Cadence Design Systems
Recurring software (~80%+ of revenue is recurring time-based licenses + maintenance) sold to chip and systems designers, plus semiconductor IP royalties/licenses and emulation/prototyping hardware (Palladium/Protium); high-margin, sticky, backlog-driven.
The thesis on this name
State of AI Compute
Highest-quality EDA franchise (88% GM, every segment double-digit, $8B record backlog, agentic-AI revenue model) — the best-calibrated, verify-confirmed lean in the EDA shift; own it, but the ~45x multiple already embeds the AI-design tailwind so accumulate on de-rating.
State of AI Compute
Long CDNS over 18-36 months on a structural-edge thesis: the EDA duopoly is a low-elasticity toll on every AI/HPC tape-out with switching costs that compound, and the market under-prices backlog durability (record $8B, $4B converting <12mo) and agentic-AI/Reality-DT optionality — but the ~46x forward multiple already…
State of AI Compute
Short leg of the EDA relative-value pair: CDNS trades ~9 turns richer than SNPS for the same duopoly; the pair harvests the spread and neutralizes the AI-capex factor. Do NOT hold CDNS as a naked long to 6% — the standa…
Earnings, margins, COGS & capex
Cadence compounds revenue mid-teens with software-grade margins: FY25 revenue $5.297B (+14%, fact) accelerated to +19% in Q1 FY26 ($1.474B, fact) as every segment grew double digits, with non-GAAP operating margin expanding to 44.7% (fact). The model is ~80%+ recurring, underwritten by a record $8.0B backlog (fact) of which ~$4.0B converts within 12 months (fact). FCF is strong (~$1.59B FY25, 20.8% Q1 FY26 margin) and ~50% is returned via buybacks (fact).
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~15¢ is cost of goods and ~39¢ operating expense, leaving ~46¢ of operating profit (~23¢ net).
Revenue trend
Margins
flat/up
up (from 41.7% Q1 FY25)
up slightly (from 29.1%)
up
seasonally low Q1; FY ~30%
COGS structure
COGS is low because the core product is software: cost of product/maintenance is mostly amortization of acquired/internal IP, cloud/compute for AI-driven tools and on-cloud EDA, and customer support. The drag on blended gross margin is the hardware business (Palladium emulation + Protium prototyping systems), which carries BOM/manufacturing cost and prints lower gross margin than software — so a hardware-heavy quarter optically compresses gross margin even as it grows. Stock-based comp is the main GAAP-vs-non-GAAP gap.
Capex
Capex is light (~3–4% of revenue) — funds R&D facilities, the internal compute/datacenter footprint for AI-driven and cloud EDA, and lab/test equipment for the hardware systems business. Not a capital-intensive model; cash generation is the story, not reinvestment in PP&E.
Latest earnings
Beat. Against its own guidance the company states it 'exceeded Q2 revenue, non-GAAP operating margin and EPS guidance'. Non-GAAP EPS $2.11 (vs $1.65 a year ago), GAAP EPS $1.33 (vs $0.59), revenue $1.584B (+24%). Sell-side aggregators DISAGREE on the top line: Zacks scores revenue as a 0.5% consensus beat and non-GAAP EPS a 2.9% beat, while MarketBeat/Quiver put consensus at ~$1.608B and call it a ~1.5% revenue shortfall. We do not resolve the conflict; the guidance beat is the clean fact. The stock closed +0.4% over the following four sessions and remains ~10% below its late-June level.
Raised for the second consecutive quarter. FY26: revenue $6.26-6.34B (from $6.125-6.225B), non-GAAP op margin 43.75-44.75% (from 43.5-44.5%), non-GAAP EPS $8.05-8.15 (from $7.85-7.95), GAAP EPS $4.76-4.86 (from $4.39-4.49), OCF ~$2.0B (from $1.875-1.975B), capex ~$240M. Q3 FY26: revenue $1.595-1.625B (+19-21% YoY), non-GAAP op margin 43.5-44.5%, non-GAAP EPS $2.01-2.07, GAAP EPS $1.11-1.17. ~$200M buyback planned in Q3; ~50% of FY free cash flow to buybacks.
- Q2 revenue
- $1,584.5M (+24.2% YoY; +7.5% sequential)
- Backlog
- $8.1B record (from $7.8B at FY25 year-end); 12-month RPO $4.2B
- Non-GAAP operating margin
- 45.5% (+270bps YoY); FY26 guided 43.75-44.75%, midpoint below FY25's 44.6%
- GAAP / non-GAAP gross margin
- 84.9% / 88.2% (disclosed, not estimated)
- Q2 free cash flow
- $582M (36.7% margin); OCF $635M, capex $53M
- Recurring revenue
- 78% of Q2 revenue; 79% TTM
- China mix
- 15% of Q2 revenue, up from 9% in Q2'25
- Product mix
- Core EDA 68% / IP 15% / SDA 17% (from 71% / 13% / 16%)
- FY26 non-GAAP EPS growth
- +13-14% guided, versus +18-20% revenue growth and +20% delivered in FY25
- Net debt
- ~$1.06B ($2,500M principal less $1,440M cash)
Growth drivers
- AI/HPC design intensity — every accelerator, custom-silicon (hyperscaler ASIC), and chiplet program drives more tool seats and longer flows; Core EDA grew ~18% in Q1 FY26 (fact)
- Semiconductor IP — fastest grower; IP up ~22–25% on HBM, UCIe (chiplets), PCIe, DDR/LPDDR6, SerDes for AI workloads (fact)
- System Design & Analysis (SDA) — multiphysics/CFD via the OpenEVA/multiphysics platform and the Ansys-vacated mind-share; grew ~35% YoY in Q4 FY25 (fact)
- Emulation/prototyping hardware — Palladium Z3 + Protium X3 at record levels as pre-silicon verification of giant AI SoCs explodes (fact)
- Agentic AI products (Cerebrus, Verisium, the Cadence.AI/JedAI stack) lifting price/value per seat and customer productivity (fact)
- Design-activity expansion — more design starts, more nodes (N3/N2/A16/A14 with TSMC), and physical-AI/robotics/digital-twin TAM via NVIDIA partnership (fact)
Reported financials — SEC EDGAR
Audited GAAP figures pulled from SEC filings · latest filing 2026-02-19. The audited primary-source spine — not financial advice.
Revenue — annual (GAAP)
Margins & balance sheet — FY’25
Bull & bear
The toll-booth read got stronger, not weaker: revenue accelerated to +24%, backlog set another record at $8.1B, and management raised the full year for the second straight quarter - while the stock de-rated ~10%, taking the forward multiple from ~47x to ~42x. You are paying less for a faster-growing, more visible business than you were six weeks ago.
- Demand accelerated rather than digested: Q2 revenue +24.2% YoY versus +19% in Q1, with the FY26 guide raised twice (to $6.26-6.34B, +18-20%) - the AI-design tailwind is showing up in revenue, not just commentary.
- Visibility deepened: record $8.1B backlog (from $7.8B at year-end), $4.2B of RPO converting inside 12 months, and 79% TTM recurring revenue - estimates are unusually de-risked for a name priced as a growth stock.
- Cash conversion is exceptional and improving: Q2 FCF $582M on $1,584.5M revenue is a 36.7% margin, OCF guided to ~$2.0B for FY26, and ~50% of FCF returns via buyback ($400M in 1H26, ~$200M planned in Q3).
- The multiple did the de-rating the old bear case warned about, without the fundamentals breaking: ~42x FY26 non-GAAP EPS versus ~47x six weeks ago, on numbers that went up rather than down. The valuation risk has been partly discharged.
- Hexagon D&E ($3.1B, closed Feb 2026) is a real TAM extension, not financial engineering: SDA is now 17% of revenue (from 16%), and the balance sheet absorbed it while ending the quarter at only ~$1.06B net debt.
Still expensive at ~42x forward, and this quarter quietly broke the cleanest part of the old bull case: for FY26 Cadence guides EPS to grow SLOWER than revenue (+13-14% vs +18-20%) with non-GAAP operating margin below FY25's actual. Growth is increasingly bought rather than compounded, and China - the named tail risk - is now a bigger share of revenue than before.
- The margin-expansion pillar has inverted for the full year: FY26 non-GAAP operating margin is guided to a 44.25% midpoint, BELOW FY25's actual 44.6%, and the Q3 guide of 43.5-44.5% compares against 47.6% actually delivered in Q3'25.
- EPS no longer compounds faster than revenue: FY26 non-GAAP EPS growth of +13-14% trails +18-20% revenue growth, versus +20% EPS growth in FY25. Non-GAAP operating expense is guided ~+20% for the year against ~+10% in FY25 - the Hexagon integration is eating the operating leverage.
- China rose to 15% of Q2 revenue from 9% a year earlier (13% in FY25). The export-control tail risk the thesis has always carried now sits on a materially larger revenue base, and it is binary rather than gradual.
- Growth is measurably more inorganic: the SDA mix gain and a large slice of the revenue acceleration trace to Hexagon D&E, headcount is up 17% YoY to 15,445, and goodwill nearly doubled to $4.91B - integration risk is now a live line item rather than a hypothetical.
- At ~42x forward non-GAAP and ~68x trailing GAAP EPS, a single soft guide still re-rates the stock 25-30% regardless of business quality; the July move (-10% from late June through a beat-and-raise) is a live demonstration that the multiple, not the fundamentals, sets the return.
- Capex intensity is climbing (from 2.7% of revenue in FY25 to a guided ~3.8% in FY26) and only ~16% of cash sits in the U.S., modestly constraining the flexibility of the buyback that underwrites the EPS story.
What it is worth
Premium-growth multiple cross-checked with a reverse-DCF: anchor on forward non-GAAP EPS (FY26 guide ~$7.90, fact) and a peer multiple vs Synopsys, then sanity-test what the price implies.
~$250–$290
a China shock or AI-capex digestion slows growth to low-double-digits and the multiple de-rates to ~30–35x forward (the dominant near-term risk is the re-rating, not the fundamentals).
~$390–$420
mid-teens growth, ~44% op margin, multiple drifts to ~45x forward as growth premium normalizes; EPS compounding carries the return.
~$460–$500
~17%+ growth sustained, op margin pushing 45%+, multiple holds ~50x forward on AI-design durability and SDA share gains.
At ~$373 / ~$102B cap, CDNS trades ~47x FY26 non-GAAP EPS (~76x trailing GAAP) — the current price implies sustained mid-teens revenue growth with continued margin expansion for years; reasonable if the AI-design tailwind + backlog hold, expensive if growth normalizes toward low-double-digits. (estimate, not advice)
SWOT
Strengths
- Half of an entrenched EDA duopoly (~30% share with Synopsys ~31%, Siemens ~13%; Big-3 >85% combined) — a near-mandatory toll on advanced chip design
- ~80%+ recurring revenue with a record $8.0B backlog (~$4.0B converting <12mo), giving rare revenue visibility for a tech name
- Software-grade economics — 44.7% non-GAAP operating margin, ~30% FCF margin, light capex, net-debt-light balance sheet
- Deep foundry/process certification (TSMC N3/N2/A16/A14) and full-flow tool breadth that make rip-and-replace impractical
- Fast-growing adjacencies (IP +22%, SDA/multiphysics +35%) diversifying beyond pure digital EDA
Weaknesses
- Premium valuation (~47x forward / ~76x trailing P/E) leaves no room for a stumble; multiple compression is the dominant near-term risk
- Hardware (Palladium/Protium) is lumpier and lower-margin, adding quarter-to-quarter optical volatility to gross margin and bookings
- Concentrated end-market — ~all revenue rides semiconductor R&D budgets, which are cyclical even if Cadence is less so than chipmakers
- China exposure (~12% of revenue, $573M FY24) sits in a live geopolitical/export-control crosshair
- Recent $140M DOJ/BIS export-control settlement (incl. $95M BIS penalty, Jul 2025) — a compliance/reputational blemish
Opportunities
- Agentic AI in design (Cerebrus, Verisium, Cadence.AI/JedAI) — monetize productivity gains as higher value-per-seat, expanding the toll
- System-level/multiphysics + digital-twin TAM via the NVIDIA partnership (physics-based simulation, AI factories, robotics/physical AI)
- Capture share/mind from Ansys-adjacent multiphysics as Synopsys digests its $35B Ansys integration
- Chiplet/UCIe + advanced-packaging design explosion lifts both EDA tool flows and IP attach
- Custom-silicon proliferation (hyperscaler ASICs, sovereign AI) multiplies design starts and tool seats
Threats
- Synopsys + Ansys creating a vertically integrated device-to-system stack that out-bundles Cadence in multiphysics
- Export-control escalation could re-restrict or permanently shrink the China market
- Semiconductor R&D-spend downturn (macro / AI-capex digestion) would slow bookings
- Multiple de-rating if AI-capex narrative cools — the stock is priced for sustained mid-teens+ compounding
- Long-tail: customer in-housing of AI-assisted design or open-source EDA eroding the low-end (low probability near-term)
Moats, dependencies & bottlenecks
Moats
very high (decades) Tools are embedded in customer design flows, methodologies, and engineer training; mid-design rip-and-replace risks tape-out — switching costs compound the longer a customer stays.
Deep co-development with TSMC/Samsung/Intel on N3/N2/A16/A14 means Cadence flows are certified for the leading edge before customers can use them — a structural barrier to new entrants.
Digital + analog + verification + IP + hardware + multiphysics under one roof lets Cadence bundle and cross-sell; IP (HBM/UCIe/PCIe/DDR) rides every new SoC.
Multi-year time-based licenses + $8.0B backlog create revenue inertia, not a true competitive moat, but materially raise the cost of customer churn.
narrow (emerging) Cerebrus/Verisium/JedAI improve with design data and reinforce stickiness, but the durability as a defensible moat (vs Synopsys' equivalent) is still being proven.
Dependencies
Revenue tracks customers' chip-design budgets; less cyclical than chip sales (R&D is protected) but not immune to a deep downturn or AI-capex digestion.
Tool relevance depends on certifying flows for each new node; loss of foundry co-development access would erode the leading-edge moat.
~12% of revenue ($573M FY24); the May–Jul 2025 license whipsaw + $140M settlement show binary, fast-moving policy risk.
NVIDIA, the hyperscaler ASIC programs, and merchant chipmakers are the marginal demand driver; concentration of AI design starts in a few buyers raises cyclicality if they pause.
EDA is a deep-IP people business; SBC is high partly to retain scarce algorithm/AI talent against Synopsys and Big Tech poaching.
Advantages
- Mission-critical, mandatory product — chips literally cannot tape out without EDA, giving low price elasticity and a tollbooth position.
- Software economics: ~86–88% gross margin, 44.7% non-GAAP operating margin, ~30% FCF margin, light capex.
- Record $8.0B backlog + ~80% recurring revenue = best-in-class revenue visibility for a high-growth tech name.
- Diversified across Core EDA, IP, SDA/multiphysics, and hardware — multiple double-digit growth vectors, not one.
- Entrenched with every leading foundry and the AI-chip ecosystem (TSMC, NVIDIA, Google), structurally hard to displace.
Weaknesses
- Valuation (~47x forward / ~76x trailing P/E) prices in perfection; de-rating risk dominates the return distribution near-term.
- Large GAAP-vs-non-GAAP gap from stock-based comp ($1.96 non-GAAP vs $1.23 GAAP in Q1 FY26) flatters headline profitability.
- China exposure (~12%) under live export-control and a fresh $140M compliance settlement.
- Hardware revenue is lower-margin and lumpier, adding quarterly noise.
- Single-end-market concentration (semiconductor R&D) — diversification within semis, but not across industries.
- Direct, equally entrenched competitor (Synopsys) now larger post-Ansys, contesting the same adjacencies.
Bottlenecks
- Premium valuation caps near-term upside and amplifies drawdown risk on any deceleration — the multiple is the governor, not the business.
- China/export-control overhang can re-restrict ~12% of revenue with little warning, capping the addressable market.
- Hardware (Palladium/Protium) supply/BOM and lumpy ordering cadence inject gross-margin and bookings volatility quarter to quarter.
- Scarce EDA/AI R&D talent — innovation pace is gated by hiring against Synopsys and hyperscalers.
- Synopsys+Ansys bundling pressure in the multiphysics/SDA adjacency Cadence most wants to expand into.
Top signals & trends
Top signals
The single best read on durability; sustained record backlog validates the under-priced-visibility thesis. A flattening backlog would be the first crack.
Any re-restriction or tightening immediately threatens ~12% of revenue; the May–Jul 2025 whipsaw is the template to watch.
Continued expansion confirms AI-tool pricing power and operating leverage; a stall would question the agentic-AI monetization story.
These adjacencies are the diversification and the Synopsys/Ansys contest zone; deceleration would narrow the bull case to core EDA.
Integration distraction is a near-term opening for Cadence; a smooth, well-bundled combined stack is a medium-term competitive threat.
Cadence's demand is a derivative of the AI buildout; any hyperscaler capex guide-down is an early warning for bookings.
Trends
Each new chip class multiplies tool seats, flow length, and IP attach — the core structural tailwind.
Shifts EDA from licensed tools to AI-driven productivity Cadence can price for — value-per-seat expansion.
Disaggregation creates new design and verification flows plus high-value IP (UCIe, HBM, SerDes) Cadence sells.
Structurally caps and destabilizes ~12% of revenue; ongoing bifurcation of the global semis market.
Reinforces the duopoly (good for pricing) but creates a larger, vertically integrated rival in multiphysics (bad for share in SDA).
Via the NVIDIA partnership, extends Cadence's TAM from chip EDA to physical-AI, simulation, and AI-factory design.
Ecosystem & competitor graph
Suppliers feed the company; customers pull from it. Line thickness shows the strength of each tie (supply-chain dependency, customer earnings contribution). Hover to isolate a tie.
Critical foundry partner — co-develops and certifies Cadence flows for N3/N2/A16/A14; the leading-edge process partner that makes the tools relevant.
Foundry partner certifying Cadence flows on its advanced nodes; second-source leading-edge ecosystem.
Foundry + IFS partner for advanced-node and packaging flow certification.
Supplier of GPUs/accelerated computing that power Cadence's AI-driven and on-cloud EDA, plus a deep co-development partner.
Google Cloud, Microsoft Azure) Compute supplier for Cadence's cloud EDA and AI workloads; Google is also a named design partner.
Supply boards/components for Palladium emulation and Protium prototyping systems — the only material physical BOM in the model.
Flagship AI-chip customer + partner; designs accelerators and physical-AI systems on Cadence flows.
Amazon, Microsoft, Meta) Custom-silicon (TPU/Trainium/Maia/MTIA-class) design teams are a fast-growing, high-value customer cohort.
CPU/GPU/accelerator designer; major EDA + IP consumer.
Custom-ASIC and networking-silicon leader; heavy EDA/IP/emulation user.
Mobile/automotive/compute SoC designer; large recurring EDA + IP customer.
Leading SoC designers across mobile, PC, and consumer — core recurring license + hardware customers.
The other half of the duopoly (~31% share) and now larger after the $35B Ansys deal — the direct, equally entrenched rival across digital EDA, IP, and (now) multiphysics.
#3 EDA player (~13% share), strong in verification, PCB, and IC packaging; part of Siemens Digital Industries Software — credible but sub-scale vs the top two.
Former independent multiphysics/CFD leader; its absorption into Synopsys is the direct competitive threat to Cadence's fast-growing SDA segment.
RF/microwave and system-level design + test; niche overlap in high-frequency/system simulation, not a full-flow threat.
Competes in semiconductor IP (CPU/GPU cores) where Cadence sells interface/foundation IP; partner in some contexts, competitor in IP attach.
OpenROAD and hyperscaler in-house flows nibble at the low end; not a near-term threat to leading-edge commercial EDA but a long-tail watch item.