Serenity：Neocloud Ecosystem Cheat Sheet Part 1/2: Bullet Point Positive vs. Negatives. Full List by Mark…

推文摘要

这条推文含有财经/市场相关信息或标的符号，核心内容是：Neocloud Ecosystem Cheat Sheet Part 1/2: Bullet Point Positive vs. Negatives. Full List by Marketcap: $CRWV ($66.2B) $NBIS ($32.85B) $IREN ($16.52B) $APLD ($9.69B) $RIOT ($7.34B) $CIFR ($7.3B) $WULF ($6.36B) $HUT($5.39B) $CLSK ($5.01B) $HIVE ($1.74B CAD) $WYF…

AI 财经深入分析

原始内容

Neocloud Ecosystem Cheat Sheet Part 1/2: Bullet Point Positive vs. Negatives. Full List by Marketcap: $CRWV ($66.2B) $NBIS ($32.85B) $IREN ($16.52B) $APLD ($9.69B) $RIOT ($7.34B) $CIFR ($7.3B) $WULF ($6.36B) $HUT($5.39B) $CLSK ($5.01B) $HIVE ($1.74B CAD) $WYFI ($1.29B) $WLAC($600m IPO) $DGXX ($393M CAD) $SLNH ($281.9M) _ Summary When comparing, a major source of alpha generation currently lies in the megawatt valuation arbitrage, which involves converting low-multiple Bitcoin power capacity into high-multiple AI hosting capacity. The second major alpha is margin generation, which involves being vertically integrated from the bottom up from GPU orchestration to software. Coreweave ( $CRWV ) - $66.2B Marketcap 🍏Positives ________________ - Sector leader by scale: quarterly revenue of $1.21B (+206.75% YoY) and EBITDA of $607.69M; on pace for $5B+ ARR in 2025. - $30B+ backlog, anchored by: - $14B Meta deal, - OpenAI - $6.3B NVIDIA GPU backstop agreement, - growng Gov contracts via CRWV Federal . - Expansion into U.S. government infrastructure is a major long-term moat if backstopped by federal workloads. - NVIDIA partnership ensures utilization floor; de-risks GPU oversupply. - Positioned as hyperscaler alternative with Tier 1 clients and national buildout footprint. Negatives ________________ - Aggressive capex: Q3 spend of $2.9–3.4B, with 2025 full-year guidance of $20–23B. - High-cost debt structure: Over $1B in projected annual interest, versus competors using low- or zero-interest convertible notes. - Q2 GAAP net loss of $291M, signaling limited profitability despite scale. - Execution risk remains high: e.g., failed $9B Core Scientific acquisition due to shareholder rejection. - Competitors like Nebius are closing the full-stack gap, weakening CoreWeave's software moat. Nebius ( $NBIS ) - $32.85B MarketCap 🍏Positives ________________ - $17.4B Microsoft deal for full-stack AI infrastructure (possibly rising to $19.4B). - Active include many enterprises such as MSFT, Shopify, Accenture, governments, and AI startups. - 71%+ gross margins on AI infra segment; profitable at the EBITDA level. - 1 GW+ powr secured, targeting dense GPU deployments across new sites. - Fully integrated software+hardware stack, which increases opex and is a moat. - $10B+ in assets, from $5.8B+ cash, ownership of companies like Clickhouse that powers Anthropic, Meta, and others growing rapidly. Negatives ________________ - Execution risks around full-stack delivery and latency SLAs could derail rollout. - Microsoft is the primary anchor on forward revenue extreme contract dependency (high concentration risk projected AI revenue) - 71% gross margin figure not representative of forward revenue or execution at scale and could lower to a more conservative 50-65% number. $IREN - $16.52B marketcap 🍏Positives ________________ - 2.91 GW power secured, diversified across North America, and expanding more than 3GW in renewable power capacity. - Full control over power, land, and data center construction = vertical efficiency and higher margin control compare to other bare metal. - Targets customers (e.g., AI labs, hyperscalers) who might bring their own Type-1 orchestration and prefer bare-metal control. - Historically the most efficient and profitable large-scale BTC miner and better positioned to handle rising GPU power/cooling requirements in B100/B200 generations for AI cloud pivot. - Targeting an ARR of over $500 million by the first quarter of 2026 by buildouting out a fleet of 23,000 NVIDIA B200/B300 + AMD MI350X GPUs. Downsides ________________ - No well-known enteprsie/hyperscaler contract visibility and limited disclosed large SLA deals yet - Not a full-stack provider; acts as Tier-1 (shell/colocation) = lower margin and tenant risk and illustrative GPU target. Meaning it is unable to capture the highest-margin revenues associated with integrated, proprietary full-stack cloud - Quoted 92% margins excludes expense recognition, and with D&A and would be much lower at scale. - The substantial capital expenditure required for the large-scale GPU purchase risks negative returns if future utilization rates fail to meet expectations $APLD 🍏Positives ________________ - $5B Macquarie financing facility, with first draw building 400MW AI campus. - Tenant: CoreWeave fully leasing Polaris Forge 1 (400MW), scalable to 1GW. - Proprietary waterless liquid cooling optimized for dense AI workloads - Capex-light via preferred equity. - AI-first buildout focused on latency, fiber, and power pairing in North Dakota. Downsides ________________ - Large-scale buildout depends on continued draws from Macquarie’s $5B facility. - 400MW CoreWeave lease is all-or-nothing; no clarity on tenant diversity. - Execution risk at Polaris Forge: weather, zoning, grid integration delays. - Exposed to single-tenant concentration and private-market GPU pricing dynamics. $RIOT 🍏Positives ________________ - 600MW idle capacity in Corsicana, TX under AI redeployment review. - Gigawatt-scale infrastructure in place, ready for rapid pivot. - Leverages one of the lowest power costs in the U.S. (ERCOT access). - TTM Gross Margin of nearly 60% as of Q2 2025, primarily derived from its core mining business. The successful implementation of this contract gives more confidence in converting additional existing mining sites into high-multiple HPC capacity. Negatives ________________ - AI pivot is purely exploratory; no firm buildout announced. - 600MW idle capacity = opportunity, but also dead capital for now. - Canceling mining expansion (8.3 EH/s) may weaken near-term revenues. - Needs permits, fiber, GPU supply, and anchor tenant before AI rollout can begin. $CIFR 🍏Positives ________________ - 10-year, $3B+ computing 168MW agreement with Fluidstack + $GOOGL - Agreement is substantially de-risked by a $1.4 billion commitment from Google to backstop the Fluidstack lease - High-margin, low-risk hosting model with leased facilities. Negatives ________________ - Fluidstack deal is 10-year but fixed-price, limiting upside if GPU rental rates increase. - $1.3B convertible debt adds dilution risk if equity markets weaken. - Site buildouts dependent on few customers, very high fill and utilization risk. _ Since the data center sector is so nuanced and broad thought it would be helpful if I wrote created a full TLDR list for newcomers to see the tradeoffs of each approach. I'll probably go back and add more stuff to $CIFR and $APLD and others in one final post since it didn't get enough justice. I just didn't realize how much time this took to write up + please correct me if any figures are off lol. Also would save me time if people helped fill out the rest lol so would appreciate it too for 2/2. ~~~ Help with the Crowdsourcing ~~~