Nebius Hikes AI Cloud Prices 21%, Memory Up 41% - shattered.io
π Nebius announced a 21% price increase for Nvidia B300 instances and a massive 41% hike for memory-backed capacity effective October 1, 2026.
π Investor sentiment turned bullish immediately following the news, driving NBIS shares up to 11.1% in pre-market trading on September 17, 2026.
π° The price hikes apply across the entire hardware stack, including a 25% jump for AMD EPYC Genoa CPUs and specific increases for H100 (17%) and H200 (20%) GPUs.
π Market analysts view the consecutive price increases as proof of Nebius's strong pricing power and fully booked capacity rather than operational struggles.
β οΈ The 41% memory price increase highlights a new binding constraint on AI infrastructure, as high-bandwidth memory supply remains tight alongside GPU silicon.
π Nebius maintains a competitive middle-ground pricing strategy, sitting below AWS list prices but above budget providers like RunPod and Lambda.
π The company has now raised prices twice in under six months (May and September 2026), indicating that wholesale chip allocation is moving faster than retail capacity growth.
π Newer GPU generations like the B300 face steeper hikes because they require proportionally more high-bandwidth memory, exacerbating supply constraints.
π‘οΈ Customers are advised to lock in reserved or multi-year contracts immediately to hedge against further volatility and potential future increases.
π The pricing asymmetry shows specialist neoclouds like Nebius repricing faster than diversified hyperscalers due to their higher reliance on GPU rental revenue.
- Shares surged 11.1% in pre-market trading as investors interpreted the price hikes as confirmation of a strong seller's market and robust demand for AI compute.
- The company successfully passed through supply constraints to customers, demonstrating significant pricing power without losing market share to competitors.
- Nebius maintains a competitive position against major hyperscalers like AWS and Google Cloud while retaining its status as a top-tier neocloud provider.
- The price increases specifically target newer, most constrained hardware (B300) and memory, indicating the company is maximizing revenue from scarce, high-value resources.
- Customers face recurring cost volatility with two price hikes in three months, creating budget uncertainty for startups and enterprises running AI workloads.
- The 41% increase in memory pricing introduces a new, unpredictable line item into AI infrastructure budgets that was previously bundled or stable.
- Repeated short-notice price adjustments (two weeks' notice) risk damaging customer relationships and incentivizing diversification away from Nebius.