Nebius Publishes Live GPU Spot Prices in European AI Push
Nebius (NASDAQ: NBIS) Group NV has made live prices for its interruptible GPU computing capacity publicly available, expanding visibility into cloud costs as it seeks more business from European artificial intelligence developers.
The Amsterdam-based company announced the move at Nvidia’s GTC event in Berlin on Wednesday. Prices are available on its website and through Shadeform, Silicon Data and GPU Finder, according to its statement.
Nebius’s published pricing lists starting spot rates of $0.79 per GPU-hour for Nvidia H100 and H200 processors, compared with on-demand rates of $4.50 and $5.40, respectively. B200 and B300 spot capacity starts at $0.99 per GPU-hour, against on-demand rates of $8.50 and $9.50. Those starting prices exclude taxes and can change with demand and available capacity.
The public display extends a pricing system Nebius announced on September 22 and said would take effect on October 8. That system calculates rates by GPU type and region, with updates as frequently as every 15 minutes. Customers previously viewed current prices when configuring computing resources in the company’s cloud console.
The potential savings come with a trade-off. Nebius can reclaim the capacity, and customers who set a maximum price can have their workloads stopped when rates exceed that limit. The company describes the service as suitable for tasks such as experimentation, data processing and short model fine-tuning jobs.
Selling spare computing capacity at variable prices is an established cloud model. Amazon Web Services also offers interruptible spot instances, recommending them for applications that can tolerate disruptions. That makes the ability to recover from interruptions an important consideration alongside the advertised hourly rate.
Nebius is also testing a rebate program that allows customers to recover value by returning idle GPUs to its spot pool. Other purchasing options include auctions for short-term capacity blocks and pay-as-you-go access, the company said.
The announcement accompanies a European expansion of its AI Builder Program, including $10 million in additional compute grants. An October 19 company post outlined support for open-source research and developer tools, connections to potential investors through Accel and other venture firms, and community events with Nvidia.
Nebius is broadening its sales channels as well. Distributor TD Synnex announced a pan-European agreement on October 20 to give its partners access to Nebius’s AI cloud for model training and inference, including reserved-capacity options and technical support.
