Compare moats
Up to three covered companies, band by band. Every call is a curated editorial judgment, never a disclosed figure — and every band carries its cited basis.
| Cerebras Systems | SanDisk | Powell Industries, Inc. | |
|---|---|---|---|
| Moat rating | narrow 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. | narrow Narrow rather than wide: the FY2025 10-K Competition section opens 'Our industry is highly competitive' and names five vertically integrated flash rivals (Kioxia, Micron, Samsung, SK Hynix, Yangtze Memory), while Item 1A concedes SanDisk's products 'are designed to be largely interchangeable with competitors' products' in a market 'often subject to declining average selling prices'. It is not 'none', because the same filing discloses a genuinely durable consumer franchise and roughly 7,900 granted patents. | none The FY2025 10-K (filed 2025-11-19) says the company believes its integration capabilities, technical and project management acumen, application engineering expertise and specialty contracting experience 'give us a sustainable competitive advantage in our markets', and that 'a significant portion of our business is from repeat customers', many times involving EPC firms 'with whom we often have long, established relationships'. Neither the filing nor Powell's record shows that edge protecting returns. Item 1A warns 'Our industry is highly competitive' and says certain competitors 'may have lower cost structures' and 'may, therefore, be able to provide their products or services at lower prices'. Item 1 says products are 'typically awarded in competitive bid situations' and that projects 'are typically non-recurring.' Stored fundamentals from Powell's filings show operating margin between -4.9% and 7.1% in every fiscal year from FY2011 to FY2022, with operating losses in FY2017 (-$19.3M on $395.9M of revenue) and FY2018 (-$9.1M on $448.7M). Operating margin has since risen to 19.7% in FY2025 ($217.9M on $1,104.3M), but the 10-K credits the gross-margin gain to 'favorable volume leverage and strong project execution in a stable pricing environment', not to a competitive edge. A self-described advantage that left operating margin no higher than 7.1% in any year from FY2011 to FY2022 is not shown to be a moat, so the band is none. |
| Moat type | intangibles ip 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. | intangibles ip The only hard-to-replicate assets the filing actually claims are intangible: it reports 'a strong position in the Consumer end market' with 'significant consumer brands and franchises globally, with valuable patent portfolios containing approximately 7,900 granted patents and approximately 3,200 pending patent applications worldwide', and adds that non-patented IP, 'particularly some of our process technology, is an important factor in our success'. Cost-scale is the wrong label because the manufacturing scale sits in Flash Ventures, a 49.9%-owned JV co-owned with Kioxia, not in a proprietary fab base. | none The 10-K describes engineering know-how, not protected technology or lock-in. It says 'We consider our engineering, project management, systems integration and technical support capabilities vital to the success of our business', that 'Powell's expertise in vacuum circuit breaker engineering is internationally recognized', and that unpatented technology, including engineering technical skills and know-how, 'is important to our overall business'. It also says 'we believe that the growth of our business will depend primarily upon the quality of our products and our relationships with our customers, rather than the extent of our patent protection', so patents are not the source. Switching costs are not shown: projects are 'typically awarded in competitive bid situations' and 'typically non-recurring', even though 'a significant portion of our business is from repeat customers'. The same filing says some competitors 'are significantly larger and have substantially greater global resources such as engineering, manufacturing and marketing', so the know-how is a capability rivals also bring to each bid, not a demonstrated moat source; the type is none. |
| Leadership | clear leader 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. | at parity At parity, not ahead: the 10-K Competition section positions SanDisk against vertically integrated suppliers Kioxia, Micron, Samsung, SK Hynix and Yangtze Memory, and Item 1A concedes its products are 'designed to be largely interchangeable with competitors' products'. Its explicit leadership claims are brand- and consumer-scoped ('industry leading consumer brand awareness and global retail distribution presence'), not technology- or share-scoped. | at parity The 10-K gives no share or rank. It says Powell competes 'with a small number of multinational competitors', naming ABB, Eaton, Schneider and Siemens Industries, and that some are 'significantly larger and have substantially greater global resources'. It adds that bid evaluation varies 'from project to project', covering technical support, engineering and manufacturing capability, equipment rating, delivered value, scheduling and price. Powell therefore competes bid by bid alongside those multinationals rather than leading the category. |
| Pricing power | moderate 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. | weak A price taker. The 10-K describes an industry 'often subject to declining average selling prices' and warns that competitors 'may utilize pricing strategies, including offering products at prices at or below cost, that we may be unable to competitively match'. Gross margin swung from 7.1% (FY2023) to 16.1% (FY2024) to 30.1% (FY2025) on the cycle rather than on sustained pricing, and FY2025 Consumer ASP per gigabyte still fell 7% 'due to pricing pressure'. | weak The 10-K says gross margin rose to 29% of revenue in FY2025 from 27% in FY2024, 'due to favorable volume leverage and strong project execution in a stable pricing environment'. The 10-Q for the quarter ended June 30, 2026 (https://www.sec.gov/Archives/edgar/data/80420/000008042026000107/powl-20260630.htm) gives the same reason for 30% in the first nine months of FY2026, so neither filing credits price. Before this upcycle, stored fundamentals from Powell's filings show gross margin between 12.8% and 21.6% in every fiscal year from FY2011 to FY2022 (FY2022: $85.0M of gross profit on $532.6M of revenue, about 16%), which fits cyclical volume leverage more than pricing power. The limits are explicit: the 10-K says 'Fixed-price contracts can limit our ability to pass these increases to our customers', and Item 1A says certain competitors 'may, therefore, be able to provide their products or services at lower prices'. The one offset is cost pass-through, not price-setting: the 10-K says 'we believe we can manage this volatility through contract pricing adjustments'. Pricing power is rated weak. |
| Summary | Cerebras bet that AI compute should be one enormous chip rather than many small ones — 'quantities of compute and memory never before assembled on a single commercial chip,' avoiding 'the latency and the power-draw induced by the traditional approach,' in the prospectus's words — and is converting that bet into an inference cloud: Q2 2026 cloud revenue grew 281% with 600 MW of data center capacity under contract and $25.4 billion of remaining performance obligations. The concentration that funded the bet is the risk that remains: two related-party Gulf customers were 86% of 2025 revenue between them, with the OpenAI MRA now layered on top. | SanDisk sells largely interchangeable NAND product into a five-rival commodity market, so its defensible edge is narrow and concentrated in intangibles: the SanDisk consumer brand and retail distribution (Consumer was $2,268M of $7,355M FY2025 revenue) plus roughly 7,900 granted patents. It owns no flash fab of its own; the 10-K says substantially all of its flash memory wafers come from Flash Ventures, its 49.9% JV with Kioxia, whose fixed costs it must fund at about half regardless of the output it takes. That structure caps both the moat and the downside cushion: gross margin ran 7.1% in FY2023, 16.1% in FY2024 and 30.1% in FY2025 as the flash cycle turned. | Powell designs, builds and services custom-engineered power-distribution packages: power control rooms, E-Houses, switchgear and medium-voltage breakers. Its FY2025 10-K says the company believes its integration, project-management and application-engineering capabilities 'give us a sustainable competitive advantage in our markets', and that a significant portion of its business comes from repeat customers, often through EPC firms with which it has long relationships. The record does not show that edge protecting returns. Work is typically won in competitive bids, projects are typically non-recurring, and the 10-K names ABB, Eaton, Schneider and Siemens Industries as principal competitors, some of which it calls significantly larger. The company says it believes growth depends on product quality and customer relationships rather than patents. Stored fundamentals from Powell's filings show operating margin between -4.9% and 7.1% in every fiscal year from FY2011 to FY2022, with operating losses in FY2017 and FY2018. In the current cycle, gross margin reached 29% in FY2025 (10-K) and 30% in the first nine months of FY2026 (10-Q, quarter ended June 30, 2026, https://www.sec.gov/Archives/edgar/data/80420/000008042026000107/powl-20260630.htm). Both filings attribute that to favorable volume leverage and strong project execution 'in a stable pricing environment', not to price increases. The same 10-Q reports backlog of $2.4 billion and says diversification into electric utility and data centers 'has reduced the cyclicality of our business.' It also cautions that as data center projects become a larger component of backlog and revenues, the product mix may shift, 'as such projects will likely require less custom engineered-to-order equipment and systems than other end markets', which is the custom-engineering skill the claimed edge rests on. |
| Chain position | Layer-4 wafer-scale accelerators plus an inference cloud — hardware vendor and specialized cloud in one node. | Merchant NAND/SSD supplier with no wholly owned fab: substantially all flash wafers come from Flash Ventures, its 49.9% JV with Kioxia across seven Japanese fabs, with an eighth due to start in calendar 2025. | Supplies custom-engineered switchgear and power control rooms to oil and gas, utility and commercial/industrial projects. Its Q3 FY2026 bookings included 'a data center project valued at over $400 million' (10-Q, quarter ended June 30, 2026, https://www.sec.gov/Archives/edgar/data/80420/000008042026000107/powl-20260630.htm). |
| Products (share / barrier) |
|
|
|
| Long-horizon vote | +0.20 at weight 0.20 · swarm bearish Editorial prior, not backtested. | -0.01 at weight 0.20 · swarm bullish Editorial prior, not backtested. | -0.06 at weight 0.20 · swarm bearish Editorial prior, not backtested. |