$NBIS Expands October 1 Price Hikes Across Compute and Inference as Capacity Squeeze Persists

Nebius ($NBIS) is extending its October 1 price hikes across both raw cloud compute infrastructure and its Token Factory Pay-As-You-Go (PAYG) Dedicated Endpoints, raising rates by roughly 16% to 21% across high-demand Nvidia ($NVDA) enterprise accelerators.
The price adjustments apply to both Hopper and next-generation Blackwell architectures:
- Nvidia H100: Increasing ~16.9% ($3.85–$4.05 base to $4.50–$4.70/hr)
- Nvidia H200: Increasing ~19%–20% ($4.50–$4.70 to $5.40–$5.60/hr)
- Nvidia B200: Increasing ~18%–19% ($7.15–$7.40 to $8.50–$8.70/hr)
- Nvidia B300: Increasing ~20%–21% ($7.85–$8.10 to $9.50–$9.70/hr)
Raising rates concurrently across on-demand clusters and managed inference endpoints demonstrates that specialized AI neoclouds continue to hold strong pricing power. Demand for both large-scale model training runs and low-latency API inference deployment remains structurally ahead of aggregate GPU supply and data center power capacity, allowing operators to expand gross margins without dampening cluster utilization.