Nvidia plans to hike rental prices for its H100 graphics processing units by 20% in 2026. The move will affect companies that lease the chips from cloud providers for artificial intelligence and high-performance computing workloads. The increase, confirmed by Nvidia, is the first notable price adjustment for the H100 line in over a year.
The size of the jump
The 20% rise applies to H100 rentals, not direct purchases. Nvidia hasn't said whether the change covers all cloud partners or only specific regions. The H100 has been the dominant AI training chip since its launch, and rental demand has stayed high as companies race to build and run large language models. A price increase of this scale could push some customers to look at alternatives, though few options match the H100's performance.
Why now
Nvidia did not give a specific reason for the 2026 hike. The company typically adjusts pricing based on production costs, supply chain constraints, and market demand. The H100 uses advanced manufacturing processes, and Nvidia has faced component shortages in the past. The announcement comes as the AI chip market grows more competitive, with rivals like AMD and Intel pushing their own products. Still, Nvidia holds the largest share of the data-center GPU market.
Impact on cloud customers
Major cloud providers — including Amazon Web Services, Microsoft Azure, and Google Cloud — offer H100 instances. Those companies will likely pass the increase on to their own customers. For startups and research labs that rely on rented H100 capacity, the 20% jump could strain budgets. Some may shift to cheaper chips or reserve capacity earlier to lock in current rates. Nvidia has not announced any special pricing for long-term contracts.
The 20% rental price rise is set to take effect at the start of 2026. How cloud providers will adjust their own pricing before then remains an open question.
