· 个股论点

解析超大规模云厂商资本支出向NeoClouds转移的逻辑,看好NBIS等标的的高回报潜力。

涉及标的:

中文翻译

NeoCloud 论点:超大规模云厂商资本支出漏斗 为什么我要投入 150 万美元以上到 NeoClouds(新型云服务商),以及为什么这可能带来 200-300%+ 的回报。 🔹 分类 Mag7(美股七巨头)合约方:$CRWV, $NBIS ✅, $WULF, $CIFR ✅ 带算力:$IREN ✅, $BITF 投机性:$WYFI, $GRRR ✅, $SLNH 转向 HPC(高性能计算)的矿工:$RIOT, $MARA, $CLSK, $HUT 论点: Mag7 的 AI 算力紧张,这是 $NVDA 设计使然。 原本流向 AWS、MSFT Azure、Google Cloud 用于传统算力的数万亿美元资本支出,当它们无法处理来自 Anthropic、OpenAI、Gemini 等的新 AI 负载时,现在将漏斗式流入 NeoClouds。 这是一个十年一遇的机会,类似于让 $NVDA 成为 4 万亿美元公司的 GPU 军备竞赛,关键在于未来 5-10 年为 AWS/Azure 等提供基础设施支持的是谁。 NBIS(来自 MSFT 的 170 亿)、CIFR / WULF(来自 GOOGL 的 30 亿)、CRWV(由 NVDA 背书)都在以数百百分比的速度扩张(NBIS 季度收入从 1.5 亿增长到可能超过 15 亿),且拥有 60-80% 的毛利率。 这种收入增长在历史上几乎闻所未闻。这主要是因为最富有的超大规模云厂商将资本支出漏斗式注入这些小公司。 NVDA / TSM(2022->):为超大规模云厂商提供 GPU CRDO / ALAB(2024 ->):超大规模云厂商获胜 -> 抛物线式增长。 NBIS/CIFR/IREN 等(2025 - )AWS/Azure 等 -> 来自 AI 算力的抛物线式增长 这就是如何获得数百百分比回报的方法,而不是在 Paypal 上进行价值投资。动量交易下一代公司。 所以看空论点通常涉及 - 执行风险(以前更偏向投机,现在像 NBIS 这样的公司有 40 多亿资金来执行) 人们总是担心执行问题,但如果微软或谷歌没有做自己的尽职调查,是不会签署如此大的 5-10 年合约的。 - 高利率(主要看你 CRWV),这就是为什么 NBIS、CIFR 等具有惊人的回报潜力。 $NBIS 在每股 138 美元以上时有 40 多亿资金(现在是 107 美元)。$CIFR 在每股 16 美元以上时有资金(现在是 11 美元)。高于当前价格的资金是一个看涨信号。 - GPU 折旧(有效的担忧,但这几乎像石油一样,即使是旧型号也保持了价值并仍然提供权益)。 - 估值(我认为我们才刚刚开始。如果 NBIS 明年规模达到 60 亿收入,75% 毛利率),260 亿市值非常小。 - NVDA 可能推出自己的 GPU 即服务并直接竞争。目前这些 NeoClouds 是 NVDA 防止对 Azure/AWS 等集中风险的答案。 - 定制超大规模云厂商芯片如 TPU、Trainium。但可能还需要几年,因为他们仍在乞求 NVDA 算力并签署了 5-10 年合约。 无论如何,像 Nebius 这样的 NeoClouds 相对于远期收入/毛利率确实被低估了。 我们仍然处于非常早期的阶段。确保像特朗普选举时的 Crypto/TSLA 或 OpenAI 发布时的 NVDA 一样搭乘 NeoCloud 浪潮。 当然,这高度投机,我不会建议全仓 YOLO,但对于这些为 Google/MSFT 等(Mag7 的现金牛)提供 AI 工作负载的小型 5-200 亿 NeoClouds,风险与回报是值得的。(交易时间范围:8 个月 - 1 年。) 这是 2025-2026 年最好的非对称 AI 基础设施交易。

英文原文

The NeoCloud Thesis: Hyperscaler Capex Funnel Why I'm putting $1.5M+ into Neoclouds, and why this might be a 200-300%+ return. 🔹 Buckets Mag7 contracts: $CRWV, $NBIS ✅, $WULF, $CIFR ✅ With compute: $IREN ✅, $BITF Speculative: $WYFI, $GRRR ✅, $SLNH Miners pivoting to HPC: $RIOT, $MARA, $CLSK, $HUT Thesis: Mag7 is AI compute strained, by design from $NVDA. Trillions of capex that normally flowed through AWS, MSFT Azure, Google Cloud for traditional compute, will now funnel into NeoClouds when they cant handle new AI loads from Anthropic, OpenAI, Gemini, etc. This is a once-a-decade opportunity, similar to the GPU arms race that made $NVDA a $4T company, on who powers the infra for AWS/Azure/etc for the next 5-10 years. NBIS (17B from MSFT), CIFR / WULF (3B from GOOGL), CRWV (backstopped by NVDA) are all scaling hundreds of percent (NBIS went from 150M quarterly revenue to likely 1.5B+) with 60-80% gross profit margins. This revenue growth is almost unheard in history. It's mainly because it's the wealthiest hyperscalers funneling capex into tiny companies. NVDA / TSM (2022->): GPU for hyperscalers CRDO / ALAB (2024 ->): hyperscaler wins -> parabolic growth. NBIS/CIFR/IREN/etc (2025 - ) AWS/Azure/etc. -> parabolic growth from AI compute This is how you get hundreds of % in return, not value investing in Paypal. Momentum riding the next generational companies. So bear thesis usually involves around - Execution Risk (before it was more speculative, now companies like NBIS have 4B+ to execute) People can always worry about execution but Microsoft or Google would not be signing such large 5-10 year contracts without their own DD. - Large interest rates (mainly looking at you CRWV), that's why NBIS, CIFR, and others have potential amazing returns. You have 4B+ in funding for $NBIS at $138+ a share (when it's $107 now). And funding for $CIFR at $16+ a share when it's $11 now. ABOVE current price funding is a bullish tell. - GPU depreciation (valid concern but it's almost like oil, even older models kept their value and still deliver equity). - Valuation (I still think we're just getting started. If NBIS scales to 6B rev next year 75% gross margin), 26B marketcap is extremely tiny. - NVDA potentially launching their own GPU-as-a-service and directly competing. Right now these Neoclouds are NVDA's answer to preventing concentration risk to Azure/AWS/etc. - Custom hyperscaler chips like TPU, Trainium. But likely years away, since they're still begging for NVDA compute and signed 5Y-10Y contracts. Regardless these neoclouds like Nebius are really undervalued relative to forward revenue/gross margins. We're still very early. Make sure to ride the Neocloud wave like Crypto/TSLA with Trump election or NVDA with OpenAI release. Of course this is highly speculative and I wouldn't YOLO full port calls, but Risk vs. Reward on these little 5-20B neoclouds powering AI workloads for Google/MSFT/etc, the cash cows of Mag7, is worth it. (Trade time Horizon: 8m - 1 year.) This is the single best asymmetric AI infrastructure trade for 2025-2026.

在 X 上查看原推 ↗