Cadence Design Systems: A Great Business Waiting on a Better Price
Summary
Cadence Design Systems is one of those rare businesses where the quality question answers itself and the price question answers itself too — in opposite directions. The company sits in an effective duopoly selling tools that virtually every advanced chip designer cannot function without. The moat is real and widening. The price, however, already assumes the next decade goes perfectly. The judgment here is watch: a fine business held on the list for a very different market day.
Gate 1 — Circle of Competence
Cadence sells electronic design automation (EDA) software, design IP, and verification hardware (emulation and prototyping) that semiconductor and electronics companies use to design chips. Revenue is anchored in multi-year subscription licenses; roughly 85% of revenue is recurring. FY2025 revenue was $5.30 billion ($4.82 billion product and maintenance, $0.48 billion services), with period-end backlog of $7.8 billion. Every advanced chip designer — NVIDIA, Apple, Qualcomm among them — relies on Cadence or Synopsys tooling. Once a team commits to a tool chain for a multi-year chip project, switching is not seriously entertained; renewal rates are effectively near 100%.
Two questions determine knowability. First, can the information be obtained? Yes — filings, CFO commentary, earnings transcripts, and industry research fully reconstruct the model and the competitive picture (a three-player field with Synopsys and Siemens EDA, effectively a duopoly at the top). Second, can the future be described without betting on specifics? Yes. Chips will still need to be designed in ten years, and the tools will come from the same two vendors. Rising chip complexity — finer process nodes, chiplets, proliferating AI accelerators — structurally increases EDA spending. Cadence is a picks-and-shovels position: one bets on total design activity, not on which chip wins. That sits comfortably within the knowable zone: software and developer tools with lock-in structures observable in public data, on platforms with established switching costs.
Gate 1: Pass.
Gate 2 — Moat
The moat has three load-bearing walls. Switching costs are the core: engineering teams spend years mastering a tool flow, and migrating an in-flight, multi-billion-dollar chip project to a competitor’s tools is effectively impossible on cost, schedule, and error risk. Intangibles follow: a tool and IP portfolio accumulated over three decades, plus certified, co-optimized relationships with foundry processes. Efficient scale completes the picture — the EDA market is split between Synopsys (~38%) and Cadence (~36%), with R&D requirements relative to market size that structurally block new entrants.
Pricing power shows up in the numbers rather than the narrative. In a structure with 85% recurring revenue and near-100% renewals, price increases stick at contract renewal. FY2025 revenue grew 14%, and FY2026 guidance calls for roughly 17% growth — driven by customers spending more per contract, not by a surge in customer count. Backlog of $7.8 billion is a record. Net margins above 20% (non-GAAP margins in the 40s) have held for years — the quantitative shadow of a moat.
The direction is widening. Harder chip design (3nm and below, 3D-IC, chiplets) and the spread of in-house chip design at systems companies — automakers, hyperscalers — simultaneously expand the customer base and tool dependence. Both duopoly members are pushing into adjacent simulation and system-design territory: Synopsys via its Ansys acquisition, Cadence via the $3.16 billion Hexagon D&E acquisition expected to close in early 2026.
AI cuts both ways, though the evidence currently favors reinforcement. The AI chip boom raises EDA spending regardless of which designer wins. AI-driven design automation (Cerebrus and the like) is only buildable by incumbents holding vast design data — a data flywheel. The long-tail risk is that LLM-based design agents eventually bypass or commoditize the tool chain itself. For now, AI features are add-ons sold on top of existing tools, and the reinforcement case is stronger.
Gate 2: Pass.
Gate 3 — Management
CEO Anirudh Devgan, an engineer by training, has run the company since 2021. Capital return policy targets 50%+ of free cash flow to shareholders; FY2025 buybacks were roughly $928 million, with no dividend. M&A has favored adjacent bolt-ons (OpenEye, BETA CAE), though Hexagon D&E at $3.16 billion is the largest deal to date — integration results bear watching. Debt is modest: $2.5 billion of senior notes issued September 2024 (4.2–4.7% coupons) against roughly $5–6 billion of equity, a D/E around 0.4–0.5, comfortably serviced.
The retained-earnings test passes decisively: over five years, market value added per dollar of retained earnings has run well above $1 (stock up more than 3x; five-year average ROE of 25.6% with a standard deviation of 4.9 points — consistency, not luck). One caveat on buybacks: they were executed at P/E levels of 70–90, well short of a below-intrinsic-value standard. These look like dilution defense against stock-based compensation (4–6% of revenue); share count has stayed roughly flat.
The deduction comes from candor. Cadence pleaded guilty to one count of conspiracy and settled with DOJ and BIS for $140.6 million in July 2025, over roughly $45 million of unauthorized exports to China’s NUDT (a sanctioned institution) between 2015 and 2021. The violations are a legacy of a prior regime and were resolved through self-disclosure and compliance remediation — but under a standard of honesty, it is not spotless. A recurrence elevates this to an outright Gate 3 failure.
Gate 3: Pass, with reservations (Hexagon integration results; any compliance recurrence).
Gate 4 — Price
Screening-derived owner earnings were $1.195 billion for FY2025 — a figure that includes the one-time $140.6 million penalty within net income of roughly $1.11 billion. Adding back the one-off and normalizing against FY2026 guidance (revenue of $6.125–6.225 billion, ~17% growth) gives a generous $1.5–1.7 billion.
Apply the 15x multiple reserved for certain growers: $22.5–25.5 billion of intrinsic value. Stretch to 20x in deference to duopoly certainty: $30–34 billion. The market cap of $102.5 billion is more than three times the most generous estimate. At ~$380 per share (P/E of 86.4), the owner-earnings yield is 1.2% — less than a third of the ~4.3% ten-year Treasury yield. There is no margin of safety. The “AI picks-and-shovels” premium has pre-paid a decade of high growth into the price; when the market has already priced in your thesis, the price gate rejects you anyway.
A target buying range — normalized owner earnings of $1.6 billion at 15x, plus a growth option — lands around $35–40 billion of market cap, or roughly $130–150 per share. That requires a decline of 60%+, realistically reachable only in a broad market break. Which is precisely why the position is kept on the watch list.
Gate 4: Fail (excellent business, insufficient price).
The Verdict
Watch. Gates 1 through 3 pass with real conviction — a knowable picks-and-shovels duopoly, a widening switching-cost moat, near-100% renewals, record backlog, and a retained-earnings test passed by a wide margin. Gate 4 fails plainly: at an owner-earnings yield of 1.2% against a 4.3% risk-free rate, the market has already paid for the future. The stock stays on the list for the collapse scenario, not the current one.
What would change it: a market-cap reset to the $35–40 billion range (~$130–150 per share, an owner-earnings yield of 4% or better) would move this from watch to a genuine candidate — or, on the downside of the thesis, sustained backlog declines across two consecutive quarters, erosion of the recurring-revenue share, a major loss of China exposure on top of its current ~12–13% of revenue, or any repeat of an export-compliance violation, which would fail Gate 3 outright.
This analysis is AI-generated, educational, and not investment advice. Figures may contain errors or be delayed. Disclaimer