Nebius announced a second price hike for Nvidia GPU and other compute instances, reflecting the rapid surge in AI training and inference demand.
Nebius, the Russian cloud‑computing arm of Sberbank, announced a second increase in prices for its Nvidia GPU‑powered and other high‑performance compute instances, citing a “rapid surge” in demand for artificial‑intelligence training and inference workloads.
Why the price hike matters
The new rates, which apply to both on‑demand and reserved instances, push Nebius’s pricing above that of many Western providers, raising concerns for startups and research labs that rely on affordable access to large‑scale GPU clusters.
Nebius officials said the adjustment reflects “the unprecedented growth in AI workloads across the region,” and is intended to sustain the company’s ability to expand its hardware inventory and maintain service reliability.
Scope of the increase
The hike targets Nvidia A100 and H100 GPU instances, as well as CPU‑only compute nodes used for data preprocessing and model serving. Nebius did not disclose the exact percentage change, but industry observers estimate a rise of roughly 10‑15 % compared with the previous pricing tier introduced six months ago.
Market reaction
Customers have expressed mixed reactions. Some large enterprises view the increase as a predictable response to soaring demand, while smaller AI startups fear that higher costs could delay or scale back experimental projects.
- Potential slowdown in AI research budgets
- Shift to alternative cloud providers abroad
- Increased focus on cost‑optimization techniques such as model pruning
Nebius said it will continue to invest in expanding its data‑center capacity and exploring partnerships with local hardware manufacturers to mitigate future price pressures.