Nebius will raise rates for select Nvidia chips by 17% to 21% beginning Oct. 1. It is the company’s second price increase in three months, reflecting the shortage of computing capacity needed to build and run AI models.
The Amsterdam-based neocloud company rents powerful chips and data center infrastructure to businesses developing AI products.
Prices for Nvidia H100 instances will climb about 17%, while rates for newer B300 GPUs will increase roughly 21% under the latest changes.
Nebius is also charging more for certain on-demand processors and memory products, extending the increases beyond GPUs.
AI capacity is getting more expensive
Rates for some CPU-only instances will rise 25%, while certain memory offerings will cost about 41% more. Advanced Micro Devices EPYC Genoa processor rates are also set to increase 25%.
Customers that reserve large clusters for several months can secure lower rates, concentrating the biggest increases among those buying capacity on demand.
The latest adjustment brings the total increase for B300 instances to roughly 56% in only a few months.
Nebius has not disclosed the expected financial impact of the new rates. CoreWeave said this week that it is signing contracts at higher prices, pointing to broader pricing strength among AI infrastructure providers.
CoreWeave raised prices by around 25% in July, though much revenue at neocloud companies remains tied to multiyear contracts.
Older AI chips still have pricing power
Nebius's increase includes Nvidia H100 chips, a processor first released in 2022. Skeptics have argued cloud providers may overstate chip value by spreading Nvidia hardware costs over six years.
Nebius's ability to charge more for H100 rentals suggests depreciation concerns have not yet weakened spot demand for older AI chips.
"AI demand remains off the charts."
Daniel Newman, Futurum Group
The company signed four customer contracts averaging more than $1B each after rapid growth in its AI cloud business.
Nebius also secured about $775M in July financing backed by deployed GPU infrastructure and contracted customer cash flows. The price increases show that AI compute remains a seller's market, especially for customers that need capacity without long-term reservations.
