The index is the moat, not the marketplace.

People look at Rillor and see a marketplace. Verified buyers, verified sellers, standardized forward contracts on complete OEM GPU systems, deposits in escrow, physical delivery on every contract. It is a good marketplace, and it is live today. But the marketplace is not the defensible thing. A well-funded competitor could stand up a contract template, a KYC flow, and an escrow relationship inside a quarter. The thing they could not stand up is the number that comes out the other end.

That number is the Rillor Compute Index. It is a 30-day rolling-blend forward price per SKU, computed only from active Rillor contracts, owned and controlled by Rillor, and licensed as a settlement feed and API to exchanges, funds, and researchers. The marketplace is the mechanism that produces the inputs. The index is the asset those inputs compound into. This is a founder-voice argument for why we built the company in that order, and why every cash-settled product a third party builds on top of our prints makes the dependency harder to unwind, not easier.

What is actually scarce here

Start with what a price is. A reference price is only worth licensing if the inputs behind it are real, recent, and hard to fake. Anyone can publish an average of public spot quotes for an H100 board. Several people do. Those numbers are directionally useful and almost worthless as a settlement basis, because the inputs are scrapeable, the trades behind them are unverifiable, and a motivated party can move the print by posting quotes it never intends to honor.

Now compare that to the inputs behind a Rillor print. Every constituent is an executed forward contract between a verified buyer and a verified seller, with a 10% deposit posted into independent escrow at execution and the balance due at delivery. Nobody prints a contract they are not standing behind with real money. The trades are dated, the SKU is standardized, the quantity is known, and the counterparties are KYC'd with the NVIDIA end-customer of record captured. You cannot reconstruct that from public data because it does not exist in public data. It exists inside the venue that originated it.

That is the whole argument compressed into one sentence: the index is defensible because its inputs are proprietary to the venue that originates the trades, and Rillor is the venue. A competitor can copy the contract. They cannot copy our order flow, and the index is computed from nothing but our order flow.

Why public spot quotes can never be the basis

It helps to be concrete about what the index is not built from. The AI hardware market is opaque on purpose. A single NVIDIA H100 80GB GPU typically trades somewhere in the range of 27,000 to 40,000 dollars, an 8-GPU H200 board lands near 315,000 dollars against roughly 215,000 for the equivalent H100 board, and lead times on new enterprise GPU systems are long enough that buyers commit capital months before silicon arrives. Those are real ranges, and they are exactly the kind of number that looks like price discovery and is not. They are scraped asks, not cleared trades. They carry no delivery commitment, no deposit, no counterparty you can name.

A forward contract is a different animal, and the regulator draws the line cleanly. The CFTC defines a forward contract as a cash transaction in which a commercial buyer and seller agree on delivery of a specified quality and quantity of goods at a specified future date. Cash settlement, by contrast, pays the buyer a cash amount based on the level of an index or price. Rillor contracts sit squarely on the forward side of that line. They are bilateral OTC forwards entered with intent of physical delivery, which is why they qualify for the CFTC forward-contract exclusion and are never exchange-listed futures. Physical delivery is not a feature we market. It is the reason the inputs are real enough to compute a settlement-grade price from.

That distinction is also what keeps the moat clean downstream. Rillor never cash-settles its own contracts. The index simply describes what is happening inside the physical market we run. Third parties may build cash-settled products against the feed they license from us, but the contracts that feed the index are physical-delivery forwards, every one of them, and that is the load-bearing fact behind the whole structure.

The loop that compounds

A marketplace has a network effect on its two sides: more buyers attract more sellers and vice versa. That is real and it is also the kind of two-sided liquidity a competitor with capital can buy their way into. The index has a different and more durable loop, and it runs on the same logic that made S&P, ICE, and LBMA into franchises.

Here is the loop. Each new contract on the marketplace adds a constituent to the print for that SKU. More constituents behind each print means tighter, more representative pricing, which makes the index more reference-worthy. A more reference-worthy index attracts more licensees: an exchange that wants to list a cash-settled future, a fund that wants to hedge compute exposure, a researcher who wants a clean depreciation series. Those licensees and the products they build pull more participants toward Rillor as the place where the price is set, which originates more contracts, which deepens the constituent count again. Index providers describe this plainly. As the provider collects more data its database becomes more valuable to customers, and more customers fund more data collection.

The marketplace loop scales liquidity. The index loop scales authority. Authority is the part nobody can buy.

What the commodity markets already proved

This is not a theory. It is the structure of every mature physical commodity, and the lesson is consistent. The operator of the underlying physical market that owns the reference price captures the durable economics.

Take ICE Brent. The Brent futures contract is a deliverable contract with an option to cash-settle against the ICE Brent Index, and that index is computed from trades in the prevailing North Sea physical and forward market. The futures price stays tethered to the underlying physical market precisely because the index is built from it. Brent is the price barometer for the large majority of internationally traded crude, and the venue that owns that reference price sits on top of the entire complex.

Gold tells the same story with a cleaner separation of roles. The LBMA Gold Price is set twice daily in an electronic auction run in 30-second rounds by ICE Benchmark Administration, while LBMA owns the intellectual property and the price is a regulated benchmark. A neutral administrator runs the mechanism, a rights holder owns the print, real participant orders set the level. That is precisely the shape of the Rillor Compute Index. A neutral operator runs the contract mechanism, Rillor owns the print, real executed forwards set the level.

And the economics are not subtle. At S&P Global, the S&P Dow Jones Indices segment is a small share of total revenue (roughly one-eighth) and carries the company's highest operating margins, earned as asset-linked fees from funds tracking the benchmarks plus royalties from exchange-traded derivatives keyed to volume. The marketplace runs the market. The index earns the royalty.

Switching costs are built downstream

The moat does not only deepen as we add contracts. It deepens as other people build on us. Once a venue lists a cash-settled B300 product that settles against RIL-GX-B300-2T, or a fund books a hedge against RIL-NVL72-GB300, the RIL- print is wired into their product. Their open interest, their margin model, their counterparties' expectations all reference our number. Switching away from a widely accepted benchmark is not a procurement decision at that point. A fund that abandons the benchmark its product is built on risks losing the very investors who came for that specific exposure. The same lock-in applies to a venue that has cleared months of contracts against a print.

This is the part that is easy to miss from the outside. Rillor does not have to operate the cash-settled venues to capture this. We do not run those products. We do not cash-settle anything. Third parties run those products downstream, against a feed they license from us, and every product they ship raises the cost of ever leaving the feed. Outside exposure is downstream only by design. Third-party venues build the cash products, Rillor keeps the physical forward and the price. If you are building one of those products, our markets page covers the feed and the API and how licensing works.

Why the index always has something to compute from

There is one failure mode a reference price built from real trades has to answer. What happens when trading is thin. A print with two constituents behind it is not a benchmark, it is a rumor. We solved this structurally rather than hoping volume shows up.

Rillor is both the neutral operator of the market and a seller and underwriter that seeds liquidity. We list forward inventory ourselves, underwrite contracts, and stand behind delivery with the same deposit and seller performance bond every other seller posts. That dual role is deliberate. It guarantees the index always has live constituent trades to compute from, even in a SKU where third-party flow is light that week. It is the same reason an exchange will support a market maker of last resort. A benchmark that goes dark intermittently is not a benchmark anyone can settle against.

People sometimes hear neutral operator plus active seller and assume a conflict. The discipline that resolves it is the same discipline that protects any administered benchmark. The index is computed by rule from executed contracts, not quoted by hand, and a contract we underwrite clears the identical KYC, escrow, deposit, and delivery requirements as any other. We make the index hard to move on purpose. Our own contracts are constituents, not thumbs on the scale.

What this means for who you are

If you are an investor, the question is not whether the marketplace can be copied. Assume it can. The question is whether the reference price can be dislodged once products settle against it, and the commodity record says it cannot, cheaply, by anyone who does not own the underlying physical flow. Rillor owns the underlying physical flow because Rillor originates the contracts the index is computed from. The take rate (2% all-in, 1% from each side, plus roughly 1,000 dollars escrow per contract) funds a profitable marketplace today. The index is the asset that compounds.

If you are a venue or a fund, the strategic read is that early licensees of a benchmark are not customers, they are co-owners of its eventual authority. The products built first define what the price means. The reference for AI compute, the GB200 NVL72 rack that connects 72 Blackwell GPUs and 36 Grace CPUs at 130 TB/s of NVLink bandwidth and 13.4 TB of HBM3e, is exactly the kind of standardized, high-value, fast-depreciating asset that needs a forward curve and has never had one. Whoever helps establish that curve is positioned inside it.

If you are a buyer or an OEM, none of this changes the contract you sign. You get a real forward at standard channel pricing or a forward order book for your inventory. The index is a byproduct of you doing exactly what you would do anyway. That is the elegance of it. The moat is built out of the ordinary business of the marketplace, and it accrues to the venue without asking participants to do anything other than transact honestly.

The short version, for anyone deciding where the compounding is: the marketplace earns the fee, and the index owns the franchise. We built the marketplace so the index would have real inputs no one else can reach. Everything downstream, every cash-settled product on every venue, settles against a number that only exists because Rillor originated the trades behind it. That is the moat, and it gets wider every time a contract clears.