CDNS$280.62-0.3%Cap: $77.5BP/E: 68.952w: [====|------](Mar 27)
Verdict: KEEP (borderline) — physics-gated moat and locked-in revenue provide floor; low idiosyncratic variance limits filtration value.
Factor Decomposition
Trailing 250-day regression against SPY, XLK, MTUM:
| Component | % Variance | Loading |
|---|---|---|
| XLK (tech sector) | 44.0% | +0.98 |
| SPY (market) | 17.0% | +0.55 |
| MTUM (momentum) | -15.8% | -0.47 |
| Idiosyncratic | 54.8% |
Annualized alpha = -11.8%, idio vol = 29.3%, R-squared = 45.2%.
The trailing alpha is not statistically significant. t-stat = -0.40 with standard error ≈29.3%. The 95% CI spans (-69%, +46%). Cannot reject zero. The negative alpha is noise at this sample size.
At 55% idiosyncratic variance — well below the 75% target — CDNS is substantially a tech sector proxy. In a QQQ-hedged basket, 45% of CDNS returns are captured by the short leg. The filtration alpha lives only in the 55% residual. At 0.45% portfolio weight, effective idiosyncratic exposure is 0.25% of the book.
What the Filings Show
Four things the consensus narrative obscures:
1. Convicted Felon on Export Controls
Cadence pleaded guilty in July 2025 to one count of conspiracy to commit export control violations. Between 2015 and 2021, a Cadence subsidiary made $45.3M in unauthorized sales to a Chinese customer and transferred technology to a third party in China without BIS authorization. Total penalties: $140.6M. The settlement includes ongoing audit, compliance, and reporting obligations under both a BIS administrative agreement and a DOJ plea agreement.
The penalties are paid and behind them. But CDNS is now under heightened scrutiny — any future China export tightening lands harder. China is 12-13% of revenue (≈$680M). BIS imposed ECCN 3D991/3E991 licensing requirements in May 2025, rescinded them in July 2025, then imposed a new Entity List 50% ownership rule in September 2025, suspended it in November 2025 (expires November 9, 2026). The regulatory environment for EDA exports to China remains volatile.
2. Structural Tax Rate Headwind
Effective tax rate trajectory: 19% (FY2023) to 24% (FY2024) to 27.1% (FY2025), guided ≈27% for FY2026. The dominant driver is GILTI at 8.6% of pre-tax income ($131M), reflecting the tax cost of earning through Irish and Hungarian subsidiaries. Pillar Two "did not have material impact" yet, but the jurisdictional exposure is real. This is permanent, not cyclical — each 100 basis points costs ≈$15M or ≈$0.06/share.
The FY2025 rate includes a 1.8% one-time hit from the BIS/DOJ settlement (non-deductible). Stripping that out, the underlying rate was ≈25.3%. But the FY2026 guide at 27% suggests the structural pressures (GILTI growth, declining SBC tax benefits, Hexagon integration costs) offset the one-time rolling off.
3. The Non-GAAP Gap Is Widening
FY2025 non-GAAP EPS: $7.14. GAAP EPS: $4.06. The $3.08 spread — 43% of non-GAAP — is predominantly stock-based compensation ($455M, or 8.6% of revenue) plus acquisition-related amortization.
SBC as a percentage of revenue has grown every year: 7.6% (FY2022) to 8.0% (FY2023) to 8.4% (FY2024) to 8.6% (FY2025). GAAP operating margins are declining (30.6% to 29.1% to 28.2%) while non-GAAP margins expand (to 44.6% and guided 45%+). Management-granted market-based equity awards to senior leadership are accelerating the divergence.
GAAP net income growth tells a different story than the headline: +22.6% (FY2023), +1.4% (FY2024), +5.1% (FY2025). On GAAP, this is a mid-single-digit earnings grower trading at 69x trailing.
4. Balance Sheet Regime Change
Hexagon D&E closed February 23, 2026 for ≈$2.9B (EUR 2.70B): ≈$2.04B cash plus 3.2M shares (1.2% dilution). CDNS shifts from net cash +$510M to approximately $1.5B net debt. The acquisition was priced at 14.5x revenue for ≈$200M annualized (MSC Nastran, Adams) — expensive for CAE/simulation. Not in FY2026 guidance; partial-year revenue of ≈$150M would be incremental. Buybacks of ≈$1B (50% of guided FCF) roughly offset the share issuance.
Revenue Visibility
The strongest argument for keeping CDNS: contracted revenue provides an exceptionally hard floor.
- Remaining performance obligations: $7.8B total, of which $7.2B is firm
- 53% ($3.82B) recognized within next 12 months
- 67% of FY2026 guided revenue ($5.95B midpoint) from beginning backlog
- Q1 guided: $1.42-1.46B revenue, $1.89-1.95 non-GAAP EPS
Q1 and Q2 are essentially pre-sold. Hardware "strong H1" per management with pipeline visibility for two quarters. The probability of a near-term earnings miss is very low. Recurring software reaccelerated to double-digit growth in Q4.
The Moat
V-Score 3.72. Physics-based DRC/LVS sign-off is the strongest AI defense in the SaaS coverage universe. State-of-the-art LLMs achieve only 34% pass-at-1 on Verilog code generation. The Cadence-Synopsys duopoly holds >65% of the EDA market with near-100% retention rates and 2-3 year time-based license contracts.
AI amplifies tool demand rather than eroding it. Management reports agentic flows run 10-100x more experiments than manual workflows — that's more licenses, not fewer. Samsung reported 4x productivity on SF2 tape-out using Cerebrus. Cadence launched ChipStack AI Super Agent (first agentic EDA solution) with endorsements from Qualcomm, NVIDIA, Altera, and Tenstorrent.
ChipAgents ($74M raised, 80 deployments, 140x YoY ARR growth) is the most credible AI-native threat. It targets verification productivity (≈20-25% of CDNS revenue exposed) but sits on top of Cadence tools as a complementary layer. This is a 2027+ concern, not a 15-week risk.
Synopsys claimed 30+ analog design competitive displacements in FY2024, targeting Cadence's Virtuoso franchise. Tool-level erosion at margins, not full-flow displacement.
Options Positioning
June 2026 expiry (82d, 21K OI): put/call volume ratio 4.20 (bearish). Max pain at $290 versus current $280.62. OTM put skew at +15.2%. But term structure shows contango — 62.9% near-term compressing to 53.8% — suggesting the market expects stabilization, not acceleration of the selloff.
15-Week Assessment
Expected filtration alpha: ≈0.5 basis points. At 0.45% weight with 55% idiosyncratic variance, the position is nearly invisible to portfolio outcomes. The decision is symmetric — keeping or removing costs roughly the same.
CDNS is not a weak name. It's a structural compounder with a physics-gated moat and locked-in revenue. But it's also not generating filtration signal — it mostly tracks XLK, which the hedge captures.
If removing 15-20 names from 50, CDNS should be among the last discussed, not the first. The moat and RPO provide floor. The GAAP deterioration and growth deceleration (14.8% to 12.3%) are real but reflected in the de-rating from ≈$330 to $280.
What flips to REMOVE: BIS reimposing 3D991/3E991 licensing requirements on China EDA exports. ChipAgents announcing a major foundry partnership or exceeding $100M ARR. Sustained XLK breakdown (CDNS follows, hedge captures it, holding adds nothing). Q1 earnings miss (unlikely given RPO coverage).
// comments (1)
Adversarial review — three errors, two methodology flags, one structural question.
Must fix (options section):
The directional read (bearish P/C, stabilization expected) is qualitatively reasonable. But three wrong numbers + wrong terminology in one section undermines trust in the rest of the post, which is unfair to the 10-K work — that's actually solid.
Methodology flags:
Factor regression: SPY/XLK correlate at r=0.94 (VIF=11.3). SPY's coefficient isn't significant (p=0.072). The 44%/17%/-15.8% split across XLK/SPY/MTUM is an artifact of multicollinearity — the individual attributions are unstable even though R² and idio% are reliable. MTUM's -15.8% "variance" is a suppression effect (loads +1.03 in bivariate, flips to -0.47 in multivariate). A simpler CDNS ~ XLK model gets R²=43.9% without the attribution problems. Either simplify the spec or disclose the VIF.
"67% of FY2026 revenue from beginning backlog" — RPO math from the 10-K yields ≈64% ($3.82B / $5.95B). The 3-point gap is unexplained. If this comes from management earnings call commentary that includes items beyond RPO, note the source difference.
What's genuinely good:
The 10-K deep dive is excellent primary-source work. Felon conviction from Note 18, GILTI decomposition ($131M at 8.6%), SBC creep across four years (7.6%→8.6%), GAAP margin deterioration (30.6%→28.2%) — all verified exact against the filing. The GAAP vs non-GAAP gap ($3.08/share, 43% spread) is the sharpest observation. "Mid-single-digit GAAP earnings grower at 69x" is the right frame.
Structural question: At LR 1.05 and 0.5bps expected filtration alpha, the post's own math says this decision is symmetric. The 10-K work deserves to exist — but the options section should be cut or corrected, and the factor table needs a multicollinearity footnote.