CoreWeave面临百亿短债重压 新兴算力云现流动性分水岭

【The Patient Investor】单季营收涨182% IREN仍亏1.86亿 — 浑水摸鱼清源

浑水摸鱼清源

CoreWeave在手现金对积压订单覆盖率仅5.1%且短期偿债缺口巨大,再融资风险显著高于手握实物资产的IREN与高预付款护航的Nebius。

AI-assisted summary of 【The Patient Investor】单季营收涨182% IREN仍亏1.86亿 — original source. Check the original for full context.

Source published:

Key ideas

  • 三大新兴云厂商年内走势分化严重:IREN大跌57%、CoreWeave跌33%,仅Nebius逆势上涨28%,但市场核心抛压聚焦于盈利与资产负债表健康度而非算力需求端。 (0:00)
  • CoreWeave面临严峻的流动性考验,其在手现金仅约41亿美元,却背负550亿美元总债务,且一年内到期的长期债务流动部分即高达105亿美元。 (5:07)
  • 现金对在手订单覆盖率指标出现显著分层:IREN以28.5%位列第一,Nebius以26.5%紧随其后,而CoreWeave仅为5.1%,股权稀释与再融资风险最高。 (10:12)
  • Nebius凭借多国业务分布、预计2027年到账的约200亿美元客户预付款,以及Avride、ClickHouse等可变现资产,构筑了三家中最稳健的资产负债表与抗周期护城河。 (7:10)

Detailed notes

## Background 本期内容由B站UP主“浑水摸鱼清源”于2026年11月17日发布,转译自频道【The Patient Investor】。视频简介聚焦三大新兴云服务商(NeoCloud)——IREN、CoreWeave与Nebius的走势分化,深入拆解跌幅最大的IREN在单季营收环比暴增182%的同时仍面临1.86亿美元经营亏损的底层财务逻辑,并对比三家的资产负债表与抗风险能力。

## TL;DR 视频横向拆解了三家NeoCloud的核心基本面与估值分歧。IREN主打买地自建机房与锁定5 GW电力的硬资产模式,估值仅约2倍P/S且债务集中在2032年之后,但经营层面尚未转正;CoreWeave客户网络已向工业与金融多元化拓展且经营利润转正,但背负550亿美元总债务且年内面临105亿美元到期偿债窗口,流动性最紧绷;Nebius估值最贵(约3.7倍P/S),但依托全球分散布局、高额预付款机制以及Avride等丰厚的可变现储备资产,资产负债表最为健康。通过现金对在手订单覆盖率指标,视频最终为不同风险偏好的投资者提供了定制化的配置指引。

## Atmosphere & style 主讲人的讲解风格严谨、冷静且极具穿透力,充斥着典型买方研究员条分缕析的数字直觉。叙事节奏层层递进,从表面跌幅与营收暴增的鲜明反差切入,剥开资本开支与商业模式的表象,直击企业流动性与偿债时间表的生死线。整个分析过程不带狂热情绪,既客观指出CoreWeave巨额短债悬顶的惊险,也透彻展示了Nebius预付款护城河的从容,带给观众一种在AI狂潮泡沫中抽丝剥茧、回归硬核财务防守的踏实感。

## Bulletin - [00:00](https://www.bilibili.com/video/BV1B5s36fEzM?t=0s) 三大新兴云厂商年内走势分化严重:IREN大跌57%、CoreWeave跌33%,仅Nebius逆势上涨28%,但市场核心抛压聚焦于盈利与资产负债表健康度而非算力需求端。 - [05:07](https://www.bilibili.com/video/BV1B5s36fEzM?t=307s) CoreWeave面临严峻的流动性考验,其在手现金仅约41亿美元,却背负550亿美元总债务,且一年内到期的长期债务流动部分即高达105亿美元。 - [10:12](https://www.bilibili.com/video/BV1B5s36fEzM?t=612s) 现金对在手订单覆盖率指标出现显著分层:IREN以28.5%位列第一,Nebius以26.5%紧随其后,而CoreWeave仅为5.1%,股权稀释与再融资风险最高。 - [07:10](https://www.bilibili.com/video/BV1B5s36fEzM?t=430s) Nebius凭借多国业务分布、预计2027年到账的约200亿美元客户预付款,以及Avride、ClickHouse等可变现资产,构筑了三家中最稳健的资产负债表与抗周期护城河。

## Key takeaways - [00:00](https://www.bilibili.com/video/BV1B5s36fEzM?t=0s) IREN单季营收环比飙升182%,调整后EBITDA利润率从14%翻倍至29%,但经营端仍录得1.86亿美元亏损。 - [01:04](https://www.bilibili.com/video/BV1B5s36fEzM?t=64s) 不同于同行租赁机房,IREN采取自持土地策略,锁定超3万英亩土地选择权与超5 GW电力指标,并依客户订单落地决定是否行权以控制资本风险。 - [02:04](https://www.bilibili.com/video/BV1B5s36fEzM?t=124s) IREN当前前瞻市销率约为2倍,估值处于低位,其35亿美元现金足以覆盖近期每年7至8亿美元的偿债需求,大额债务均在2032年之后到期。 - [04:06](https://www.bilibili.com/video/BV1B5s36fEzM?t=246s) CoreWeave客户结构不仅囊括OpenAI与科技巨头,还深入工业制造、金融交易与网络安全领域,有效分散了纯大模型业务的周期风险。 - [05:07](https://www.bilibili.com/video/BV1B5s36fEzM?t=307s) CoreWeave经营利润虽已转正至1亿美元且市销率仅约1.2倍,但550亿美元的巨额债务与年内105亿美元的到期额构成严重资金缺口隐患。 - [06:09](https://www.bilibili.com/video/BV1B5s36fEzM?t=369s) 得益于英伟达的强力背书与专属资产信贷便利,CoreWeave近期可转债票面利率大幅压降至2.5%,缓解了再融资利率成本。 - [07:10](https://www.bilibili.com/video/BV1B5s36fEzM?t=430s) Nebius业务分散在英国、法国、芬兰等多个国家,极大降低了对美国本土单一算力数据中心周期的依赖。 - [07:10](https://www.bilibili.com/video/BV1B5s36fEzM?t=430s) Nebius大额长期订单高度绑定预付款模式,预计2027年将回流200亿美元预付款,有效减轻对高息债务融资的依赖。 - [08:11](https://www.bilibili.com/video/BV1B5s36fEzM?t=491s) Nebius持有Avride自动驾驶、TripleTen教育以及ClickHouse等高价值独立股权,提供了极佳的流动性备选变现通道。 - [09:11](https://www.bilibili.com/video/BV1B5s36fEzM?t=551s) Nebius单季净利润转正至5,000万美元,手握90亿美元流动资产且债务覆盖率近半,支撑其高达3.7倍的市销率估值。 - [10:12](https://www.bilibili.com/video/BV1B5s36fEzM?t=612s) 现金对订单覆盖率对比显示,IREN(28.5%)与Nebius(26.5%)资金充裕度远胜CoreWeave(5.1%),后者面临更高的增发稀释风险。 - [11:12](https://www.bilibili.com/video/BV1B5s36fEzM?t=672s) 三家厂商投资画像明确:追求硬资产与高成长选IREN,承受高波动博取最大弹性选CoreWeave,追求资产负债表纪律与稳健选Nebius。

## Detailed walkthrough ### [00:00](https://www.bilibili.com/video/BV1B5s36fEzM?t=0s) IREN:自持土地策略、超高增速与亏损现状

作为当下一批备受瞩目的新兴云厂商(NeoCloud),市场对于究竟该买入哪一家存在激烈争论。主讲人开篇指出,IREN 年内跌幅最为剧烈,达到了 57%。但深入其基本面会发现,该公司的营收增长极其惊人:单季营收环比增长高达 182%(并非同比概念),调整后 EBITDA 利润率更是从上季度的 14% 飙升至本季度的 29%,订单积压(backlog)也在急速膨胀。

与其他竞争对手相比,IREN 展现出了截然不同的商业模式。主讲人指出:“What separates them from CoreWeave and Nebius is that IREN likes to own the land where their data center sits.” IREN 不希望通过长期租赁来运营机房,因为租赁机房意味着每隔几年就要面临租金上涨的压力,不仅运营成本受制于人,而且机房物理位置无法自由变动。为了将控制权牢牢掌握在自己手中,IREN 采取“买地自建”策略,在整条供应链上占据土地所有权。更重要的是,他们在全球锁定了超过 30,000 英亩的土地选择权(land option),并且锁定了超过 5 吉瓦(GW)的电力资源,大部分区域都紧邻充足的电力供应。这种策略极具灵活性:如果拿下微软或亚马逊等大客户的长期合同并拿到预付款,IREN 就会行使选择权买地建设机房;如果市场需求不及预期,他们仅仅损失一笔微小的期权权利金(tiny premium)。

估值方面,IREN 目前的前瞻市销率(Forward P/S)仅约为 2 倍。考虑到公司当前营收增速达 237%,明年预计增速为 200%,后年预计仍有 98%,这一估值水平被主讲人评价为“非常、非常便宜”。然而,IREN 的硬伤在于尚未实现盈利:经营亏损达 1.86 亿美元(相较于一年前的 6.2 亿美元明显收窄),净亏损为 2.63 亿美元(对比去年同期的 6.84 亿美元同样有大幅减亏)。资产负债表方面,IREN 账面拥有 35 亿美元现金,但面临约 110 亿美元的总债务。好在债务期限结构非常平缓,绝大多数到期日都在 2032 年或更晚,近期每年仅需偿付 7 亿至 8 亿美元,现有的 35 亿美元现金储备完全足以应对短期的债务到期。

```mermaid flowchart LR accTitle: IREN's Land Option Strategy accDescr: Flowchart illustrating how IREN acquires customer prepayments before exercising land options to build data centers.

subgraph S1["1. 储备锁定"] A["锁定全球 30,000+ 英亩土地选择权"] B["配套锁定 5+ GW 电力供应"] end

subgraph S2["2. 商务验证"] C{"大客户谈判<br/>(微软 / 亚马逊)"} end

subgraph S3["3. 执行与风控"] D["获得长期合同与预付款"] E["行使土地选择权买地建机房"] F["仅损失极少量期权费,放弃建厂"] end

A --> C B --> C C -->|"谈判成功"| D --> E C -->|"需求不足"| F ```

- 年内跌幅:57% - 营收环比增速:182% - 调整后 EBITDA 利润率:从 14% 提升至 29% - 资源储备:超 30,000 英亩土地选择权、超 5 GW 电力指标 - 估值:约 2 倍前瞻市销率 - 盈利状况:经营亏损 1.86 亿美元,净亏损 2.63 亿美元 - 现金与负债:35 亿美元现金储备对约 110 亿美元总债务;近期年均还本付息 7 亿至 8 亿美元,大笔债务在 2032 年之后到期

---

[04:06](https://www.bilibili.com/video/BV1B5s36fEzM?t=246s) CoreWeave:客户多元化、经营转正与严峻的短期债务压力

相比于 IREN,CoreWeave 年内跌幅为 33%,表现相对抗跌。CoreWeave 最大的亮点在于其极为多元化的客户结构。他们不仅绑定了顶尖的大模型独角兽和科技巨头——如 OpenAI、谷歌、微软、Anthropic,而且业务版图已经广泛拓展至纯 AI 模型之外的多个领域。主讲人极为赞赏这一点:“So they're more diversified. They're not all in LLMs and all in AI, which I personally like very, very much with CoreWeave.” 具体的多元化布局包括:

- 网络安全与边缘网络:为 CrowdStrike、Cloudflare 等巨头提供算力服务; - 工业与实体 AI(Industrial & Physical AI):涵盖宾利、阿斯顿·马丁、日产等车企; - 金融服务:服务于纳斯达克、摩根士丹利、简街资本(Jane Street)等顶尖金融机构。

在财务表现上,CoreWeave 的经营利润已经成功转正至 1 亿美元。其市销率约为 1.2 倍,比 IREN(约 2.2 倍)还要低廉。虽然 2027 年预计增速放缓至 100%、2028 年降至 64%,低于 IREN 的增长预期,但就 1.2 倍的估值倍数而言依然具有吸引力。

然而,CoreWeave 隐藏的重大隐患在于其资产负债表。公司账面仅有约 41 亿美元现金及现金等价物,却背负着高达约 550 亿美元的总债务。最为致命的是短期还款集中度:一年内到期的长期债务流动部分(current portion of long-term debt)高达 105 亿美元,另外还有 20 亿美元其他流动负债和 83 亿美元长期负债。更细化的债务到期时间表显示,2027 年到期债务近 110 亿美元,2028 年到期 54 亿美元,2029 年到期 47 亿美元,2030 年之后还有约 200 亿美元。与 IREN 债务大部分分布在 2032 年之后截然不同,CoreWeave 的现金流缺口迫在眉睫。

缓解其流动性危机的一个积极信号是其融资成本在急剧下行。2023 年 CoreWeave 的借贷成本曾高达 15%,随后回落至 7.8%,近期发行的可转换债券票面利率仅为 2.5% 左右。这为公司在未来通过低息再融资、拉长债务期限以削减利息开支提供了可能。主讲人指出,CoreWeave 拥有英伟达的强力背书,在最糟糕的情况下,英伟达极有可能施以援手提供低息贷款。此外,CoreWeave 针对特定大客户(如微软)设立了独立的资产信贷便利工具(asset facilities),债务挂钩于具体合约实体而非母公司自身,即使单一项目发生波动也不会直接击垮母公司,但其总体现金与债务错配的风险依旧不容忽视。

---

[07:10](https://www.bilibili.com/video/BV1B5s36fEzM?t=430s) Nebius:全球化布局、丰富资产与预付款护城河

Nebius 是三家新兴云厂商中唯一在年内录得正回报的标的,涨幅约为 28%,但由于前期剧烈波动,距离历史最高点也已回调近 60%。Nebius 的核心竞争优势首先体现在极其分散的地理布局上。主讲人强调,如果美国本土的数据中心投资情绪降温,Nebius 并不会全盘皆输,其业务版图广泛分布于英国、法国、西班牙、以色列、印度、芬兰、冰岛和爱沙尼亚等多个国家,能够有效规避单一市场的政策与市场风险。

另一个被主讲人反复称赞的护城河是“预付款机制”(prepayments)。Nebius 高度专注于签署包含大额预付款的长期合同,预计在 2027 年将有约 200 亿美元的预付款入账。这种模式使公司在交付巨额积压订单时,无需过分依赖高息借贷来支撑资本开支。在业绩指引方面,Nebius 预计年营收将从当前的约 30 亿美元飙升至 2027 年的 130 亿美元(实现近 5 倍爆发式增长),2027 年预计增速接近 300%,2028 年预计增速仍达 112%。

此外,Nebius 拥有其他两家公司完全不具备的变现后备资产——大量可拆分出售或抵押的独立业务和股权: - 自动驾驶业务 Avride; - 在线教育与培训平台 TripleTen; - 数据与 AI 标注平台 Toloka; - 开源分析型数据库 ClickHouse 的股权。 主讲人指出,相关研报显示这些股权资产的实际价值远高于账面估值。如果 Nebius 未来在云计算主业扩张中遭遇资金紧张,可以随时将这些资产抵押贷款或直接出售变现。

财务指标上,Nebius 本季度实现经营利润转正,单季录得 5,000 万美元净利润(前一季度亏损 1.75 亿美元)。估值方面,Nebius 是三者中最贵的,市销率约为 3.7 倍。资产负债表上,公司拥有约 90 亿美元的流动资产(包含 74 亿美元现金及现金等价物,以及 16 亿美元主要由股权构成的长期投资),面对的是 140 亿美元长期债务和 81 亿美元其他长期负债,债务总额约为 220 亿美元。现金与投资对长期债务的覆盖率接近 50%,整体资产负债表相对稳健。

---

[10:12](https://www.bilibili.com/video/BV1B5s36fEzM?t=612s) 资金压力终极对比与三大标的投资画像

| 对比维度 | IREN | CoreWeave | Nebius | |:--- |:--- |:--- |:--- | | **年内涨跌表现** | 下跌 57% | 下跌 33% | 上涨 28%(距高点回撤 ~60%) | | **市销率估值 (P/S)** | 约 2.0 - 2.2 倍 | 约 1.2 倍 | 约 3.7 倍 | | **核心商业护城河** | 自持土地与机房资产,5 GW 电力储备 | 顶级客户网络,非 AI 工业/金融渗透 | 跨国分散布点,高预付款占比,可变现子公司 | | **经营利润状态** | 亏损 1.86 亿美元(持续收窄) | 盈利 1 亿美元 | 盈利 5,000 万美元 | | **现金及长期投资** | 约 35 亿美元现金 | 约 41 亿美元现金 | 约 90 亿美元(74 亿现金 + 16 亿股权) | | **总债务/长期负债** | 约 110 亿美元(主债期在 2032 年后) | 约 550 亿美元(年内到期约 105 亿) | 约 220 亿美元 | | **现金对积压订单覆盖率** | **28.5%(排名第一)** | **5.1%(排名垫底)** | **26.5%(排名第二)** |

主讲人分享了一张关键信息图表,该图表通过计算“在手现金+现金等价物+长期投资”对“积压订单(backlog)”的覆盖比率,直观反映哪家公司未来最有可能被迫增发股票稀释股东,抑或陷入流动性危机: - **IREN** 的覆盖率达 **28.5%**,表现最为优秀; - **Nebius** 紧随其后,覆盖率为 **26.5%**; - **CoreWeave** 则出现严重断层,覆盖率仅有 **5.1%**。

结合 CoreWeave 仅 41 亿美元的现金与年内即将到期的 105 亿美元债务,其短期的再融资与现金流压力在三家中最为紧绷。针对不同风险偏好的投资者,主讲人给出了清晰的画像与配置结论:

1. **IREN(实物资产与成长型配置)**:“If you looking more at the real assets real estate play with some growth I would say IREN is the best.” 适合看重土地和电力等硬资产所有权、看好高增速且能容忍公司暂时亏损的投资者。 2. **CoreWeave(高风险高弹性投机)**:“If you looking for the maximum upside I would say CoreWeave is the riskiest due to the balance sheet.” 尽管债务压力最大,但有英伟达低息贷款兜底的预期、独立合约信贷结构以及极其多元的优质客户群,一旦再融资顺利推进,其股价反弹空间最大。 3. **Nebius(稳健与纪律型标的)**:“If you're looking for a safer option for something that's a little bit more disciplined, a good CEO and better balance sheet in general and decent growth and decent margins, I would say Nebius is the safest.” 尽管 3.7 倍的 P/S 估值最高,但得益于充沛的现金、预付款模式以及多项可变现资产,Nebius 在面临行业逆风时具备最高的确定性。

## Notable quotes > [10:12](https://www.bilibili.com/video/BV1B5s36fEzM?t=612s) If you're looking for a safer option for something that's a little bit more disciplined, a good CEO and better balance sheet in general and decent growth and decent margins, I would say Nebius is the safest. — 主讲人

> [10:12](https://www.bilibili.com/video/BV1B5s36fEzM?t=612s) If you looking for the maximum upside I would say CoreWeave is the riskiest due to the balance sheet. — 主讲人

> [10:12](https://www.bilibili.com/video/BV1B5s36fEzM?t=612s) If you looking more at the real assets real estate play with some growth I would say IREN is the best. — 主讲人

## Mentioned IREN, CoreWeave, Nebius, 微软 (Microsoft), 亚马逊 (Amazon), OpenAI, 谷歌 (Google), Anthropic, CrowdStrike, Cloudflare, 宾利 (Bentley), 阿斯顿·马丁 (Aston Martin), 日产 (Nissan), 纳斯达克 (Nasdaq), 摩根士丹利 (Morgan Stanley), 简街资本 (Jane Street), 英伟达 (NVIDIA), Avride, TripleTen, Toloka, ClickHouse

## Video description 【The Patient Investor】 单季营收涨182% IREN仍亏1.86亿 The BEST NeoCloud Stock To Buy! Nebius VS IREN VS CRWV

|三家都在跌 但跌法不一样 三家新兴云厂商今年表现分化得厉害。IREN年内跌了57%,CoreWeave跌33%,Nebius反而还涨28%,不过离历史高点也回撤了近60%。跌得最狠的那家,收入增速反而是最猛的:单季营收环比涨182%,调整后EBITDA利润率从14%拉到29%,订单积压也在快速膨胀。市场杀的是盈利和资产负债表,不是需求。这个区别很关键,决定了后面谁能熬、谁要先融资。

|IREN:买地自建 估值最便宜 IREN最不一样的地方是不租机房,自己买地自己建。全球锁定了超过3万英亩的土地选择权,配套超过5吉瓦电力,先拿到微软、亚马逊这类大客户合同和预付款,再行使选择权买地开工;谈不下来就损失一笔权利金,风险可控。估值只有约2倍前瞻市销率,而明年营收预计还能翻两倍。问题在盈利:经营亏损1.86亿美元,净亏损2.63亿美元,虽比一年前的6.2亿明显收窄,但还没转正。35亿现金对约110亿总债务,好在到期日大多在2032年之后,近期每年只还7到8亿,现金够覆盖。

|CoreWeave:客户最散 债务最急 CoreWeave的客户结构是三家里面最健康的,OpenAI、谷歌、微软、Anthropic都在里面,还切进了CrowdStrike、Cloudflare的网络安全业务,以及阿斯顿·马丁、日产这类工业客户和纳斯达克、摩根士丹利的金融业务,不全押在大模型上。经营利润已经转正到1亿美元,估值约1.2倍市销率,比IREN还便宜。真正的雷在债务:41亿现金对约550亿债务,其中约105亿是今年就要还的流动部分,2027年到期约110亿,2028年54亿,2029年47亿。融资成本倒是在降,2023年还要15%,最近的可转债只付2.5%,英伟达在背后撑着,再融资空间是有的。

|Nebius:最贵 但家底最杂 Nebius是三家里面唯一还在涨的,估值也最贵,约3.7倍市销率。贵有贵的道理:今年营收30亿,明年指引130亿,等于一年翻四倍多,2027年预计增长近300%,2028年还有112%。地理上也是最分散的,英国、法国、西班牙、以色列、印度、芬兰、冰岛、爱沙尼亚都有布点,美国市场出问题不至于全盘皆输。手里还攥着一堆可以卖的资产:自动驾驶公司Avride、在线教育平台TripleTen,还有Toka和ClickHouse的股权,需要钱的时候可以抵押或变现。2027年有约200亿预付款到账,不用全靠借债撑订单。财务上90亿现金加长期投资,对约220亿长期债务,覆盖率接近一半。经营利润也转正了,净赚500亿(单位应为百万级),前一季还亏1.75亿。

|一个指标分出谁最缺钱 把在手现金和长期投资,跟订单积压做对比,能看出谁最可能被迫增发稀释股东。IREN覆盖率约28.5%,是三家里最好的;Nebius约26.5%,紧随其后;CoreWeave只有5.1%,差距不是一点半点。这个数字配合它今年到期的105亿债务一起看,压力就很直观了。三家的排序大致是:想要实物资产加成长,选IREN;想要最大弹性、能承受波动,选CoreWeave,英伟达背书和分散的债务结构是缓冲;想要稳一点、管理层纪律好、资产负债表干净,选Nebius。

接下来建议的操作指引: 1. CoreWeave(CRWV)2027年有约110亿美元债务到期,盯住其后续可转债发行利率是否维持在2.5%附近,利率明显抬升说明再融资窗口在收窄。 2. IREN关注经营亏损能否从1.86亿美元继续收窄至转正,同时看3万英亩土地选择权是否有新的大客户预付款落地。 3. Nebius明年的130亿美元营收指引是关键验证点,若季度营收环比增速掉到50%以下,3.7倍市销率的估值就撑不住了。

[0:03] 三家新兴云厂商对比框架与年内涨跌 [1:04] IREN自持土地与电力储备策略 [2:04] IREN估值与亏损收窄情况 [3:04] IREN债务到期结构与现金覆盖 [4:06] CoreWeave客户多元化与经营转正 [5:07] CoreWeave 550亿债务与近期到期压力 [6:09] CoreWeave融资成本下降与英伟达背书 [7:10] Nebius全球布点与预付款模式 [8:11] Nebius可出售资产与经营利润转正 [9:11] Nebius估值与资产负债表覆盖率 [10:12] 现金对订单覆盖率排名 [11:12] 三家排序与各自适合的投资者

[Source](https://www.bilibili.com/video/BV1B5s36fEzM)

中文简报

CoreWeave面临百亿短债重压 新兴算力云现流动性分水岭 CoreWeave在手现金对积压订单覆盖率仅5.1%且短期偿债缺口巨大,再融资风险显著高于手握实物资产的IREN与高预付款护航的Nebius。 视频横向拆解了三家NeoCloud的核心基本面与估值分歧。IREN主打买地自建机房与锁定5 GW电力的硬资产模式,估值仅约2倍P/S且债务集中在2032年之后,但经营层面尚未转正;CoreWeave客户网络已向工业与金融多元化拓展且经营利润转正,但背负550亿美元总债务且年内面临105亿美元到期偿债窗口,流动性最紧绷;Nebius估值最贵(约3.7倍P/S),但依托全球分散布局、高额预付款机制以及Avride等丰厚的可变现储备资产,资产负债表最为健康。通过现金对在手订单覆盖率指标,视频最终为不同风险偏好的投资者提供了定制化的配置指引。 三大新兴云厂商年内走势分化严重:IREN大跌57%、CoreWeave跌33%,仅Nebius逆势上涨28%,但市场核心抛压聚焦于盈利与资产负债表健康度而非算力需求端。 (0:00) CoreWeave面临严峻的流动性考验,其在手现金仅约41亿美元,却背负550亿美元总债务,且一年内到期的长期债务流动部分即高达105亿美元。 (5:07) 现金对在手订单覆盖率指标出现显著分层:IREN以28.5%位列第一,Nebius以26.5%紧随其后,而CoreWeave仅为5.1%,股权稀释与再融资风险最高。 (10:12) Nebius凭借多国业务分布、预计2027年到账的约200亿美元客户预付款,以及Avride、ClickHouse等可变现资产,构筑了三家中最稳健的资产负债表与抗周期护城河。 (7:10)

Reading summary

主讲人对比了 IREN、CoreWeave 和 Nebius 三家新兴云厂商(NeoCloud)的商业模式、财务状况与偿债压力,指出三者在重资产自持、高杠杆弹性与现金流防御能力上呈现出显著分化。 商业模式分歧:自持硬资产与客户网络 (0:00) 在基础设施布局上,IREN 采取“买地自建”的重资产策略,在全球锁定了超过 30,000 英亩土地选择权及 5 GW 以上电力供应。通过在大客户签署长期合同并支付预付款后再行使期权买地建厂,IREN 规避了机房长期租金上涨和运营受制于人的风险;若需求不达预期,损失也仅限于少量期权费。与此相对,CoreWeave 则侧重于多元化的客户生态,除 OpenAI、谷歌、微软等 AI 巨头外,已广泛渗透至 CrowdStrike 等网络安全公司、宾利等实体工业车企以及简街资本等顶级金融机构。Nebius 则选择跨越英国、法国、西班牙、芬兰等多个国家分散布点,借此规避单一区域市场的政策与投资周期风险。 债务期限错配:CoreWeave短债与IREN长债 (4:06) 各厂商在资产负债表与偿债节奏上面临截然不同的流动性环境。CoreWeave 尽管单季经营利润转正至 1 亿美元,但背负约 550 亿美元总债务,账面现金仅约 41 亿美元,其中一年内到期的长期债务流动部分高达 105 亿美元,短期再融资压力突出。虽然其借贷成本已从 2023 年的 15% 降至近期可转债的约 2.5%,且有英伟达潜在信贷兜底预期与独立信贷隔离工具,但现金错配隐患依然显著。相比之下,IREN 尽管尚未盈利(单季经营亏损 1.86 亿美元,净亏损 2.63 亿美元),但其 35 亿美元现金面对的是年均仅 7 亿至 8 亿美元的近期还本付息支出,约 110 亿美元的总负债绝大部分在 2032 年及之后才到期。 Nebius防御机制:预付款与可变现资产 (7:10) Nebius 构筑了更具防御性的财务闭环,其单季录得 5,000 万美元净利润,年营收预计从当期约 30 亿美元大幅扩张至 2027 年的 130 亿美元。主讲人重点指出其“预付款机制”优势,公司高度聚焦附带大额预付款的长期合约,预计 2027 年将有约 200 亿美元预付款入账,极大减轻了对外部债务融资的依赖。此外,Nebius 账面拥有 74 亿美元现金与 16 亿美元长期投资,且手握自动驾驶公司 Avride、教育培训平台 TripleTen、数据标注平台 Toloka 以及数据库 ClickHouse 股权等多项独立后备资产。主讲人指出,相关股权资产实际价值被认为远高于账面值,可在面临资金压力时作为抵押或出售变现。 订单现金覆盖率与三类投资画像 (10:12) 在衡量后续流动性安全与股东稀释风险的“在手现金及投资对积压订单(Backlog)覆盖率”指标上,IREN 以 28.5% 位列第一,Nebius 以 26.5% 紧随其后,而 CoreWeave 仅为 5.1%,存在明显的兑付断层。估值维度上,CoreWeave 市销率约 1.2 倍最为低廉,IREN 前瞻市销率约 2 至 2.2 倍,Nebius 则为 3.7 倍。主讲人据此归纳了三类投资配置画像:IREN 属于适合看重土地与电力硬资产自持、能承受阶段性亏损的实物资产与成长型标的;CoreWeave 债务风险最高但反弹弹性最大,属于高风险投机标的;Nebius 则凭借充沛现金、预付款模式与健康资产负债表,成为行业逆风下最具纪律性与确定性的防御之选。 新兴云厂商的高成长叙事受到资本开支与偿债节奏的双重制约。各厂商能否在行业波动中兑现高增速,核心取决于其短期再融资能力、大客户预付款到位情况以及经营现金流转正的速度。

Source quotes

[10:12] 主讲人: If you're looking for a safer option for something that's a little bit more disciplined, a good CEO and better balance sheet in general and decent growth and decent margins, I would say Nebius is the safest. [10:12] 主讲人: If you looking for the maximum upside I would say CoreWeave is the riskiest due to the balance sheet. [10:12] 主讲人: If you looking more at the real assets real estate play with some growth I would say IREN is the best.

Original transcript

[0:00] Which NeoCloud is the best one to buy in this market? Is it Iron? Is it Nibius? Or is it even CoreWeave? That's what I'm going to answer in today's video. I'm going to be comparing all three of them, tell you what I like about each one, what I don't like about each one of them, talk about the risks, and then compare them and tell you which one I believe is the best. And I really hope you enjoy it as always. The first one is Iron. I've done a deeper dive on Iron and each one of them, you could check them out over here. is down 57% year to date. What I really like about IRON is the revenue growth. The company is growing 182% in revenues, not year on year, quarter over quarter. They're expanding adjusted EBITDA margin like crazy. Last quarter adjusted EBITDA was 14%. Right now it's 29%. And the backlog has them from Corweave and Nibius is that IRON likes to own the land where their data center sits. So they like to own where they operate. They like to own the real estate. They like to be in control of their own destination and their own buildings in general. They don't want to be hooked up to those leases where every few years they have to increase the lease payment and they have to keep, you know, paying rent on where they operate. And you can't change where you operate. So IRON has been taking the land wherever they operate. And they also have like a land option on over 30,000 acres all over the world. So if they can get a good customer like Microsoft or Amazon, get them on a contract, get some prepayment, and if they can get a good deal, then they can exercise the option, buy the land and build the data center. But if they don't get enough interest, then they lose a tiny premium for this option. So they have an option of over 30,000 acres and over 5 gigawatt of secured and a lot of it is very close to power in general. So that's a really good strategy from Iron. I also like the valuation. The trading at two times sales, two times forward sales, is not expensive at all for a company that's growing 237% in revenues. Next year is expected to grow 200%, and the year after is 98% gross. Two times sales is very, very cheap for Iron. Now, what I don't like about Iron is the fact that the company is still unprofitable, operating profit basis, it's posting an operating loss of 186 million. Now this is an improvement from last year, they were putting up 620 million. Now they're putting up 186 million as an operating loss. So this is a big improvement, but they're not yet profitable. And if we use net income, they're losing 263 million, but it's still an improvement from 684 million a year ago, pretty good improvement, but it's the company still unprofitable. I also don't really like the balance sheet of iron. The company has three and a half billion dollars in cash against about 11 billion dollars in total debt. So they don't have as much in cash. They have a lot of debt. Now, the good thing is that debt is well spread out and the maturities, they don't really come up until 2032 or after. So in the near term, they only have 700 to 800 million dollars due every single year. And the $3 billion cash position is more than enough for them to cover the near-term maturities. And this is my quick analysis on IRON. Now, I'm going to compare it to CoreWeave. CoreWeave is down less than IRON year-to-date. It's down 33% on the year. What I like about CoreWeave is how diversified their customer base is. They have a lot of the big players like OpenAI, Google, Microsoft, Entropic, and they have some really big customers. But they're also diversified away from AI. for CrowdStrike and CloudFlare. They're industrial fiscal AI like Bentley, Aston Martin, Nissan, also in financial services like NASDAQ, Morgan Stanley, Jane Street. So they're more diversified. They're not all in LLMs and all in AI, which I personally like very, very much with CoreWeave. And this is something that IRN doesn't have yet. IRN is mostly in the big players and in the AI stuff. I also like that CoreWeave is profitable on an operating profit basis. They were able to flip it positive to $100 million, which is great. I also like the valuation on CoreWeave. CoreWeave's valuation is about 1.2 times sales. Iron's valuation is about 2.2 times. So CoreWeave is a little bit cheaper than Iron. But CoreWeave is expected to grow a little bit less than Iron. CoreWeave is expecting 100% in 2027 and 64% grows in 2028. It's still very, very cheap for the multiple. [5:00] but not as cheap as iron in this case, as iron is expected to grow more. Now, the worst thing about CoreWave is the balance sheet. So the company has $4 billion in cash-in-cash equivalent against approximately $55 billion in debt. They have about $10.5 billion in the current portion of long-term debt. So that's actually due this year. They have $2 billion in other liabilities. long-term abilities have $8.3 billion. Total is $55 billion against a $4.1 billion cash position. And the worst part about it is a lot of it is due in the near term. So they have almost $11 billion due in 2027. They only have $4.1 billion in cash. Iron, as I told you, has mostly longer-term maturities that can deal with the short-term stuff. Most of it is due after 2032. But for CoreWeave, So they have $11 billion due in 27, 5.4 in 28, 4.7 in 29. And then after 2030, they have about $20 billion, which is not bad. But some of it is due in the near term. Now, the good thing is, as I told you guys, CoreWeave has been improving in terms of the cost of debt. So the cost of debt a few years ago in 2023 was 15%. It was 7.8%. And they even issued convertible debt recently at about a 2.5%. rate. So they can most likely refinance some of the debt at even lower interest rates than they had and save on interest expense and extend the maturities. And again, this one is backed by NVIDIA. So I think if they needed help, NVIDIA will most likely help them with a lower interest rate loan. So that's not really as bad, but still, I think IRON has a better balance sheet in this case. And that's what I like and dislike about CoreWeave. I like everything about it, but the Now, the last one is Nibius. Nibius is the only one that's still up 28% year to date. It's down a lot from all time high, almost 60%. What I like about Nibius is the diversification. So the company is not all in the United States. If the data center sentiment in the United States is bad, they still have in the UK, they have in France, Spain, Israel, India, and Finland, and Iceland, and Estonia, and then they have them all over the world. So the company is very, very diversified. much. I also like that the company is focused a lot on contracts that have like a prepayment thing. For example, they have about $20 billion of prepayments in 2027, which will help them not, you could say, take on a lot of debt to meet the backlog as a lot of it is prepaid. I also like the growth with Nibius because Nibius is expecting $13 billion this year, 27. Last year, which is this So this is almost a, I mean, it's like a 5X in growth year over year. Absolutely crazy. I also like that the company has so many of those other businesses they can sell. And this is something that Iron and CoreWeave don't have. This one has a company called AVRide, which is an autonomous vehicle. If they need the money to ramp up clouds, they could sell it. Same thing with Triple 10, which is like some kind of an online school. You could learn stuff. They also have Toloka and ClickHouse. I read a lot of reports that they're worth much more on their books. So I also like those equity stakes where if the company needs to take a loan against it, monetize it, sell it, they have a second option to do it. I also like that Nibius posted a positive operating profits quarter. They made $50 billion in that profit after losing $175 million the quarter prior to that. So I still like that very much. Now, the valuation on Nibius is the highest. I think iron is trading at somewhere around 2.2 times sales. Coreweave is trading at 1.2 times sales. And iron is the most expensive at 3.7 times sales. But if we look at the growth of Nibius, Nibius has much more growth. In 27, they're expecting almost 300%. And in 28, 112%. So it's not really as expensive as it looks. is I would say in the debt and the balance sheet in general. And the company has about $9 billion in cash and cash equivalent, 7.4 in cash and cash, and $1.6 billion in long-term investments, which is mainly the equity stakes. Now, they also have $14 billion in long-term debt and about $8.1 billion in other long-term liabilities. So this is about $22 billion in long-term debt against $9 billion in cash and long-term investment. So almost, [10:00] percent of it is covered. I would have liked to see a little bit more cash, but that's the only major risk that I see with Nibias. Now, I posted an infographic, which I believe is really nice. And this is mainly the company's total cash and cash equivalent and long-term investments against the backlog. So what percentage of the backlog could be funded through the cash on hand? And then the rest, they would have to either be prepayments or they would have to issue shares. So this metric would show you which companies will likely dilute more in the future or will be a little bit more trouble. The best one was IRON. IRON's backlog relative to the company's cash position is about 28.5% covered. The second one was Nibius at 26.5%. And the worst one was CoreWeave at 5.1%. As I told you guys, CoreWeave has by far the worst balance sheet between those two or between those three, and they also have the nearest term maturities. So it's really hard for me to say which one is the best. I would say if you looking more at the real assets real estate play with some growth I would say iron is the best If you looking for the maximum upside I would say CoreWeave is the riskiest due to the balance sheet But they have the NVIDIA backing and they have some quality customers. And CoreWeave also has different asset facilities for different contracts. So, for example, they have a debt facility with, let's say, Microsoft. If the deal fails, then this asset facility would fail, but it wouldn't really affect CoreWeave. because each debt facility is tied to like a different corporation related to this specific contract. So it's not really to the parent. It's more based on the contract. So that looks risky, but it's not as risky as it is. But I think it has the most upside. The last one is Nibius. If you're looking for a safer option for something that's a little bit more disciplined, a good CEO and better balance sheet in general and decent growth and decent margins, I would say Nibius This is the safest, I would say, iron is a great role state play and the one with the most upside but more risk is core weave. This is how I would rank them. It was not financial advice. Hope you enjoyed the video. If you enjoy it, please press the like button and maybe consider subscribing so I'll talk to you in another video.