Injective Launches Onchain Futures Market Using Nvidia GPU Cost Data

Injective has launched a new perpetual futures market designed to let traders speculate on changes in the cost of renting high-end Nvidia GPUs. The market is linked to the Super Intelligence Compute Index (SICI), which tracks rental prices for five Nvidia chip models and settles positions using USDC as margin.

The initiative aims to translate volatility in AI infrastructure—especially GPU rental rates—into a tradable on-chain risk product. Injective says the index is updated hourly using a trailing seven-day average and is intended both for traders seeking direction and for companies that want to hedge computing expenses.

Key takeaways

• Injective’s Super Intelligence Compute Index (SICI) tracks rental prices for five Nvidia GPUs: H100, B200, H200, A100, and RTX 4090.
• The index is weighted most heavily toward the H100 (40%) and B200 (25%), with hourly updates based on a trailing seven-day rental-price average.
• Traders can take long or short positions on whether the index rises or falls using USDC margin, with leverage up to 10x.
• Injective expands on an earlier market launched in August 2025 that was tied to a single GPU model (Nvidia H100) by broadening the framework into a five-chip basket.
• The launch comes amid heightened scrutiny of whether AI “neocloud” providers can sustain returns as infrastructure costs rise.

How the Super Intelligence Compute Index works

According to Injective’s announcement, SICI is designed to measure the rental pricing of Nvidia GPUs commonly used for AI workloads. The index covers five models: H100, B200, H200, A100, and RTX 4090.

Injective states that the weighting scheme prioritizes the most widely demanded accelerators: the H100 holds the largest weight at 40%, followed by the B200 at 25%. The remaining weight is distributed across H200, A100, and RTX 4090, though Injective’s post highlights the H100 and B200 proportions as the dominant components.

The index calculation is based on rental-price data sourced from compute marketplaces compiled by London-based market-data firm LO:TECH. Injective says the index is updated hourly and uses a trailing seven-day average, a method intended to smooth short-term spikes that can distort day-to-day pricing.

On-chain, the perpetual futures market lets participants place bets on whether the index will move upward or downward. Injective adds that USDC is used as margin, and that the product supports up to 10x leverage.

From single-GPU pricing to a multi-GPU benchmark

This new launch builds directly on Injective’s earlier step into GPU-linked derivatives. In August 2025, the network introduced a market tied specifically to Nvidia H100 rental prices. With SICI, Injective shifts from a single-asset reference point to a broader basket spanning five different GPUs.

That change matters because it potentially reduces reliance on one particular chip’s market dynamics. AI demand can rotate across GPU types depending on performance characteristics, availability, and pricing on rental markets. A multi-GPU index may therefore better reflect the overall economics of renting compute for model training and inference than a single-model benchmark.

Injective frames the product as a way to mirror real-world pricing drivers in AI infrastructure: if rental rates rise and fall with demand, then a futures market linked to those rates can provide an instrument for both directional trading and risk management.

Why GPU rental volatility is becoming a financial issue

The arrival of a GPU-price-linked derivatives market comes at a time when many AI infrastructure providers are under pressure to prove they can generate returns after large capital expenditures and ongoing operational costs.

Research firm Gartner warned in October that more than 80% of specialist AI cloud providers—often referred to as “neoclouds”—could cease operating independently by 2030 due to difficulties sustaining adequate returns. The concern is not simply whether demand for AI computing exists, but whether providers can price access in a way that covers the economics of building and running data center infrastructure.

Elon Musk echoed similar worries Sunday, saying more than 200 neocloud providers could face failure or consolidation if lenders become reluctant to finance companies whose revenue depends heavily on AI data center output.

Commentary around these risks extends beyond the provider side. Major technology firms have also been restructuring and cost-cutting as they invest heavily in AI systems. Bloomberg reported Monday that Microsoft is offering voluntary buyouts to about 60,000 eligible employees as part of cost containment, while it forecasts roughly $175 billion in capital expenditures for 2026.

Meanwhile, Meta—described in the same coverage as having already invested more than $100 billion in AI—cut 8,000 jobs in May as part of a restructuring effort tied to controlling spending.

Who the product may serve: hedgers and risk takers

Injective’s stated motivation centers on the idea that rental markets for GPUs behave like operating input markets for AI builders. Companies developing AI models often rent computing power from cloud and marketplace providers such as CoreWeave and Nebius rather than purchasing Nvidia hardware outright.

In that setting, rental rates function as a variable cost. When those costs rise, it can disrupt budgets for training runs and iterative development. When rental prices fall, it can change the calculus for running larger experiments or deploying inference at scale.

Injective suggests that a futures market tied to the rental-price index could help hedge against higher computing costs. At the same time, GPU providers could use the same type of pricing-linked instrument to protect against declines in rental revenue.

For traders, the immediate use case is straightforward: they can take positions on whether the index value will increase or decrease, using USDC margin and leverage up to 10x. For long-horizon market participants, the bigger question is whether the index pricing mechanism and the smoothing method (hourly updates using a trailing seven-day average) produce a benchmark that remains stable enough to support hedging strategies—especially during periods of rapid shifts in AI demand.

What remains to be seen is how widely the SICI-linked market will be used—whether by AI infrastructure operators seeking hedges, by trading firms looking for compute-cost exposure, or by more traditional crypto derivatives participants seeking a new underlying tied to real-world data center economics.

Investors and builders watching this space should track not only adoption and liquidity for SICI-linked perps, but also whether broader concerns about neocloud profitability and AI infrastructure spending translate into sustained volatility in GPU rental prices—the input this market is built to reflect.