跳到主内容
@wquguru
精选75MarkTechPost(RSS)行业动态

2026年五大GPU Neocloud厂商算力与定价对比

Best GPU Neoclouds 2026: CoreWeave, Nebius, Lambda, Crusoe, and Groq Ranked by Published Pricing and Contracted Power

原文
发到 X

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.

“新云”(neocloud)标签目前涵盖五家商业模式截然不同的公司。CoreWeave 和 Nebius 是上市公司,按季度公布业绩并向美国证券交易委员会(SEC)提交报告(CoreWeave 提交 Form 10-Q,Nebius 作为外国私人发行人提交 Form 6-K)。Lambda 和 Crusoe 是非上市公司,正朝着首次公开募股(IPO)迈进。Groq 自 2024 年起在其 LPU 架构上运行 GroqCloud;在 2025 年 12 月将该技术授权给 NVIDIA 后,独立的 Groq 将重心完全转向推理云基础设施。本文从买家真正关心的角度对它们进行比较:已公布的定价、部署及签约的电力容量、硬件路线图、合同结构以及独立的质量评级。

TL;DR

TL;DR(太长不看版)

  • 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.
  • 在 SemiAnalysis ClusterMAX 2.0 中,CoreWeave 是唯一获得铂金评级的 GPU 云服务提供商,并在 H100、H200 和 B200 上享有溢价列表定价。
  • Nebius 是此处唯一公布 B300 按需定价的提供商;Lambda 公布了最低的 B200 按需费率。
  • Lambda 仍公布最便宜的 B200 按需价格($6.69/GPU-hr),但没有竞价实例层级。
  • Crusoe 是此处唯一在价目表上提供 AMD MI300X/MI355X 的提供商。
  • Groq 现为 NVIDIA 云合作伙伴,并计划在其 LPU 推理云旁增加 NVIDIA GPU 容量。

CoreWeave Q2 rev $2.575B, +112% YoYNebius AI cloud ARR $3.0BLambda B200 on-demand $6.69/hr, lowest publishedCrusoe 4.9 GW contracted, 40 GW+ pipelineGroq 13 DCs, 54 MW → 200+ MW in 2027CoreWeave first Vera Rubin NVL72 bring-upNebius $5.0B converts priced Aug 19Crusoe only AMD MI300X/MI355X on a rate cardLambda Michel Combes CEO, May 2026Groq $350M Series A at $3.5B, Aug 17CoreWeave Q2 rev $2.575B, +112% YoYNebius AI cloud ARR $3.0BLambda B200 on-demand $6.69/hr, lowest publishedCrusoe 4.9 GW contracted, 40 GW+ pipelineGroq 13 DCs, 54 MW → 200+ MW in 2027CoreWeave first Vera Rubin NVL72 bring-upNebius $5.0B converts priced Aug 19Crusoe only AMD MI300X/MI355X on a rate cardLambda Michel Combes CEO, May 2026Groq $350M Series A at $3.5B, Aug 17

CoreWeave Q2 营收 25.75 亿美元,同比增长 112%Nebius AI 云年度经常性收入(ARR)30 亿美元Lambda B200 按需 $6.69/小时,为最低公布价格Crusoe 签约 4.9 GW,管道项目超 40 GWWroq 13 个数据中心,54 MW → 2027 年达 200+ MWCoresweave 首个 Vera Rubin NVL72 上线Nebius 50 亿美元可转债于 8 月 19 日定价Crusoe 价目表上仅有 AMD MI300X/MI355XLambda Michel Combes 任 CEO,2026 年 5 月Groq 8 月 17 日以 35 亿美元估值完成 3.5 亿美元 A 轮融资CoreWeave Q2 营收 25.75 亿美元,同比增长 112%Nebius AI 云年度经常性收入(ARR)30 亿美元Lambda B200 按需 $6.69/小时,为最低公布价格Crusoe 签约 4.9 GW,管道项目超 40 GWWroq 13 个数据中心,54 MW → 2027 年达 200+ MWCoresweave 首个 Vera Rubin NVL72 上线Nebius 50 亿美元可转债于 8 月 19 日定价Crusoe 价目表上仅有 AMD MI300X/MI355XLambda Michel Combes 任 CEO,2026 年 5 月Groq 8 月 17 日以 35 亿美元估值完成 3.5 亿美元 A 轮融资

Marktechpost · GPU neocloud ledger · 2026

Marktechpost · GPU 新云账本 · 2026

CoreWeave / Nebius / Lambda / Crusoe / Groq

CoreWeave / Nebius / Lambda / Crusoe / Groq

Figures checkedAugust 21, 2026Primary sources where available

数据核对日期:2026 年 8 月 21 日主要来源:尽可能采用原始出处

MetricCoreWeaveNASDAQ: CRWVNebiusNASDAQ: NBISLambdaPrivate · IPO H2’26 (rep.)CrusoePrivate · ~$10B+GroqPrivate · $3.5B
Published priceNVIDIA HGX H100 · $/GPU-hr$6.16node price ÷ GPUs$3.85HGX, on-demand$3.99SXM, 8x instance$3.90HGX, on-demand—per-token only
Q2 2026 revenuelatest reported$2.58B+112% YoY · net loss $626M$582M+454% YoY · ARR $3.0Bn/d>$520M FY Sep-25 (The Information)n/dprivate; ~$3B raise at ~$30B reported (Jul)n/d$1B raised in 2026 (Jun + Aug)
Poweractive / contracted1.5 GW4.2 GW+ contracted · 51 DCs~1 GWconnected target YE26 · 5 GW contracted targetn/d15 US DCs · KC site 24→100+ MW~0.2 GWAbilene energized · 4.9 GW contracted (Jun)54 MW→ 200+ MW in 2027 · 13 DCs
ClusterMAX 2.0SemiAnalysis tierPlatinumOnly Platinum provider, 2 ratings runningGoldTop of Gold tierSilverNo spot tier; on-demand self-serveGoldOnly AMD on the rate cardNot ratedInference API at time of rating
Anchor dealsdisclosedMeta, Anthropic, OpenAI~$104B backlog (+>$25B early Q3)Microsoft $17.4B (to $19.4B)Meta $3B + up to ~$27BMicrosoft multibillion, GB300 NVL72NVIDIA ~$1.5B lease-back (rep.)Oracle/OpenAI Abilene 1.2 GWMicrosoft 900 MW campusNVIDIA licensed LPU tech (~$20B rep.)6M+ developers on GroqCloud
Next-gen siliconstatus Aug 2026Vera Rubin NVL72: first bring-up + validation (Q2)Vera Rubin NVL72 units received; validatingVera CPU + Quantum-X800 CPO launch partnerVera CPU launch support; MI355X via salesNVIDIA LPX (Groq tech); NCP certified Aug 12; GPU capacity planned
Discounts & egresspublishedReserved up to 60% off · egress freeCommit up to 35% off · network egress free (Std object egress $0.015/GiB)Sales-quoted · egress free · no spotSales-quoted · egress freePer-token pricing · clusters sales-led
Funding posture2026$35–39B capex guide · $31B recourse debt · $3.1B term loan · $1B Jane Street$4.3B converts (H1) · $5.0B converts priced Aug 19 · $775M secured · >$9B prepayments expected$1.5B+ Series E (Nov) · $1B credit facility (May) · $350M pre-IPO (rep.)$1.375B Series E at >$10B · $7.1B JV construction loan (Abilene ph.2)$650M (Jun) + $350M Series A at $3.5B (Aug), NVIDIA participating
指标CoreWeave(纳斯达克:CRWV)Nebius(纳斯达克:NBIS)Lambda(私有·IPO 2026年下半年(代表))Crusoe(私有·约100亿美元以上)Groq(私有·35亿美元)
公布价格 NVIDIA HGX H100 · 美元/GPU-小时节点价格 ÷ GPU数量 = 6.16美元3.85美元HGX,按需3.99美元SXM,8卡实例3.90美元HGX,按需—仅按token计费
2026年第二季度收入 最新报告25.8亿美元+ 同比增长112% · 净亏损6.26亿美元5.82亿美元+ 同比增长454% · ARR 30亿美元n/d > 5.2亿美元 FY Sep-25(The Information)n/d 私有;约30亿美元估值融资30亿美元(7月)n/d 2026年(6月和8月)融资10亿美元
电力 活跃/已签约1.5 GW 4.2 GW+ 已签约 · 51个数据中心~1 GW 连接目标至2026年底 · 5 GW 签约目标n/d 15个美国数据中心 · KC站点从24MW增至100+ MW~0.2 GW Abilene已通电 · 4.9 GW 已签约(6月)54 MW→ 2027年增至200+ MW · 13个数据中心
集群 MAX 2.0 SemiAnalysis评级铂金 唯一铂金提供商,2项评级运行中黄金 黄金级最高白银 无竞价层级;按需自助服务黄金 费率表上仅有AMD未评级 评级时的推理API
锚定交易 披露Meta、Anthropic、OpenAI 约1040亿美元积压订单(+>250亿美元 第三季度初)微软 174亿美元(至194亿美元)Meta 30亿美元 + 高达~270亿美元微软数十亿,GB300 NVL72 NVIDIA ~15亿美元售后回租(代表)Oracle/OpenAI Abilene 1.2 GW 微软900 MW园区NVIDIA授权LPU技术(~200亿美元代表)600万+开发者在GroqCloud
下一代硅片状态 2026年8月Vera Rubin NVL72:首次启动+验证(第二季度)Vera Rubin NVL72单元已接收;正在验证Vera CPU + Quantum-X800 CPO发布合作伙伴Vera CPU发布支持;MI355X通过销售NVIDIA LPX(Groq技术);NCP认证8月12日;计划GPU容量
折扣与出口流量 公布预留最多节省60% · 出口流量免费承诺最多节省35% · 网络出口流量免费(标准对象出口流量0.015美元/GiB)销售报价 · 出口流量免费 · 无竞价销售报价 · 出口流量免费按token定价 · 集群由销售主导
资金状况 2026年350–390亿美元资本支出指引 · 310亿美元追索债务 · 31亿美元定期贷款 · 10亿美元Jane Street43亿美元可转债(上半年)· 50亿美元可转债定价于8月19日 · 7.75亿美元担保 · 预计预付款超过90亿美元15亿美元+E轮(11月)· 10亿美元信贷额度(5月)· 3.5亿美元IPO前(代表)13.75亿美元E轮估值超100亿美元 · 71亿美元合资建设贷款(Abilene二期)6.5亿美元(6月)+ 3.5亿美元A轮估值35亿美元(8月),NVIDIA参与

Cheapest published

最便宜的公布价格

$3.85

$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.

Nebius的H100按需价格。Lambda拥有最低的B200价格为6.69美元;Crusoe拥有最低的H200价格为4.29美元。Nebius是唯一列出B300按需价格的供应商。

Premium, and why

溢价及其原因

+10–15%

+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.

SemiAnalysis 指出,在托管集群上,CoreWeave 的每 GPU 小时定价高于 Nebius、Crusoe 和 Lambda。它是唯一两次获得 Platinum ClusterMAX 认证的提供商。

What capacity really clears at

实际出清产能位于

$40M+

$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.

Nebius Q3 短期交易的每兆瓦(MW)价格,而其 2026 年基础价格为 ~$12M。四个 Q2 交易平均每个超过 $1B;其中 70% 为预付款。

Marktechpost · AI Dev Signals

Marktechpost · AI Dev Signals

Sources: CoreWeave Q2’26 release + pricing page · Nebius Q2’26 results, shareholder letter, pricing page · Lambda pricing page · Crusoe pricing page, Jun 9 ’26 release · Groq newsroom Dec ’25–Aug ’26 · SemiAnalysis ClusterMAX 2.0

来源:CoreWeave Q2’26 发布稿及定价页面 · Nebius Q2’26 财报、致股东信及定价页面 · Lambda 定价页面 · Crusoe 定价页面、Jun 9 ’26 发布稿 · Groq 新闻室 Dec ’25–Aug ’26 · SemiAnalysis ClusterMAX 2.0

1. Who these companies are in 2026

1. 这些公司在 2026 年的定位

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. (CoreWeave Q2 2026 release, CoreWeave AI data centers page)

CoreWeave(纳斯达克代码:CRWV)报告 2026 年第二季度营收为 25.75 亿美元,同比增长 112%,积压订单约为 1040 亿美元。该积压订单不包括 Q3 初新增的超过 250 亿美元的新承诺。截至 6 月 30 日,活跃功率达到 1.5 GW,签约功率达到约 3.7 GW;CoreWeave 的数据中心页面现在显示其在 51 个数据中心拥有 4.2 GW+ 的签约功率。该公司在本季度完成了 NVIDIA Vera Rubin NVL72 的行业首次上线和验证。净亏损为 6.26 亿美元,主要由 6.4 亿美元的净利息支出驱动。(CoreWeave Q2 2026 发布稿,CoreWeave AI 数据中心页面)

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. (CNBC)

2026 年全年指引为营收 124–132 亿美元,资本支出 350–390 亿美元。管理层预计年底前活跃功率将超过 1.85 GW。(CNBC)

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.

Nebius Group(纳斯达克代码:NBIS)报告 2026 年第二季度集团营收为 5.823 亿美元,同比增长 454%。AI 云营收为 5.749 亿美元,同比增长 514%,年化运行收入为 30 亿美元。AI 云板块调整后 EBITDA 利润率为 49.7%。Q2 资本支出约为 57 亿美元,公司季度末持有现金 80 亿美元。Nebius 将其 2026 年底签约功率目标提高至 5 GW,并计划从 2027 年起每年部署超过 1 GW 的功率。

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 (J.P. Morgan lead arranger) 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 Information has reported an IPO target of H2 2026, but no public S-1 exists as of this writing.

Lambda 为非上市公司。它于 2025 年 11 月完成了由 TWG Global 领投的超过 15 亿美元的 E 轮融资,并于 2026 年 5 月完成了 10 亿美元的优先担保信贷额度(J.P. Morgan 担任牵头安排人)。SoftBank International 和 Sprint 前首席执行官 Michel Combes 于 2026 年 5 月出任 CEO;联合创始人 Stephen Balaban 转任 CTO。Lambda 与微软签署了一份价值数十亿美元、为期多年的协议,涵盖数万个 NVIDIA GPU,包括 GB300 NVL72 系统。《The Information》曾报道其 IPO 目标时间为 2026 年下半年,但截至目前尚未公开 S-1 文件。

Crusoe is private. It raised a $1.375 billion Series E at a valuation above $10 billion in October 2025, led by Valor Equity Partners and Mubadala Capital. In July 2026, Bloomberg reported 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.

Crusoe 是一家私营公司。2025年10月,在Valor Equity Partners和Mubadala Capital的领投下,它以超过1000亿美元的估值完成了13.75亿美元的E轮融资。2026年7月,彭博社报道Crusoe正在洽谈以约300亿美元的估值融资约30亿美元。Crusoe为Oracle和OpenAI建设位于德克萨斯州阿比林的1.2GW园区(这是Stargate项目的第一个站点)。2026年3月,它宣布为微软建设一个独立的900MW阿比林园区,使整个站点的总容量达到约2.1GW。2026年6月9日,Crusoe表示其签约的人工智能基础设施在美国五个园区及Crusoe Cloud上总计达到4.9GW,另有超过40GW的开发管线。

更进一步:量化金融体系

看懂新闻只是起点——沿量化金融路径,把它变成能交付的工程能力

进入量化体系 →

相似阅读

另一事件,读法相近