CoreWeave wins if you need proven, large-scale H100 capacity with enterprise SLAs today. Nscale wins if you want next-generation GB300 GPUs, European data sovereignty, and the lowest energy costs in the market. Both are deeply partnered with Microsoft. CoreWeave is the established public company ($88B backlog, $5B+ revenue); Nscale is the fast-moving challenger with $4.5B raised and Goldman Sachs/JPMorgan underwriting its H2 2026 IPO. For investors: CoreWeave (CRWV) is liquid now; Nscale is the pre-IPO upside play.
By the Numbers
Head-to-Head Comparison
CoreWeave and Nscale are both full-stack AI hyperscalers — but they are at different stages, with different GPU generations, different geographies, and different cost profiles.
| Factor | CoreWeave (CRWV) | Nscale |
|---|---|---|
| IPO Status | ✅ Public (Nasdaq: CRWV) | ⏳ Pre-IPO (H2 2026 target) |
| Valuation / Market Cap | $35B market cap (Apr 2026) | $14.6B (Series C, Mar 2026) |
| GPU Generation | H100 / H200 (current gen) | ✅ GB300 / GB200 (next-gen, 4-5× faster) |
| GPU Count | ✅ 250,000+ (deployed) | 200,000 (deploying 2026) |
| Contract Backlog | ✅ $88B+ (unfilled, Q1 2026) | $23B+ (Microsoft alone) |
| Revenue | ✅ $5B+ annualized (public filings) | Not disclosed (early-stage ops) |
| Data Center Locations | 32 DCs, US-focused | ✅ Europe (Norway, UK) + US (Texas) |
| Energy Cost | Standard US/EU grid rates | ✅ 3¢/kWh (Norwegian hydro, 7× below avg) |
| Cost Advantage | ~30-40% below AWS/Azure | ✅ ~62% below comparable providers |
| Primary Clients | Microsoft, Meta, OpenAI, IBM | Microsoft, OpenAI (Stargate Norway), ByteDance |
| Founding Year | 2017 | 2024 |
| Total Funding | ✅ $12.1B raised (+ IPO proceeds) | $4.5B+ (Series A through C + debt) |
| Key Investors / Backers | Magnetar, Coatue, Public Markets | NVIDIA, Dell, Aker ASA, Citadel, Goldman, JPMorgan |
| Infrastructure Model | Owned data centers + leased colocation | ✅ Fully vertically-integrated (own land, power, DCs) |
| Data Sovereignty | US-focused | ✅ European sovereign AI (UK/Norway) |
| Maturity | ✅ 7 years, proven ops, public co. | 2 years, executing fast, board credibility |
Deep Dive: Infrastructure & Strategy
CoreWeave
CoreWeave is the established neocloud that proved the model works. Founded in 2017, it pivoted from crypto mining to GPU cloud, built 250,000+ NVIDIA GPUs across 32 data centers, signed $88 billion in contracts, and went public on Nasdaq in March 2025 at a $23B IPO valuation.
Its position as Microsoft's primary AI infrastructure partner — a relationship worth tens of billions — cemented its status as the hyperscaler-tier GPU cloud alternative to AWS and Azure. Meta and OpenAI are also major customers.
- 250,000+ H100/H200 GPUs across 32 data centers
- $88B+ contract backlog (Q1 2026)
- $5B+ annualized revenue, public filings
- Kubernetes-native orchestration, InfiniBand networking
- 32 US and European data center locations
- SLA-backed enterprise reliability — proven production ops
Nscale
Nscale is the challenger disrupting on three fronts: next-generation hardware, European sovereignty, and energy economics. Founded in 2024, it raised $4.5B+ in under 18 months and is deploying 200,000 NVIDIA GB300 GPUs — the generation after H100s, with 4-5x the performance per unit.
The Nscale thesis: if you build AI infrastructure on cheap Norwegian hydro (3¢/kWh vs. 21¢ European average), use the latest GPU generation, and vertically integrate everything from energy to orchestration, you can undercut established providers while delivering better compute.
- 200,000 NVIDIA GB300/GB200 GPUs — deploying through 2026
- Norwegian hydro power at 3¢/kWh (62% cost advantage claimed)
- $23B+ Microsoft contract (Texas + Norway deployments)
- Goldman Sachs + JPMorgan hired for H2 2026 IPO
- Sheryl Sandberg + Nick Clegg on board (governance credibility)
- Stargate Norway: 100,000 GPUs for OpenAI by end of 2026
Platform Profiles
CoreWeave — Platform Facts
Nscale — Platform Facts
Investment Comparison
Both companies have become the AI infrastructure trade of 2026 — one is already liquid, the other is the most anticipated GPU cloud IPO of the year.
CoreWeave (CRWV) — Bull Case
CoreWeave is the proven play. Public since March 2025, it has verifiable revenue ($5B+ annualized), a massive backlog ($88B+), and the Microsoft relationship that underpins its model. The question isn't whether AI infrastructure demand is real — it is. The question is whether CRWV is fairly valued relative to that demand.
- Liquid — anyone can buy CRWV today
- $88B backlog provides multi-year revenue visibility
- Microsoft partnership is structural, not transactional
- Only pure-play AI hyperscaler in public markets
- Scale advantage: 250K GPUs already running
Bear case: Significant debt load from GPU financing. Customer concentration risk (Microsoft ~65% of revenue). H100s will age relative to next-gen hardware. Competition from Nscale (GB300) and others on cost.
Nscale — Pre-IPO Bull Case
Nscale is the higher-risk, higher-upside bet. Two-year-old company, $14.6B valuation, and a $23B Microsoft contract as the anchor. Goldman Sachs and JPMorgan hired as underwriters. Sheryl Sandberg and Nick Clegg on the board. This is what a pre-IPO company looks like when it's seriously preparing to go public.
- GB300 fleet is next-gen — 4-5× the compute per unit vs H100
- Energy cost moat: 3¢/kWh Norwegian hydro (vs $0.10+ elsewhere)
- European sovereign AI is a structural demand driver (regulation)
- Vertically integrated: own the land, power, DCs, and software
- Goldman + JPMorgan hiring signals serious IPO preparation
Bear case: Two years old — massive execution risk on $4B+ buildout. Revenue concentration: Microsoft = most of revenue. Public markets may not support a $14B+ valuation for a company with limited operating history. IPO timing risk.
🏆 Bottom Line: Two Different Bets
For compute buyers: Use CoreWeave if you need proven H100 clusters today. Use Nscale if you want next-gen GB300 capacity and are willing to wait for Nscale's deployments to come online through 2026 — the performance and cost economics are compelling.
For investors: CoreWeave (CRWV) is the liquid, established play with $88B in backlog and public financials. Nscale is the pre-IPO bet on whether a two-year-old company can execute on $4B+ of infrastructure buildout and justify a $14B+ IPO valuation in H2 2026. Nscale's strategic investor list (NVIDIA, Goldman, Citadel) suggests institutional conviction, but the execution risk is real.
The big picture: Both companies are positioned for the same structural wave — AI inference and training demand is growing faster than any single company can supply. The GPU cloud market doesn't have a winner-take-all dynamic. CoreWeave and Nscale can both win.
Frequently Asked Questions
Track the Nscale IPO
Nscale is targeting an IPO in H2 2026. Get alerts when the S-1 is filed and when the IPO date is set.
View All AI IPOs 2026 → Nscale Company Profile