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 | Duolingo | NVIDIA | |
|---|---|---|---|
| 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 Wide is ruled out by the company's own Item 1A, which concedes "low switching costs" and that "Low barriers to entry allow start-up companies with lower costs and less pressure for profitability to compete with us." It sits well above none because Q4 DAUs grew 30% to 52.7M, DAU/MAU rose to 39.6% from 34.7%, and marketing took only 12% of FY2025 revenue, though that ratio was flat versus FY2024 and gross margin fell to 72.2% from 72.8%. It stays narrow rather than eroding because Q1 FY2026 DAUs and paid subscribers each grew 21%. | wide Wide: the CUDA software platform and its developer installed base create high switching costs across the AI-training stack, defended by $76.7B cumulative R&D and reinforced by ~71% FY2026 gross margins and Data Center revenue (~90% of total) up 68% YoY — durable, hard-to-replicate advantages per the FY2026 10-K. |
| 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 intangible is the consumer brand, not patents: Item 1 says "For many, Duolingo has become synonymous with language learning" and that growth is "organic, primarily driven by word-of-mouth and brand buzz," while the IP section discloses only "two patent applications." Switching costs are ruled out by the filing's own "low switching costs" caption. The data-scale support is the filing's own characterisation, and the brand is admittedly damageable: an April 2025 AI memo "may have contributed to unfavorable publicity, adverse impacts on the Company's brand and social media presence, and a deceleration in user growth." | intangibles ip The durable edge rests on proprietary IP and software: the full-stack CUDA development platform running on all NVIDIA GPUs plus hundreds of proprietary domain libraries/SDKs/APIs, and $76.7B cumulative R&D yielding "inventions that are essential to modern computing" (NVIDIA invented the GPU in 1999). This is reinforced by developer-ecosystem network effects and CUDA switching costs, per the FY2026 10-K Business section. |
| 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. | clear leader Leadership holds within consumer app-based language learning, anchored on one externally checkable claim in Item 1: the app "is also the top-grossing app globally in the Education category on both Google Play and the Apple App Store," with "over 250 total language courses to more than 130 million monthly active users" as of 2025-12-31. The broader "leading market position" claim is the filing's own characterisation, and no third-party share statistic appears anywhere in the 10-K, so a band is recorded and no percentage transcribed. The band does not extend to assessment, literacy, math, music, chess or offline learning, where the filing claims no position. | clear leader FY2026 Data Center revenue was $193.7B (~90% of the $215.9B total), up 68% YoY on the Blackwell ramp, and NVIDIA describes itself as "a data center scale AI infrastructure company reshaping all industries." The 10-K frames named rivals as parties who "provide or intend to provide" GPUs/accelerators — incumbent-leader positioning. |
| 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 Every mechanism is cost, mix or volume, never price. FY2025 gross margin fell to 72.2% from 72.8% on "increased AI costs used in features like Video Call" and an advertising mix shift; the Q1 FY2026 gain to 73.0% came from "continued reductions in per-unit AI costs," with margin guided to about 71.0% in Q2 and "approximately 69.0% by Q4." The filing concedes it "may not be able to fully offset such higher costs through price increases." Subscription revenue rose 44% to $873.4 million "primarily due to an increase in the average number of paid subscribers," and the cap is self-imposed: "We intentionally do not put our learning content behind a paywall." | strong FY2026 gross margin was 71.1% (75.0% in FY2025); per the 10-K MD&A the ~3.9pt decline reflects the Hopper HGX→Blackwell full-system mix shift and a one-time $4.5B H20 excess-inventory/purchase-obligation charge, not competitive price erosion. A low-70s% hardware gross margin evidences strong pricing power. |
| 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. | Duolingo's moat is essentially one asset — a consumer brand strong enough that the filing claims people "search for the term 'Duolingo' much more often than 'learn Spanish'" — sitting on real but not defensive scale. It converts into economics through organic acquisition: sales and marketing took 12% of FY2025 revenue while Q4 DAUs grew 30% to 52.7 million and DAU/MAU rose to 39.6% from 34.7%. It is not wide, because the company supplies both disqualifiers itself: "low switching costs" and "Low barriers to entry allow start-up companies with lower costs and less pressure for profitability to compete with us." Width and pricing power diverge: Duolingo will not price against its own free product, grows subscription revenue by adding subscribers ($873.4 million, up 44%), and guides gross margin down to roughly 69% by Q4 FY2026 "as AI feature use expands in our products." | NVIDIA pairs market-leading accelerated-computing hardware (the Blackwell data-center platform) with a proprietary full-stack software moat — CUDA plus hundreds of domain libraries — funded by $76.7B of cumulative R&D, and its "large and growing number of developers and installed base... strengthens our ecosystem and increases the value of our platform for our customers" (FY2026 10-K). Competition is intensifying from AMD, Intel and Huawei, and from hyperscalers (Alphabet, Amazon, Microsoft) designing internal AI silicon, but rivals must overcome NVIDIA's entrenched CUDA software ecosystem to displace it. |
| Chain position | Layer-4 wafer-scale accelerators plus an inference cloud — hardware vendor and specialized cloud in one node. | A consumer AI application and net buyer of model inference. The filing ties its margin line directly to inference cost, attributing the FY2025 subscription gross-margin decline to "increased AI costs used in features like Video Call" and guiding margin to approximately 69.0% by Q4 FY2026 "as AI feature use expands," while the Q1 FY2026 beat came from "reductions in per-unit AI costs." | Upstream compute-platform supplier: NVIDIA sells full-stack data-center systems (GPU + Arm CPU + DPU + NVLink/InfiniBand networking + CUDA software) to cloud providers and enterprises, and relies on third-party foundry/assembly-test-packaging partners (e.g., SPIL, Amkor, Wistron, Fabrinet). Per the FY2026 10-K, several of its largest customers (hyperscalers such as Amazon, Alphabet, Microsoft) are simultaneously customers and emerging competitors developing internal accelerated-computing silicon. |
| Products (share / barrier) |
|
|
|
| Long-horizon vote | +0.20 at weight 0.20 · swarm bearish Editorial prior, not backtested. | +0.13 at weight 0.20 · swarm bearish Editorial prior, not backtested. | +0.42 at weight 0.20 · swarm bullish Editorial prior, not backtested. |