Bitget launches GPU rental-price perpetuals for traders

Bitget has launched H100USDT and B200USDT pre-market compute perpetuals, which track Nvidia H100 and B200 GPU rental prices.

The move gives retail traders access to perpetual contracts linked to GPU rental-price indices, a market that until now has largely been confined to institutional and infrastructure circles.

According to Bitget, the instruments track hourly rental costs for Nvidia's H100 and B200 graphics processing units, which are widely used in artificial intelligence workloads. The contracts are quoted against indices expressed in US dollars per GPU-hour and are settled in USDT.

Trading is available around the clock with leverage of up to 10 times. The products are based on H100 and B200 GPU Rental Price Indices published by Silicon Data, which are designed to show standardised average hourly rental costs across a range of GPU cloud and compute providers.

Compute market

GPU access has become an increasingly important part of the artificial intelligence economy as demand for model training and inference has grown. Rental prices for top-end chips can shift with supply constraints, cloud availability and changing demand from AI developers.

That has driven interest in turning compute pricing into a market benchmark instead of leaving it fragmented across bilateral deals and cloud platform listings. Bitget said the indices standardise variables such as GPU specifications, rental duration, geography and rental terms to create a clearer reference point.

The launch comes as traditional derivatives markets also begin to target the sector. CME Group is scheduled to introduce futures tied to H100 and B200 GPU rental prices, suggesting compute costs are emerging as a distinct market signal tied to the broader build-out of AI infrastructure.

Bitget's approach brings that theme into crypto-linked trading through perpetual contracts rather than dated futures. The structure allows traders to take positions on shifts in GPU rental prices without buying hardware or entering direct cloud rental arrangements.

Broader offering

The new contracts sit within Bitget's wider Universal Exchange model, which combines crypto markets with tokenised and other asset exposures. In that context, GPU compute adds a category tied to the economics of artificial intelligence infrastructure rather than to digital tokens or conventional financial assets alone.

Bitget says it serves more than 125 million users and offers access across crypto tokens, tokenised stocks, exchange-traded funds, commodities, foreign exchange and precious metals. The addition of compute-linked products suggests the exchange is testing how far traders will engage with infrastructure costs as a standalone market.

Retail access could also broaden price discovery in an area where transparency has been limited. GPU pricing has often varied between providers and regions, and market participants have had few ways to express a direct view on the cost of compute itself.

For traders, the appeal is likely to depend on whether GPU rental prices become volatile enough to support active speculation and hedging. For the wider market, the launch is another sign that AI-related inputs such as chips, energy and data centre capacity are being treated as financial variables in their own right.

Bitget's leadership accompanied the launch with a factual note.

"AI has created enormous demand for chips, data centers and energy, and computing capacity, making compute an increasingly important part of the digital economy. As compute develops a more transparent market price, crypto-native infrastructure creates an opportunity to make that market accessible in new ways. Bringing GPU rental-price perpetuals to Bitget reflects the UEX vision of giving users access to both established assets and emerging markets within a single trading environment," said Gracy Chen, Chief Executive Officer of Bitget.