GPU Clouds and Neoclouds: CoreWeave, Lambda, Nebius — 2026-09-19
CoreWeave and Nebius both announced aggressive price increases this week, signaling sustained AI demand and tightening GPU supply. CoreWeave is contracting new capacity at higher rates with a $25+ billion backlog, while Nebius is raising prices up to 21% effective October 1. The sector also saw a major IPO filing from Nscale, a London-based AI data center operator backed by Nvidia.
GPU Clouds and Neoclouds: CoreWeave, Lambda, Nebius — 2026-09-19
Top developments
CoreWeave Locks in Higher-Priced Contracts; $25 Billion Backlog Drives Revenue
CoreWeave announced on September 16 that it has continued to contract new compute capacity at materially higher prices since June 30, 2026. The company has secured approximately 4.2 gigawatts of contracted power capacity, with net new customer commitments exceeding $25 billion. The latest contracts are priced at roughly $40 million per megawatt, a significant increase from earlier 2026 pricing. This reflects CoreWeave's pricing power in a market where hyperscalers and AI labs are competing aggressively for constrained GPU supply.
Nebius Raises GPU Rental Prices by Up to 21% Effective October 1
Nebius announced on September 17 that it will raise prices on its GPU cloud services by an average of approximately 20% starting October 1, 2026. The increases span NVIDIA H100, H200, B200, and B300 GPUs, with specific hikes of 17% for H100 (from $3.85/hr to $4.50/hr), 20% for H200, 19% for B200, and 21% for B300 (from $7.85 to $9.50/hr). The move, coupled with strong AI demand signals, drove Nebius stock up nearly 9% in premarket trading.
Nscale Files for U.S. IPO; Third Major Neocloud to Pursue Public Markets
Nscale, a London-based AI data center operator spun off from a cryptocurrency mining business in early 2024 and backed by Nvidia, filed for a U.S. IPO on September 18, seeking up to $3 billion. The filing underscores investor appetite for pure-play GPU rental infrastructure amid the AI boom and follows CoreWeave and Nebius to public markets.
H100 Availability Tightens; Spot Premiums Narrow
According to the Cloud GPU Rental Price Index updated September 18, H100 availability stands at 33% of listings, indicating persistent scarcity. While H100 on-demand prices have compressed to roughly $3.38–$6.16/hr across neocloud providers (down from peaks earlier in 2026), newer GPUs command steep premiums: B200 rentals run $3.99–$5.50/hr on dedicated clouds but spike to ~$9.36/hr on AWS capacity blocks. The tightness in B200, H200, and MI300X supply (at 16%, 18%, and 7% availability respectively) is driving enterprises toward mature H100 clusters despite efforts to diversify.
Local view
Chinese financial media (新浪财经, 搜狐, 17173) emphasized that Nebius's pricing power—even on legacy H100 hardware in its fourth year on market—signals a structural shift in bargaining power toward GPU supply operators. Analysts noted that the price increases coincide with Intel CEO Pat Gelsinger's disclosure that CPU supply can only meet 50% of customer demand, reinforcing sector-wide scarcity narratives that justify both neocloud and hyperscaler price hikes.
Context & numbers
- CoreWeave contracted backlog: ~$99.4–$104 billion (as of Q1–Q3 2026)
- CoreWeave debt load: ~$25 billion at
8% blended cost ($2 billion annual interest) - Nebius Q1 2026 revenue: Tracking at elevated growth; pricing increases effective Oct 1 likely to expand margins
- H100 spot/committed pricing: Down to $1.80–$2.40/hr with 1-year commitments; on-demand $3.38–$6.16/hr
- B200/H200 availability: 16% and 18% of provider listings; commanding 40–60% premiums over H100
- MI300X availability: 7% of listings; most scarce accelerator in the market
- Nscale IPO target: $3 billion; backs trend of neocloud capital raises via public markets
On the radar
- Nebius pricing effective date: October 1, 2026 — watch for customer pushback and competitive response from Lambda, Together, and others
- CoreWeave power ramp: Company targeting 1.7 GW of active power by end-2026; track quarterly updates on deployment velocity and utilization rates
- Saturn Cloud + NVIDIA Run:ai integration (announced Sept 18): Pairing GPU orchestration with inference-focused monetization; may signal neocloud focus shifting from training toward higher-margin inference products
- CFO budgeting cycles: PYMNTS reported (Sept 12) that AI capacity contracts are becoming embedded in corporate financing — watch for Q4 2026 enterprise capex commitments as sign of sustained demand or pullback
Freshness note: All sources dated Sept 12–19, 2026. Pricing snapshots reflect on-demand market rates and are subject to rapid change; reserved and spot rates vary by term and utilization tier.
This content was collected, curated, and summarized entirely by AI — including how and what to gather. It may contain inaccuracies. Crew does not guarantee the accuracy of any information presented here. Always verify facts on your own before acting on them. Crew assumes no legal liability for any consequences arising from reliance on this content.
Powered by
