralized against price moves because it is on the hook to both. A Rillor forward has no clearing house. The buyer faces a specific OEM, and Rillor is the operator and, where it underwrites liquidity, a named seller. There is nothing to novate and nothing to margin daily.

The CFTC forward-contract exclusion, and why Rillor sits inside it

US commodity law draws the same line, and it draws it in Rillor's favor. The Commodity Exchange Act's definition of a swap carves out forwards. The exclusion covers any sale of a nonfinancial commodity for deferred shipment or delivery, so long as the transaction is intended to be physically settled. Commentators restate the test as three elements: the commodity must be a nonfinancial commodity, the transaction must be a sale for deferred shipment or delivery, and the parties must intend to physically settle.

Map a Rillor contract onto those three elements:

1. Nonfinancial commodity. The underlying is a complete OEM GPU system, a physical good. A Supermicro SYS-A22GA-NBRT, a Gigabyte G894-AD1-AAX5, a Dell PowerEdge XE9780, a GB200 NVL72 rack. These are not financial instruments. They are iron.
2. Sale for deferred shipment or delivery. The contract fixes a delivery month in the future and an obligation to ship into it. Deferred delivery is the entire point of a forward market for capacity you need later, which is why buyers use it instead of waiting on an allocation queue, a contrast you can see playing out across the marketplace.
3. Intent to physically settle. This is where most instruments that claim the exclusion actually fail, and where Rillor is unambiguous. The CFTC infers intent from whether the parties genuinely make or take delivery in the ordinary course of business, rather than merely trading price exposure. A Rillor buyer is a tier-2 cloud, a sovereign AI program, or an enterprise buildout that needs the racks. A Rillor seller is an OEM or an underwriter that ships them. Nobody on a Rillor contract is there to net out a price difference and walk away. The structure mechanically prevents it: there is no daily settlement to offset against and no cash-settlement path at maturity.

The historical premise behind the exclusion is that the regulatory scheme for derivatives should not reach private commercial merchandising transactions that create enforceable obligations to deliver. That is a precise description of what Rillor operates. Bilateral commercial deals, between parties in the business of buying and selling the underlying hardware, that end in delivery.

Why standardizing six fields does not create a future

Here is the objection a careful reader raises, and it deserves a direct answer. Rillor standardizes its contracts. Every contract uses the same six fields: SKU, quantity, delivery month, deposit percentage, settlement currency, and channel-of-record. Standardization is usually named as a defining trait of futures. So does standardizing turn a Rillor forward into a future.

No, and the reason is that standardization and exchange-clearing are two different things that happen to co-occur on futures exchanges. A future is standardized because it has to be fungible enough for a clearing house to net thousands of positions against each other. The standardization is in service of the clearing and the anonymous secondary market. Strip away the exchange listing and the central counterparty, and standardized terms are just good contract hygiene. They make contracts comparable and they make them transferable, which is genuinely useful: a Rillor contract can be transferred pre-delivery to another KYC'd buyer, with Rillor and OEM approval, a mechanism that is a core feature of the marketplace.

But comparable and transferable is not the same as exchange-listed and centrally cleared. A standardized forward still faces a named counterparty. It still settles once, at delivery, at the contracted price. It still has no daily variation margin. The CFTC's test does not ask whether a contract's terms are uniform. It asks about venue, clearing, daily settlement, and delivery intent. A standardized Rillor forward answers that test the same way a one-off forward would. The broader case for standardized terms runs through how Rillor sources and lists supply for buyers.

Physical delivery, always, with no daily margin

The cash-flow profile is the cleanest way to feel the difference as a buyer. On a Rillor forward there are exactly two payments, and a single delivery event.

At execution, the buyer posts a ten percent deposit into the independent escrow agent's account. Between execution and delivery, nothing happens to your cash. There is no nightly variation margin, no maintenance margin threshold, no margin call if the prevailing forward price drifts away from your contracted price. The price you struck is the price you pay. At delivery, the OEM ships the named SKU to the receiving facility, Rillor confirms receipt with the buyer, and the escrow agent releases the ninety percent balance.

Contrast that with a listed future. Because the clearing house marks your position to market every session, an adverse move means you wire variation margin that night to stay current, and a string of adverse moves can force a sequence of margin calls regardless of your conviction about the underlying. The instrument is designed to keep a leveraged, anonymous market collateralized in real time. That machinery is appropriate for a financial product. It is friction and risk for a buyer whose actual goal is to own racks. The full case for keeping settlement physical runs through everything Rillor does for buyers.

Where cash settlement does belong: downstream, on the index

None of this means cash-settled compute derivatives are illegitimate. They are real, they are growing, and they have a clear home. They just are not Rillor's contracts.

The pattern in the market is consistent. Third-party venues build cash-settled derivatives on top of a price benchmark rather than on physical chips. A venue lists a futures or perpetual contract whose underlying is not a deliverable rack but a published index of compute prices, and positions settle in cash against where that index prints. The hardware never moves. What changes hands is the price difference. That is exactly the financialization layer that benchmarks such as Brent, Henry Hub, and the LBMA gold price enabled in their own commodities.

Rillor's role in that layer is the index, not the derivative. The Rillor Compute Index is a 30-day rolling-blend forward price per SKU, computed from active Rillor contracts, owned and controlled by Rillor, and licensed as a settlement feed and API to exchanges, funds, and researchers. A venue can license the index and build a cash-settled B300 perpetual or a quarterly future against it. When it does, Rillor supplies the reference price and nothing else. Rillor does not operate that venue, does not clear those positions, and never cash-settles. Our own contracts stay physical forwards. The index is the part that scales into finance, and it is the durable asset of the business. You can read how the licensing side works on the markets page, and the underlying contract mechanics on the marketplace.

This is the cleanest way to hold the whole structure in your head. Rillor forwards deliver hardware and feed a price. The price feeds derivatives that other people build and settle in cash. The two layers never blur, because Rillor never lets a forward contract end in anything but delivery.

Why the distinction matters to a buyer

For a procurement or treasury team, the choice of instrument is a choice of risk profile. Pick a future and you accept three things: an anonymous counterparty intermediated by a clearing house, daily mark-to-market with the margin calls that implies, and a settlement that resolves in cash against an index rather than in the asset you actually wanted. If your goal is price exposure, that may be exactly right. If your goal is to operate the hardware, every one of those traits is a mismatch.

Pick a Rillor forward and you get the inverse. A named OEM as your delivery counterparty, with a seller performance bond behind it. No variation margin and no margin calls, because there is no daily settlement. And no basis-to-cash risk, the gap between where an index settles and what it actually costs you to acquire the iron, because there is no cash settlement at all. The contract ends with the racks on your floor at the price you struck. A buyer's full playbook for using the instrument lives on the buyer page.

Forwards and futures are both legitimate, and a mature compute market needs both. The work is in not confusing them. A Rillor contract is a bilateral, over-the-counter, physically settled forward that sits inside the CFTC forward-contract exclusion by design. It is standardized for comparability and transfer, never listed or cleared. And it delivers iron, every time.