Traders Can Now Bet 10x on Nvidia GPU Rents, H100 at 40%

Injective launched a market on 26 October that lets traders bet on the hourly rental price of five Nvidia GPUs with up to 10x leverage, the company said in a blog post. The market tracks its new Super Intelligence Compute Index (SICI), in which the Nvidia H100 carries a 40% weight, the largest of the five chips.

The contract is a perpetual future: a bet on a price that has no expiry date, settled in cash, so traders never take delivery of any GPU time. Injective says the market runs on its mainnet, uses the USDC stablecoin as margin, and can be offered by any app built on Injective.

AI Daily Briefing · video

Today’s stories in about a minute (27 Oct 2026).

What the Index Tracks

SICI tracks the hourly rental price of five Nvidia GPUs at fixed weights, according to Injective: the H100 SXM 80GB at 40%, the B200 at 25%, the H200 at 15%, the A100 SXM 80GB at 10% and the GeForce RTX 4090 at 10%. Injective says the H100’s weight means its rental price moves the index more than any other chip.

The 4090 is a consumer graphics card. Injective says it is included because developers rent it by the hour to run open-source models, fine-tune small ones and run agents, so the index covers inference demand as well as frontier training.

How It Is Priced

The index is built and priced by LO:TECH, which Injective describes as a London-based capital markets firm that runs market making and institutional market data across more than 80 venues. Each GPU enters at a LO:TECH reference price in US dollars per GPU-hour, drawn from compute marketplaces, with on-demand and spot listings pooled.

Each chip is rebased to 100 at the 22 July 2026 daily fixing, according to Injective. The published index is the trailing seven-day average of the hourly basket, struck every hour, with the 19:00 UTC print as the daily fixing. Injective says a change to the weights ships as a new version beside the old one, never over the live index.

Why Injective Says It Built It

“AI developers face rising operating costs when compute becomes scarce, while GPU providers face declining rental revenues when prices fall,” Injective wrote. It says the index lets traders take positions on compute prices and gives AI developers and infrastructure providers a way to hedge that exposure.

SICI is Injective’s second GPU market. In August 2025 it launched what it called the first onchain market for Nvidia H100 rental rates, built with Squaretower, and it says SICI widens the basket from one GPU to five. Injective also lists pre-IPO perpetuals on OpenAI, SpaceX, Anthropic and Perplexity, and says 47% of its active perpetual markets are tied to real-world assets.

Connecting the Coverage

The launch comes as GPU rents have been rising on other indices. Our piece on the H100 rental index reported on 24 October that Silicon Data’s H100 index read $2.88 per GPU-hour, up 44% since December by our arithmetic. Our piece on B300 rents reported the next day that Silicon Data’s B300 index was up 51.9% since 31 May, against 16.9% for the B200 and 5.5% for the H100.

Our piece on the GPU forward curve reported that Silicon Data chief executive Carmen Li said its curve had turned upward this year, a sign, in her words, that “the price could go up in the future, next 36 months.” And our piece on the neocloud stocks reported on 26 October how shares of GPU cloud providers split after Elon Musk’s warning.

The Structural Read

As we read it, the weights make SICI mostly a Hopper-generation index: the H100 and H200 together are 55% of it by our arithmetic, against 25% for the Blackwell-generation B200, using the weights and descriptions Injective published.

As we read it, the seven-day average slows the index down: by our arithmetic, a jump in GPU rents that lasts one day enters the published index at about one-seventh of its size, because Injective says SICI is the trailing seven-day average of the hourly basket.

As we read it, the market adds a second public read on compute prices: our earlier piece reported Silicon Data’s forward curve turning upward this year, while SICI, as Injective describes it, is built from on-demand and spot rental listings.

Three Implications

FOR AI DEVELOPERS Injective says SICI gives developers a way to hedge rising compute costs; its launch post does not say whether any developer has done so.

FOR GPU PROVIDERS Injective names falling rental revenue as the risk providers face when prices drop, and says the index lets them hedge that exposure.

FOR TRADERS The perpetual allows up to 10x leverage, according to Injective; by our arithmetic, at full leverage a 10% move against a position equals its whole margin, before fees.

The Business Engineer Lens

This story maps onto the Business Engineer framework Tokenomics: The Economics of AI.

The framework argues that the token is at once the unit of cognition, compute, price and value, and that “Almost no other commodity in modern economic history has played all four roles at once.”

As we read it, SICI prices the step below the token: the GPU-hour that serves it, now quoted as an index and traded like a commodity. The framework was published in May 2026 and does not discuss Injective or SICI.

What Is Not Established

Injective’s launch post does not give the index’s current level, trading volume, open interest or fees, and we did not read the market’s onchain data. We could not read LO:TECH’s own description of the index, so the methodology here is as Injective describes it. Silicon Data’s indices, cited above for context, are a different product: Injective’s post names LO:TECH, not Silicon Data, as the index’s builder and pricer.

At the full 10x, a 10% move in the index against a position equals its whole margin, before fees; that is our arithmetic, not a figure from Injective. The launch post does not say who is trading the market or whether any GPU provider or AI developer has used it to hedge.

The Bottom Line

Injective has turned the hourly rent of five Nvidia GPUs into one index that can be traded with up to 10x leverage, with the H100 at 40% of the basket. It is an onchain product, priced by LO:TECH, launched in a year when GPU rents on Silicon Data’s indices have risen.