**2026 Neocloud GPU Price and Capacity Guide: CoreWeave, Nebius, Lambda, Crusoe, Groq**
The “neocloud” label now covers five companies with very different business models. CoreWeave and Nebius are public companies that publish quarterly results and file reports with the SEC (CoreWeave on Form 10-Q, Nebius as a foreign private issuer on Form 6-K). Lambda and Crusoe are private and heading toward IPOs. Groq has run GroqCloud on its LPU architecture since 2024; after licensing that technology to NVIDIA in December 2025, the independent Groq refocused entirely on inference-cloud infrastructure. This article compares them on what actually matters to a buyer: published pricing, deployed and contracted power, hardware roadmap, contract structure, and independent quality ratings.
**TL;DR**
– CoreWeave is the only Platinum-rated GPU cloud in SemiAnalysis ClusterMAX 2.0 and commands premium list pricing on H100, H200, and B200.
– Nebius is the only provider here publishing B300 on-demand pricing; Lambda publishes the lowest B200 on-demand rate.
– Lambda still publishes the cheapest on-demand B200 ($6.69/GPU-hr) but has no spot tier.
– Crusoe is the only provider here with AMD MI300X/MI355X on its rate card.
– Groq is now an NVIDIA Cloud Partner and plans to add NVIDIA GPU capacity alongside its LPU inference cloud.
**Quick Reference Table**
| Metric | CoreWeave (NASDAQ: CRWV) | Nebius (NASDAQ: NBIS) | Lambda (Private · IPO H2’26) | Crusoe (Private · ~$10B+) | Groq (Private · $3.5B) |
|——–|————————–|———————–|——————————-|—————————|————————|
| **Published price NVIDIA HGX H100 · $/GPU-hr** | $6.16 (spot $2.46) | **$3.85** (preemptible $2.15) | $3.99 | $3.90 | — |
| **Q2 2026 revenue** | $2.58B (+112% YoY, net loss $626M) | $582M (+454% YoY, ARR $3.0B) | n/d (> $520M FY Sep-25) | n/d (private; ~$3B raise at ~$30B reported) | n/d ($650M raised in 2026) |
| **Power (active / contracted)** | 1.5 GW / 4.2 GW+ | ~1 GW / 5 GW target YE26 | n/d / 15 US DCs, KC site 24→100+ MW | ~0.2 GW / 4.9 GW (Jun) | 54 MW → 200+ MW in 2027 |
| **ClusterMAX 2.0 (SemiAnalysis tier)** | Platinum (only Platinum) | Gold | Silver (no spot) | Gold | Not rated |
| **Next-gen silicon status Aug 2026** | Vera Rubin NVL72: first bring-up + validation (Q2) | Vera Rubin NVL72 units received; validating | Vera CPU + Quantum-X800 CPO launch partner | Vera CPU launch support; MI355X via sales | NVIDIA LPX (Groq tech); NCP certified Aug 12; GPU capacity planned |
**Cheapest published**
– $3.85: H100 on-demand at Nebius.
– Lambda has the lowest B200 at $6.69; Crusoe the lowest H200 at $4.29.
– Nebius is the only one listing B300 on-demand.
**Premium, and why**
– +10–15%: SemiAnalysis says CoreWeave commands that per GPU-hour over Nebius, Crusoe and Lambda on managed clusters. It is the only Platinum ClusterMAX provider, twice.
**What capacity really clears at**
– $40M+ per MW on Q3 short-term deals at Nebius, versus ~$12M on its 2026 base. Four Q2 deals averaged > $1B each; 70% were prepaid.
—
### 1. Who These Companies Are in 2026
**CoreWeave (Nasdaq: CRWV)** reported Q2 2026 revenue of $2.575 billion, up 112% year over year, with a revenue backlog of roughly $104 billion. That backlog excludes more than $25 billion of new commitments added in early Q3. Active power reached 1.5 GW and contracted power reached about 3.7 GW as of June 30; CoreWeave’s data center page now states 4.2 GW+ contracted across 51 data centers. The company completed the industry’s first bring-up and validation of NVIDIA Vera Rubin NVL72 during the quarter. Net loss was $626 million, driven by $640 million in net interest expense. Full-year 2026 guidance is $12.4–13.2 billion in revenue and $35–39 billion in capex. Management expects more than 1.85 GW of active power by year end.
**Nebius Group (Nasdaq: NBIS)** reported Q2 2026 group revenue of $582.3 million, up 454% year over year. AI cloud revenue was $574.9 million, up 514%, with annualized run-rate revenue of $3.0 billion. The AI cloud segment posted a 49.7% adjusted EBITDA margin. Q2 capex was about $5.7 billion, and the company ended the quarter with $8.0 billion in cash. Nebius raised its year-end 2026 contracted power target to 5 GW and plans to deploy more than 1 GW per year starting in 2027.
**Lambda** is private. It closed a Series E of more than $1.5 billion led by TWG Global in November 2025 and a $1 billion senior secured credit facility in May 2026. Michel Combes, former CEO of SoftBank International and Sprint, became CEO in May 2026; co-founder Stephen Balaban moved to CTO. Lambda has a multibillion-dollar, multi-year agreement with Microsoft covering tens of thousands of NVIDIA GPUs, including GB300 NVL72 systems. The IPO target is H2 2026.
**Crusoe** is private. It raised a $1.375 billion Series E at a valuation above $10 billion in October 2025. In July 2026, it was reported that Crusoe was in talks to raise roughly $3 billion at a valuation near $30 billion. Crusoe builds the 1.2 GW Abilene, Texas campus for Oracle and OpenAI (the first Stargate site). In March 2026 it announced a separate 900 MW Abilene campus for Microsoft, bringing the full site to about 2.1 GW. On June 9, 2026, Crusoe said its contracted AI infrastructure reached 4.9 GW across five US campuses and Crusoe Cloud, with a development pipeline above 40 GW.
**Groq** launched GroqCloud in March 2024 around its LPU inference chip. It changed shape in December 2025, when NVIDIA licensed Groq’s inference technology in a transaction reported at about $20 billion. Groq stayed independent, and GroqCloud kept running. Since then, Groq raised $650 million in June 2026 and a $350 million Series A in August 2026 at a $3.5 billion valuation, with planned NVIDIA participation. Groq now operates 13 data centers and expects to scale from 54 MW to 200+ MW in 2027. On August 12, 2026 it joined the NVIDIA Cloud Partner program and said it plans to bring NVIDIA accelerated computing online “in the future.”
—
### 2. Published On-Demand Pricing (Per GPU-Hour, USD)
Prices below were read directly from each provider’s pricing page on August 21, 2026. CoreWeave lists per-node prices; per-GPU values are the node price divided by GPU count. CoreWeave’s page also shows the identical per-GPU figure as its “inference single GPU” price. Groq does not publish GPU-hour pricing; GroqCloud is priced per token, and NVIDIA GPU cluster access is a planned offering handled by sales.
| GPU | CoreWeave | Nebius | Lambda | Crusoe |
|—–|———–|——–|——–|——–|
| NVIDIA GB300 NVL72 | Contact sales | Contact sales | — | — |
| NVIDIA HGX B300 | Contact sales (spot $4.48) | **$7.85** (preemptible $4.30) | — | — |
| NVIDIA GB200 NVL72 | $10.50 (4-GPU node $42.00) | Contact sales | — | Contact sales |
| NVIDIA HGX B200 | $8.60 (8-GPU node $68.80; spot $4.26) | $7.15 (preemptible $3.95) | **$6.69** | Contact sales |
| NVIDIA HGX H200 | $6.31 (spot $2.62) | $4.50 (preemptible $2.45) | — | **$4.29** |
| NVIDIA HGX H100 | $6.16 (spot $2.46) | **$3.85** (preemptible $2.15) | $3.99 | $3.90 |
| NVIDIA A100 80GB SXM | $2.70 | — | $2.79 | $2.30 |
| AMD MI355X | — | — | — | Contact sales |
| AMD MI300X | — | — | — | $3.45 |
| Network egress | Free | Free (Standard object egress $0.015/GiB) | Free | Free |
| Committed-use discount | Up to 60% | Up to 35% | Sales-quoted | Sales-quoted |
**What the table shows**
– **CoreWeave charges a premium and publishes the most complete card.** Its H100 rate of $6.16/GPU-hr is about 60% above Nebius and Lambda. SemiAnalysis reports CoreWeave commands roughly 10–15% more per GPU-hour than Nebius, Crusoe, Lambda, and Fluidstack on managed Slurm or Kubernetes clusters, and that its pricing sits closer to the big hyperscalers. CoreWeave is also the only one of the four publishing a GB200 NVL72 on-demand price.
– **Nebius undercuts CoreWeave on Blackwell and is the only one with B300 on-demand.** Its B200 rate of $7.15 is below CoreWeave’s $8.60, and its $7.85 B300 rate has no published competitor in this group. Preemptible rates run roughly 45% below on-demand.
– **Lambda holds the cheapest published B200 instance at $6.69.** Its 1-Click Clusters (16 to 2,000+ GPUs, 2-week to 1-year terms) are priced higher: $9.86, $9.36, and $8.87 per B200-hour at 16, 64, and 256+ GPUs respectively. Lambda has no spot tier. Third-party trackers note Lambda raised its H100 SXM rate from $2.99 to $3.99 over 2025–2026 as demand outran supply.
– **Crusoe is the only AMD option and the cheapest H200 at $4.29.** Its H100 rate of $3.90 is close to Nebius and Lambda. B200, GB200, and MI300X require a sales conversation.
– **List prices understate the market.** Nebius disclosed that its Q2 2026 deals averaged more than $20 million in annual contract value per MW, that the 2026 base is priced near $12 million per MW, and that Q3 short-term capacity deals are clearing above $40 million per MW. The company also ran its first capacity auction.
—
### 3. Deployed and Contracted Power
| Provider | Active / connected power | Contracted or planned | Notable sites |
|———-|————————–|———————-|—————|
| CoreWeave | 1.5 GW active (Q2 2026); >1.85 GW expected by YE 2026 | ~3.7 GW contracted at June 30; 4.2 GW+ per current company page | 51 data centers, North America and Europe |
| Nebius | Up to ~1 GW connected targeted by YE 2026 | 5 GW contracted by YE 2026; >1 GW/yr from 2027 | New Jersey, Missouri, Oklahoma, Alabama, Kansas City, UK, Finland, France, Israel, Estonia, Iceland |
| Lambda | Not disclosed; 15 US data centers per company | Microsoft deal (tens of thousands of GPUs); Kansas City site 24 MW → 100+ MW | Liquid-cooled US sites; 10,000+ Blackwell Ultra factory with Quantum-X800 CPO |
| Crusoe | Abilene phase 1: two ~100 MW buildings energized | 4.9 GW contracted across data centers and cloud (June 2026); Abilene 1.2 GW (phase 2 by YE 2026) + 900 MW Microsoft campus (first building mid-2027); 1.8 GW Wyoming campus | Five US campuses incl. Abilene TX, two more in Texas, Missouri; Wyoming; Iceland, Norway |
| Groq | 54 MW | 200+ MW in 2027 | 13 data centers across North America, Europe, Middle East, APAC |
—
### 4. Hardware Roadmap and Access to Next-Gen Silicon
**Vera Rubin**: CoreWeave says it completed the industry’s first bring-up and system-level validation of Vera Rubin NVL72 in Q2 2026. Nebius says it has received its first Vera Rubin NVL72 systems and is validating compute, networking, and orchestration ahead of production availability. Lambda was a GTC 2026 launch partner for NVIDIA’s Vera CPU and Quantum-X800 InfiniBand co-packaged optics. Crusoe announced launch support for the Vera CPU and NVIDIA Nemotron 3 models. None of the four publishes Rubin pricing yet.
**Blackwell Ultra (B300 / GB300)**: Nebius is the only provider with a public B300 on-demand price. CoreWeave lists GB300 NVL72 and HGX B300 as contact-sales, with a B300 spot rate of $35.84 per 8-GPU node. Lambda’s Microsoft contract includes GB300 NVL72 systems. Nebius’s Microsoft contract was reported to cover more than 100,000 GB300 chips.
**AMD**: Crusoe is the only one of the five with AMD Instinct MI300X on its on-demand card and MI355X available via sales. CoreWeave, Nebius, and Lambda remain NVIDIA-only on their public pages. Groq’s LPU is the other non-NVIDIA accelerator in this group, but it is sold per token, not per GPU-hour.
**Groq LPU and NVIDIA LPX**: NVIDIA announced its LPX platform, incorporating licensed Groq inference technology, at GTC 2026. Groq says its capital will fund fit-out of existing data centers with LPX systems and NVIDIA GPU clusters. Groq’s own LPU remains the engine behind GroqCloud’s token API.
—
### 5. Contract Structure and Who the Anchor Customers Are
| Provider | Anchor customers (disclosed) | Contract model |
|———-|——————————|—————-|
| CoreWeave | Meta, Anthropic, OpenAI, Microsoft (historically), plus Cognition, Databricks, Hudson River Trading, Runway, Jane Street | Multi-year take-or-pay; ~$104B backlog; reserved up to 60% off; on-demand and spot published |
| Nebius | Microsoft ($17.4B base, up to $19.4B); Meta ($3B Nov 2025; second agreement up to ~$27B, Mar 2026); Reflection, Cohere, a US quant firm | 3 tiers: short-term 3–6 month premium deals, 1–3 year core contracts, long-term investment-grade contracts; ~70% of Q2 deals prepaid |
| Lambda | Microsoft (multibillion), NVIDIA (reported ~$1.5B lease-back of ~18,000 GPUs) | On-demand self-serve; 1-Click Clusters 2 weeks–1 year; reserved via sales |
| Crusoe | Oracle/OpenAI (Abilene lease), Microsoft (900 MW campus) | On-demand, spot, reserved; plus long-term data center leases as a developer |
| Groq | 6M+ developers, Fortune 500 enterprises | Per-token serverless inference; NVIDIA GPU clusters planned, sales-led |
The Nebius–Meta structure is worth reading closely. The March 2026 agreement includes $12 billion of dedicated 5-year cluster orders starting early 2027. A further order gives Meta access to unsold capacity on specified clusters, and obliges Meta to buy that capacity if Nebius cannot sell it to others. That is effectively a floor on utilization for the covered clusters. (SEC 6-K)
—
### 6. Which Provider Fits Which Workload
– **Frontier-scale training (10,000+ GPUs, multi-year)**: CoreWeave and Nebius are the two public providers here with SEC-reported financial scale and demonstrated NVL72 rack-scale operating experience. CoreWeave has the reliability track record and Platinum rating. Nebius has better committed pricing and is selling 2027 capacity now. Both have Vera Rubin units in hand.
– **Mid-size training and fine-tuning (16–512 GPUs, weeks to months)**: Lambda’s 1-Click Clusters and Nebius’s on-demand B200/B300 are the most accessible. Crusoe’s H200 at $4.29 is the cheapest Hopper-class option for memory-heavy jobs.
– **Budget-sensitive experiments**: Nebius preemptible H100 at $2.15 and CoreWeave spot H100 at $2.46 per GPU are the lowest published rates here. Lambda has no spot tier.
– **AMD GPU evaluation**: Crusoe is the only choice in this group.
– **Low-latency inference at high token volume**: GroqCloud’s per-token LPU endpoints remain the distinctive offering. Nebius Token Factory and Crusoe Managed Inference are the GPU-based alternatives with published per-token rates.
– **Buyers needing a public balance sheet**: Only CoreWeave and Nebius publish quarterly results and file with the SEC. Both companies’ annual statements are audited; the Q2 figures used here are unaudited quarterly results.
—
### Sources
1. CoreWeave, “CoreWeave Reports Strong Second Quarter 2026 Results,” Aug 11, 2026 — CoreWeave Q2’26 release and pricing page
2. CNBC, “CoreWeave (CRWV) Q2 earnings report 2026,” Aug 11, 2026 — CNBC Q2’26 transcript
3. Nebius, “Nebius reports second quarter 2026 financial results” (PDF), Aug 12, 2026 — Nebius 6-K and investor deck
4. Nebius, “Letter to shareholders Q2 2026” (PDF), Aug 12, 2026 — Nebius shareholder letter
5. Nebius Group Form 6-K (Meta agreement), Mar 2026 — SEC filing
6. Nebius Group Form 6-K (secured financing), Jul 2026 — SEC filing
7. Lambda pricing page — company pricing page
8. Lambda, “$1 Billion Senior Secured Credit Facility,” May 7, 2026 — facility announcement
9. Lambda, “Leadership Team” announcement, May 2026 — leadership change
10. Lambda, Microsoft agreement, Nov 3, 2025 — partnership disclosure
11. Crusoe pricing page — company pricing page
12. Crusoe, Series E announcement — funding round release
13. Crusoe, 900 MW Microsoft campus, Mar 27, 2026 — project announcement
14. Groq, NVIDIA licensing agreement, Dec 2025 — licensing news
15. Groq, $650M raise, Jun 22, 2026 — raise announcement
16. Groq, $350M Series A, Aug 17, 2026 — Series A close
17. CNBC, NVIDIA–Groq deal, Dec 24, 2025 — deal coverage
18. SemiAnalysis, “ClusterMAX 2.0,” Nov 2025 — platform evaluation
19. SemiAnalysis ClusterMAX CoreWeave review — analyst review
20. DCD, Nebius–Meta and capacity targets — capacity reporting
21. DCD, Lambda pre-IPO funding report — pre-IPO round
22. TechCrunch, Groq $350M pivot — funding round analysis
23. SiliconANGLE, Crusoe $3B talks — raise coverage
24. CoreWeave, AI data centers page (51 DCs, 4.2 GW+ contracted) — infrastructure overview
25. Crusoe, “Contracted AI Infrastructure Capacity Approaches 5 Gigawatts,” Jun 9, 2026 — capacity metrics
26. Groq, “Groq Becomes an NVIDIA Cloud Partner,” Aug 12, 2026 — partnership announcement
27. Nebius Group Form 6-K, pricing of $5.0B convertible notes, Aug 20, 2026 — convertible notes pricing
28. Newmark, $7.1B construction loan for Abilene JV — project finance
29. Warren, Wyden, Blumenthal letter to DOJ/FTC, Feb 4, 2026 — policy inquiry
—
Thank you for reading
