护城河对比

The FY2026 10-K competition section describes tool-level qualification lock-in as the structural fact of the industry: 'semiconductor manufacturers must make a substantial investment to qualify and integrate new capital equipment into semiconductor production lines. As a result, once a semiconductor manufacturer has selected a particular supplier's equipment and qualified it for production, the manufacturer generally maintains that selection for that specific production application and technology node as long as the supplier's products demonstrate performance to specification in the installed base.' The filing then sizes what that incumbency is worth: of 23,232,690 thousand dollars of FY2026 revenue, 8,347,202 thousand - about 36% - was customer-support-related revenue and other, which the same document defines as 'sales of customer services, spares, upgrades, and non-leading-edge equipment' sold back into the installed base. A qualified position plus a third of revenue annuitised against it is a durable barrier rather than a cyclical one.

来源：sec.gov

The FY2025 20-F positions Nebius as 'one of the few global, at scale, multi-tenant clouds purpose built for AI,' but the same filing names its central dependency plainly: 'We currently rely on Nvidia for the GPU chips we use' — a purpose-built neocloud renting a supply the hyperscalers it competes with also control, which is a real but narrow position.

来源：sec.gov

One-of-a-kind silicon sold to very few buyers: the IPO prospectus discloses G42 at 24.0% of 2025 revenue (85.0% in 2024), MBZUAI at 62.0% of 2025 revenue, and a December 2025 MRA with OpenAI representing 'a substantial portion of our projected revenues over the next several years' — against a competitive field the same document names as NVIDIA, AMD, Intel, hyperscaler accelerators and the major clouds.

来源：sec.gov

The 10-K grounds the advantage in the cost of requalification, not in scale or patents: a customer that has qualified a competitor's tool is hard to win, and conversely Lam's own qualified positions persist for that application and node. The same section adds that the company must keep serving 'our installed base of customers through the delivery of high-quality and cost-efficient spare parts', naming the installed base as the asset being defended.

来源：sec.gov

The moat the filing describes is engineering: designs that 'optimize power and cooling efficiency, lower latency' plus 'a consistent track record of being one of the first-to-deploy the latest generation of NVIDIA GPU chips' — density and time-to-deploy at scale, not a switching lock.

来源：sec.gov

Wafer-scale integration is the moat's substance: the prospectus contrasts the WSE-3's 4 trillion transistors on 46,225 mm² of silicon with NVIDIA's B200 at 208 billion on roughly 1,600 mm² — a commercial chip category with one occupant.

来源：sec.gov

The 10-K claims position but not primacy: 'We believe we are in a strong position with our leadership and expertise in deposition, etch, and clean markets.' The same document names Applied Materials as primary competitor in deposition and Applied Materials, Hitachi and Tokyo Electron as primary competitors in etch, which is a shared-leadership structure rather than a clear single leader.

来源：sec.gov

The 20-F's own claim is 'one of the FEW global, at scale' AI clouds — a differentiated challenger to the hyperscalers it names as the competitive field, not a claimed leader of it; the first-to-deploy record is a follower's speed advantage, not category leadership.

来源：sec.gov

Scoped to the category the record supports: wafer-scale integration has no other commercial occupant per the prospectus's own comparison, and the Q2 release claims — attributed, via AMD and AWS partnerships — leadership in disaggregated inference, a technique the prospectus defines.

来源：sec.gov

Real but bounded, and the filing is explicit that the recent gain was mix rather than price. Gross margin as a percent of total revenue rose 180 basis points to 50.5% in fiscal 2026 from 48.7% in fiscal 2025 (gross margin of 11,725,308 thousand dollars on revenue of 23,232,690 thousand, against 8,979,059 on 18,435,591; 7,052,791 on 14,905,386 in fiscal 2024), and the MD&A attributes that: 'Gross margin as a percentage of revenue increased in fiscal year 2026 compared to fiscal year 2025 largely due to favorable customer mix, partially offset by aluminum and steel tariff-related spend.' Against it, the same filing says manufacturers evaluate suppliers on 'overall cost of ownership', that the spares business competes with third-party spare parts providers, and that the customer list is a handful of very large buyers (Micron, Samsung, SK hynix and TSMC named as the most significant customers).

来源：sec.gov

The filing frames a 'highly competitive' market with 'frequent introduction of new or improved solutions' and a GPU cost base set by a single supplier (NVIDIA) — a renter of compute competing on efficiency has little list-price control.

来源：sec.gov

The CFO states Q2 'significantly improved core gross and operating margins compared to a year ago' — improving, from a base where three named customers hold most of the revenue and the leverage that comes with it.

来源：sec.gov

Upstream wafer-fabrication-equipment supplier; the 10-K names Micron, Samsung, SK hynix and TSMC as its most significant customers across FY2024-FY2026, so Lam sits one step above the memory and foundry capacity that AI compute depends on.

Layer-8 neocloud — a purpose-built AI compute provider reselling NVIDIA silicon at scale.

Layer-4 wafer-scale accelerators plus an inference cloud — hardware vendor and specialized cloud in one node.

• Clean systems (EOS, DV-Prime, Da Vinci, SP Series, Coronus families) 挑战者 · 中等 来源：sec.gov
• Customer Support Business Group (spares, upgrades, service, Reliant) 领先 · 中等 来源：sec.gov
• Deposition systems (ALTUS, SABRE, VECTOR, Striker families) 前三 · 深 来源：sec.gov
• Etch systems (Kiyo, Flex, Vantex, Syndion, Argos families) 前三 · 深 来源：sec.gov

• Nebius AI cloud (GPU compute + managed AI/ML platform) 挑战者 · 中等 来源：sec.gov
• Toloka / Avride / TripleTen (non-cloud segments) 未知 · 低 来源：sec.gov

• CS wafer-scale systems (WSE-3) 领先 · 深 来源：sec.gov
• Cerebras inference cloud 挑战者 · 中等 来源：sec.gov

编辑先验，未经回测。

查看它究竟如何投票 →

编辑先验，未经回测。

查看它究竟如何投票 →

编辑先验，未经回测。

查看它究竟如何投票 →