Financebotresearch desk研究台

SNPS › analyze

SNPS · Analyze

Semiconductors

Date: 2026-07-28 · Price: $383.82 · Mkt cap: $73.49B · EV: $81.85B · Sector: Semiconductors (EDA) Verdict: 🟡 WATCH — 6.0/10. Great core business, mediocre consolidated company, fair-not-cheap price. Next catalysts: Q3 FY26 earnings 2026-08-26 · Investor Day 2026-09-30.

A portfolio-specific passage was removed from the public build.


0. The One-Paragraph Answer

Synopsys owns one of the widest moats in enterprise software — chip designers cannot switch EDA vendors without risking a tape-out, and foundries mandate specific Synopsys tools for signoff. That fortress is intact. But the company that owns the fortress is not the company it was two years ago. It spent $35B on Ansys, levered a debt-free balance sheet with $10B of debt, issued 23% more shares, and in exchange delivered FCF per share of $10.36 in FY26E versus $10.22 in FY22 — four years, flat. Consolidated ROIC fell from ~21% to ~5%. Design IP, 18–20% of revenue, is losing share to Cadence at a rate of 46 percentage points of growth. Management's credibility was destroyed by a −36% single-day print and a securities class action naming the sitting CEO and CFO. Against all of that, the stock is at 26x FY26 non-GAAP EPS versus Cadence at 42x — a headline discount that shrinks to ~15–18% once you measure both on cash flow instead of adjusted earnings. My weighted fair value is $371, essentially the current price. This is a good business trading at a fair price. The framework calls that "wait / watch," not "value."


1. Fundamentals — General Health

Headline financials (fiscal year ends October)

FY22 FY23 FY24 FY25 FY26E (guide)
Revenue $4.62B $5.32B $6.13B $7.05B $9.625–9.705B
Gross profit $3.72B $4.29B $4.88B $5.43B
Gross margin 80.5% 80.6% 79.6% 77.0%
GAAP operating income $1.16B $1.33B $1.36B $0.91B ~11.6% margin
Non-GAAP op margin ~36% ~37% 38.5% 37.3% ~41%
Net income (GAAP) $0.98B $1.23B $2.26B* $1.33B
GAAP EPS $6.29 $7.92 $14.51* $8.04 $2.49–2.91
Non-GAAP EPS ~$12.78 $14.72–14.80
FCF $1.60B $1.51B $1.27B $1.35B ~$2.00B
Interest expense $1.7M $2.7M $36.8M $446.7M ~$533M
R&D $1.59B $1.85B $2.08B $2.48B

*FY24 net income inflated by the Software Integrity Group divestiture gain — not operating.

The owner's-eye view — the single most important table in this report

FY22 FY23 FY24 FY25 FY26E
Diluted shares 156.5M 155.2M 155.9M 165.7M ~193M
Revenue / share $29.52 $34.28 $39.31 $42.56 $50.08
FCF / share $10.22 $9.73 $8.14 $8.15 $10.36

Revenue per share is up 70% over four years. FCF per share is up 1.4%.

That is the whole bear case in one line. The framework's rule — "growth funded by dilution isn't growth for me" — is dispositive here. Synopsys bought $2.4B of Ansys revenue with $10B of debt and 32M new shares, and the cash that reaches a shareholder's pocket per unit of ownership has not moved since 2022. FY26E finally gets back to the FY22 level, and only because restructuring-driven cost cuts are landing.

FCF & CAGR

  • FCF 3yr CAGR: −5.5%. OCF 3yr CAGR: −4.4%. Yahoo only carries 4 annual periods; the framework asks for 5–8yr and I could not source a clean 8yr FCF series without roic.ai, so treat the CAGR as 3yr and directionally negative.
  • FCF declined from FY22 to FY25 in absolute dollars, during the largest semiconductor design boom on record.
  • FY26E FCF of $2.0B is depressed by restructuring cash ($300–350M pre-tax charge phasing through FY27) and integration costs. Normalized FY28 FCF, if management hits its "mid-30s% unlevered FCF margin" target, is ~$3.0–3.2B. That is the bull case's load-bearing assumption.
  • Cash conversion is poor and worth flagging. FY26E non-GAAP net income is ~$2.85B against guided FCF of $2.0B. For a software company that should convert at or above 100%, a 70% conversion ratio says the non-GAAP number is doing real work.

Capital allocation — where the FCF went

Use FY22 FY23 FY24 FY25
Capex $139M $192M $140M $169M
R&D $1.59B $1.85B $2.08B $2.48B
M&A $422M $298M $157M $16.68B
Buybacks $1.10B $1.16B $0 $0
Debt issued 0 0 0 $14.33B
Debt repaid $77M $3M $3M $864M
Dividends

Read: a disciplined allocator through FY23 (R&D + steady buybacks, no debt) that made one enormous bet in FY25 and stopped everything else to pay for it. Buybacks went to zero for two full fiscal years while stock-based comp ran at ~10% of revenue — that is a pure, unoffset transfer from shareholders to employees. Buybacks resumed only in Q2 FY26 at $263M ($250M ASR + $50M open-market) against ~$950M of annual SBC. The dilution is still net-negative and will remain so until repurchases roughly triple.

Balance sheet & leverage

FY24 FY25 Q2 FY26
Total debt $684M $14.29B $10.84B
Cash $3.90B $2.89B $2.48B
Net debt net cash $11.40B $8.36B
Debt / Assets 5.2% 29.6% ~22%
Goodwill $3.45B $26.90B
Equity $8.99B $28.33B
  • $3.5B of debt retired in roughly three quarters. That is genuine, fast deleveraging and the best thing management has done since the deal closed.
  • Tangible book value is deeply negative. Goodwill alone ($26.9B) exceeds total equity ($28.3B) less acquired intangibles. Any book-value-based valuation is meaningless here — see §4.
  • Interest expense of ~$533M/yr against $2.0B FCF is a 27% claim on cash flow that did not exist two years ago.
  • No published net-leverage target. Management says "prioritize deleveraging" qualitatively. I could not find a numeric goal — a gap worth pressing at the Sept 30 Investor Day.

🚩 Fundamentals scorecard

Test Result Grade
FCF CAGR (3yr) −5.5% 🔴
FCF/share trajectory Flat 4 years 🔴
Revenue CAGR (3yr) +15.2% (but +2.8% organic FY25) 🟡
Net income CAGR (3yr) +10.6% (FY24 distorted by divestiture gain) 🟡
Gross margin trend 80.5% → 77.0%, −3.5pts 🟡
Shares outstanding 152M → 193M, +26% 🔴
Debt / Assets 5% → 30% → 22%, deleveraging fast 🟡
Capital allocation Excellent pre-FY25; one $35B bet since 🟡
Backlog / visibility $11.0B, 83% recurring revenue 🟢
Cash conversion ~70% of non-GAAP NI 🔴

Data gaps flagged: 8yr FCF series unavailable from Yahoo (4 periods only); FY25 R&D-to-revenue and full segment history would need roic.ai or the 10-K; the FY2025 10-K itself could not be retrieved (SEC returns 403 to both WebFetch and agent-browser) — segment data below is sourced from Synopsys's own Q2 FY26 Financial Supplement, which is primary but not the 10-K.


2. Moat & Competitive Advantage

Quantitative base — ROIC is the finding

NOPAT Invested capital ROIC
FY24, GAAP $1.12B $9.01B 12.4%
FY24, non-GAAP $1.94B $9.01B 21.5%
FY25, GAAP $0.75B $41.81B 1.8%
FY25, non-GAAP $2.16B $41.81B 5.2%

Synopsys paid $35B for a business that took consolidated ROIC from ~21% to ~5%. Some of that is the accounting artifact every large acquirer suffers — goodwill inflates the denominator instantly while synergies arrive over years. But run the incremental math and it does not rescue the deal:

Ansys incremental returns NOPAT on $32.8B incremental IC ROIC
At ~40% adj. op margin, no synergies $0.97B 3.0%
Plus the full $1B eventual synergy target $1.79B 5.5%

Cost of the debt used to buy it: ~5.3%. Estimated WACC: ~9%. Even assuming management captures every dollar of synergy it has promised, the Ansys acquisition earns roughly its cost of debt and materially less than its cost of capital. This is the sharpest single indictment in the analysis and it is not something a recovery in Design IP fixes. The defense — that the market values cash flow, not accounting capital, and that Ansys is a genuinely high-quality #1 asset in CAE — is fair but does not change the arithmetic on the price paid.

Gross margin: 80.5% → 77.0% over three years. Most of the 3.5pt decline is Ansys mix and acquired-intangible amortization running through COGS; underlying tool margins are likely stable in the low 80s. Not evidence of moat erosion — but not evidence of strength either.

Moat sources, named

Source Strength Why
Switching costs Very wide Tools are sequentially dependent — changing synthesis forces re-running P&R, timing signoff, and physical verification. You cannot A/B test one tool; you evaluate a whole flow, on a project where a single mask set costs millions. Realistic switch: 12–24 months and 8-figure all-in cost for a large SoC team, and impossible mid-project. Teams switch at node transitions or never. Emulation hardware (~$50M systems, 5–7yr depreciation) adds a half-decade lock.
Intangibles / foundry certification Wide — and underrated Foundries co-develop PDKs with EDA vendors ~24 months before production, then specify which tools customers must use for tape-out. TSMC requiring PrimeTime for 7nm signoff is why PrimeTime holds 90%+ share — customers didn't choose it. The barrier is self-reinforcing: you can't get early PDK access without being trusted, and you can't earn trust without shipping silicon. Synopsys cites 1,400+ tape-outs at ≤16nm. This functions as quasi-regulatory protection no amount of capital shortcuts.
Efficient scale Moderate–wide The EDA+IP TAM is only ~$22.4B sitting underneath $805B of semiconductors. A market that small, requiring 30–35% of revenue in R&D, structurally supports about three players. The Big 3 have held 75–90% share for over a decade.
Cost advantage Narrow R&D intensity is the barrier, not unit cost. Zero advantage over Cadence. Chinese state-backed vendors face no return-on-capital constraint at all.
Network effects Narrow, indirect Verification-IP ecosystems, foundry reference flows, and a labor market trained on Synopsys/Cadence flows. Real but secondary.

Adversarial stress-test — how a well-funded rival attacks

Frontal assault on the full flow: fails. Replicating synthesis → P&R → timing → physical verification at 2nm accuracy, then getting TSMC to certify and mandate all of it, is a 10+ year project with a chicken-and-egg certification loop. No customer risks a $500M tape-out on unproven signoff. This is why the moat is real.

The attack that actually works: beachhead where AI resets the algorithm and no foundry certification applies. Don't attack signoff — attack RTL generation, testbench creation, and verification debug. Verification is 50%+ of design effort, sits upstream of signoff, and nobody needs TSMC's permission to write a better testbench generator. This is exactly where the funding is going: ChipAgents has raised $74M (Series A1 $50M, Feb 2026 — with Micron, MediaTek and Ericsson participating, i.e. strategic customers, not just VCs); Silimate, PrimisAI, Astrus in the same lane; 80 AI/EDA startups raised $8.4B in Q1 2026 alone. These don't kill Synopsys — they take the fastest-growing, least-defended slice and force competition on product rather than lock-in. That is margin pressure, not extinction.

The hyperscaler path. Google or Nvidia doesn't build EDA. It hires 200 EDA engineers, builds internal AI design agents on top of existing Synopsys licenses, cuts seat count 30%, and uses that leverage at renewal. Caps pricing power on the largest and fastest-growing accounts.

Where Synopsys is genuinely vulnerable: 1. Design IP is not moated — see §3. 2. Customer/foundry concentration. The FY25 blowup was substantially one "major foundry customer" (analysts infer Intel; Synopsys has never named them and this is unverified) deferring orders. A business one customer's roadmap slip can break has less moat than tool-share figures imply. 3. The balance sheet is now a constraint. $10.8B debt and $533M/yr interest mean Synopsys cannot buy its way out of the next problem the way it bought Ansys. 4. Execution, not competition, caused the damage. Management's own words: "certain roadmap and resource decisions that did not yield their intended results." A moat protects you from rivals; it does not protect you from yourself.

Competitive landscape — is "duopoly" still right?

Partly. Accurate for core digital implementation and signoff. Obsolete for the company Synopsys has become — ~31% of revenue now sits in a four-way simulation market (Siemens+Altair, Cadence+MSC/BETA-CAE/Hexagon D&E, Dassault, Keysight) where Ansys leads but is no longer the incumbent-by-default. Synopsys bought the CAE leader and both main rivals immediately moved to contest it.

Market-share estimates conflict badly across sources (31/30/13%, 38/36%, and 46/35% all appear for SNPS/CDNS/Siemens); methodologies are undisclosed. Only the "Big 3 ≈ 75–90%" aggregate is safe to use.

The head-to-head quarter is the cleanest evidence available:

SNPS Q2 FY26 CDNS Q2 CY26
Revenue $2,276M, +42% (all Ansys) $1,584M, +24%
Core EDA growth ~9% guided FY26 +18% y/y
IP growth −6% y/y +40% y/y
Non-GAAP op margin 39.5% 45.5%
Backlog $11.0B $8.1B (record)

Synopsys IP down 6% while Cadence IP grew 40% in overlapping categories (HBM, PCIe, UCIe, LPDDR6) is not a market problem. It is a share problem.

Disruption vectors, 5–10 year arc

Vector Read Verdict
AI-designed chips Bull mechanism is real: more design starts, verification intensity growing ~15%/yr, and Synopsys is monetizing via subscription + consumption pricing for AI compute + agents priced as headcount-equivalents. AMD/Microsoft autonomous debug workflows show 25–40% cycle-time cuts. Bear mechanism is equally real: if an agent replaces three verification engineers, willingness-to-pay anchors to the headcount replaced — and the agent layer sits above the certified flow, making it far more contestable than signoff. Synopsys's AgentEngineer has only 20 customers evaluating; Cadence's Cerebrus has 1,000+ tape-outs. TAM-accretive through ~2030; genuine seat-compression risk beyond. Synopsys is behind on AI commercialization despite claiming pioneer status.
Chinese domestic EDA China revenue $989.5M (16% of FY24) → $814.3M (12% of FY25), −17.7% (−22% ex-Ansys); Q2 FY26 $240.4M = 10%. Export controls imposed 2025-05-29, rescinded 2025-07-02 for rare-earth concessions — but the demand damage outlived the rule, because customers learned US EDA is a political dependency. Domestic capability: ~70% flow breadth, ~30% depth at advanced nodes; Empyrean H1 2026 revenue ~$70M (+13%, profit −92%) vs Synopsys at $9.7B. Not displacement this decade — but permanent loss of ~6pts of revenue base, handed to subsidized players climbing the node ladder. Unrecoverable even if controls reverse.
RISC-V The one place open-source genuinely bites — and Synopsys just conceded it: on 2026-01-14 it agreed to sell Processor IP Solutions (ARC, ARC-V, DSP, NPU, ASIP) to GlobalFoundries (~$40M FY26 revenue impact, closing 2H26). Nvidia has 40+ RISC-V cores in Blackwell/Rubin; Qualcomm bought Ventana; the spec sits outside US export control. Honest retreat from an unwinnable #4 position — good capital allocation, but the fastest-growing IP category is now someone else's.
Open-source EDA (OpenROAD/LibreLane) Real progress, but lives at mature nodes, academia, and low-volume shuttles. Excellent open-source synthesis is useless if TSMC won't sign off the tape-out. The Kimi K3 demo (2026-07-17) — autonomous RTL→tape-out in 48h — knocked 7.85% off SNPS, but CDNS fell 9.47%, and it ran on Nangate 45nm, generations behind the frontier. Negligible this decade. Watch as a talent/standards pipeline, not revenue competition.
Hyperscaler in-house silicon Google TPU v7, Trainium 3, Maia 200, MTIA — custom ASIC ~44.6% CAGR. Every new in-house team is a new EDA customer that previously spent $0 on tools. Strong tailwind for tools; mixed-to-negative for IP (a 500-person silicon team eventually builds its own SerDes).

Evergreen assessment — 7.5 / 10

  • Core EDA tools franchise alone: 9/10. Close to unkillable through 2036. Switching costs measured in tape-out risk; foundry certification is a trust-and-time barrier capital cannot shortcut; 15 years of stable Big-3 share through the mobile, cloud, and AI transitions.
  • Ansys / simulation: 8/10 as a business, but with materially more competition than when Synopsys bought it.
  • Design IP: 4/10 — genuinely at risk of structural impairment.
  • Weighted by the ~51% EDA / 31% Ansys / 18% IP mix, discounted for execution risk and leverage: 7.5.

Synopsys will exist and be profitable in 2036. Whether it compounds at the rate its valuation implies is a different question, and turns almost entirely on whether AI expands or compresses tool consumption.


3. Design IP — The Segment That Broke, In Detail

Segment revenue and margin, from Synopsys's own Q2 FY26 Financial Supplement:

Adjusted segment op margin FY24 FY25 Q1 FY26 Q2 FY26
Design Automation 38.7% 41.7% 47.3% 43.3%
Design IP 38.3% 23.9% 16.2% 24.4%

In FY24 the two segments earned identical margins. They are now different businesses. Why IP is structurally inferior to EDA tools:

Dimension EDA tools Design IP
Revenue recognition Ratable multi-year subscription Substantially upfront license + NRE + royalty
Renewal dynamic Automatic, 3–7% escalators Re-license per design, per node — every time
Switching cost Enormous (flow-wide) Low — swap one block at the next design
Foundry role Foundries mandate your tool Foundries compete with you via turnkey IP
FY25 margin 41.7% 23.9%

The core asymmetry: switching EDA vendors risks your tape-out; switching SerDes vendors risks a design review. IP is a components business wearing software clothing — no sequential dependency, no certification mandate, no ratable smoothing.

Is the turnaround credible? Partially. For: IP bottomed Q1 FY26 ($407M, 16.2%), rebounded +12% sequentially to $454.2M at 24.4%; >90% win rate on PCIe 7.0 with 18 new licenses; UCIe/HBM4 pipeline; a new Customer-Owned Tooling monetization model targeting signed agreements by FY26-end. Against: management itself calls FY26 "transitional and muted"; the long-term "mid-teens growth" objective is aspirational after two years of decline; and one quarter of sequential recovery off a trough is not a trend.

I would not underwrite the IP recovery. Underwrite EDA + Ansys and treat IP as free optionality.


4. Valuation

Where the stock actually trades

Metric SNPS CDNS Note
Price $383.82 ~$338.61
FY26E P/E, non-GAAP 26.0x ~41.8x The 38% "discount" everyone quotes
P/E, GAAP (FY26E) 142x Amortization + SBC + restructuring
Fwd P/E (Yahoo, FY27 basis) 22.2x Watchlist's "22.2x" is FY27, not FY26
EV / FY26E FCF 40.9x ~48x The real discount: ~15–18%, not 38%
EV / normalized FY28E FCF ~27x ~43x If synergies land
EV / Revenue 8.47x ~14.5x
FCF yield 2.72% ~2.0%

This resolves the central tension. The sell-side's "38% cheaper than Cadence" is largely a leverage-and-amortization artifact. Cadence has no $10B debt and no $1.6B/yr of acquired-intangible amortization, so its non-GAAP EPS is a much closer proxy for cash. Measure both on cash flow and SNPS is only ~15–18% cheaper on FY26 numbers. It becomes genuinely cheap only if you accept the normalized-FY28 FCF bridge — which is exactly the assumption management has not yet earned the right to be trusted on.

Models applied

Graham's Intrinsic Value — computed, then discounted to near-zero weight.

Inputs Result
GAAP TTM EPS $4.30, BVPS $159.20 $124
FY26E GAAP EPS $2.70, BVPS $159.20 $98
FY26E non-GAAP EPS $14.76, BVPS $159.20 $230

All three are close to meaningless and I am flagging that rather than using them. BVPS of $159.20 is >90% goodwill and acquired intangibles — tangible book value is deeply negative. Graham's formula assumes book value represents recoverable assets; here it represents the price paid for Ansys. And GAAP EPS is crushed by amortization of that same purchase. Graham systematically penalizes asset-light software regardless. Weight: 10%, and only as a reminder that there is no asset floor under this stock.

Bogle's Expected Return = dividend yield + earnings growth ± P/E change. - Dividend yield: 0% (no dividend, none planned). - Earnings growth: FY26 non-GAAP EPS $14.76 vs FY25 ~$12.78 = +15.5%. Sustainable 5yr rate: revenue ~8–10%, margin expansion 41% → mid-40s adds ~2pts, net buyback ~1% → call it 12%. - P/E change: 26.0x today vs a 5yr historical forward multiple in the 35–45x range. Reversion to 30x over 5 years = +2.9%/yr; flat at 26x = 0%; compression to 22x = −3.3%/yr. - Bogle expected return: 8.7% – 14.9%/yr, base case ~12%. Respectable, but note it inherits the non-GAAP EPS definition and therefore ignores ~$950M/yr of SBC.

DYT and DDM: N/A. No dividend.

Reverse DCF — what must SNPS do to justify $383.82? At a 10% discount rate and 3% terminal growth, EV of $81.85B requires terminal FCF of $5.73B. From FY26E's $2.0B that is 14.1%/yr over 8 years or 11.1%/yr over 10 years. Demanding but not absurd for a duopolist with an $11B backlog guiding to mid-40s operating margins. The price is not pricing in a disaster — it is pricing in success, just less success than the sell-side assumes.

Fair value build

Method Bear Base Bull Weight
A. Multiple on FY27E non-GAAP EPS (22x/28x/33x × $16.0/$16.8/$17.5) $352 $470 $578 30%
B. Normalized FCF ($2.6/3.0/3.4B × 22x/28x/33x, less $8.36B net debt, 193M sh) $253 $392 $538 35%
C. SBC-charged owner earnings (50% of ~$1.0B SBC charged as real cost) $218 $306 $407 25%
D. Graham $98 $164 $230 10%

Fair value range: $300 – $470. Weighted base case: $371. Current price $383.82 sits just above the weighted midpoint. SNPS is fairly valued — not cheap.

Method C deserves a word: SBC runs ~9.9% of revenue (~$950M/yr) while buybacks are $263M/quarter at best. Non-GAAP EPS adds SBC back entirely; FCF adds it back too (it's non-cash). Neither captures that shareholders are being diluted ~1.5–2%/yr net. Charging half of SBC as a real economic cost is the honest middle ground between "ignore it" (management's view) and "charge all of it" (which double-counts against the buybacks funded from the same FCF).

The street's $564 mean target implies ~34x FY27 non-GAAP EPS — it assumes the IP recovery lands, ignores SBC entirely, and re-rates the multiple back toward Cadence. Every one of those is a choice, not a fact. Note also that the lowest target on the street is $414.77, above spot — there is no bearish sell-side anchor below the current price, which means no downgrade cushion if 8/26 disappoints.


5. Sentiment & Intelligence

The de-rating was one day, not a slow bleed

2025-09-10: SNPS fell −35.84% in a single session ($604.35 → $387.78) — the worst day since its 1992 IPO. Today's $383.82 is essentially the same price as the close on the day of the crash, ten months later. In between it round-tripped to $500.82 (2026-04-24) on the Elliott news and gave it all back in the June–July AI de-rating.

What caused it (Q3 FY25, reported 2025-09-09): revenue $1.740B vs ~$1.77B consensus; non-GAAP EPS $3.39 vs ~$3.75; Design IP revenue $428M, −8% y/y, missing ~$140M of expected IP revenue; Design IP adjusted margin collapsed 36.7% → 20.1%; FY25 EPS guidance cut from $15.11–15.19 to $12.76–12.80. Management named China export restrictions, a major foundry customer pulling orders, and "roadmap and resource decisions that did not yield their intended results."

The CDNS divergence is real but smaller than assumed

52wk closing high Date Now Drawdown
SNPS $636.02 2025-08-04 $388.99 −38.8%
CDNS $416.39 2026-06-02 $338.61 −18.7%

(The $651.73 in Watchlist.md is the intraday high; closing basis is $636.02.) Cadence is down nearly a fifth from its own high, set only eight weeks ago. And Cadence's just-reported Q2 was a penny EPS beat on a revenue miss ($1.584B vs $1.608B est) — the press called it "strong," but both companies raised FY guidance this quarter. The operational divergence is narrower than the narrative.

What actually explains the gap, ranked: 1. Segment mix (dominant). Cadence has no Design IP equivalent. SNPS carries a 20%-of-revenue segment down 6% y/y with margins that fell from 36.7% to 20.1%. The market pays a premium for narrative purity. 2. Credibility damage (the real answer). A −36% day plus a securities class action permanently raises the discount rate applied to management guidance. This is why a beat-and-raise on 2026-05-27 still produced −8.6% while nine analysts raised targets. The market no longer takes SNPS guidance at face value. 3. Ansys optics. GAAP $0.09 vs non-GAAP $3.35 per share probably matters more than actual integration failure, which is tracking fine. 4. Kimi K3 is not the explanation — CDNS fell harder on that day (−9.47% vs −7.85%).

Guidance has been raised three times and never cut

Date FY26 Revenue FY26 non-GAAP EPS
2025-12-10 (Q4 FY25) $9.56–9.66B $14.32–14.40
2026-02-25 (Q1 FY26) reiterated raised $14.38–14.46
2026-05-27 (Q2 FY26) raised $9.625–9.705B raised $14.72–14.80

The move from $500 (April) to $384 (now) happened while estimates were rising. That is multiple compression, not estimate compression — and it is the strongest single argument for the bull case.

Q2 FY26 (reported 2026-05-27) — the quarter the market rejected

Revenue $2.276B (+41.9%) · Design Automation $1.821B · Design IP $454.2M (−6% y/y, +12% q/q) · non-GAAP EPS $3.35 · GAAP EPS $0.09 · DA adj. margin 43.3% · IP adj. margin 24.4% · backlog $11.0B. Q3 guide: $2.410–2.460B revenue, $3.63–3.69 non-GAAP EPS.

Management tone was optimistic, arguably too much so: "the IP segment bottomed in Q1, and has begun its recovery… we expect sequential quarterly improvements throughout the second half." On China: "the design start environment in China remains challenged given all the restrictions." On Ansys: "by the end of this fiscal year, we will have achieved about half of our committed synergies."

Analyst positioning — and why it cuts both ways

Consensus buy, 25 analysts, mean target $564.43 (high $650, low $414.77) → +47% implied. Even the lowest target is 8% above spot. Recent actions: Benchmark initiated Buy $570 (7/16); Piper Sandler upgraded Neutral→Overweight, $450→$550 (6/23); on 2026-05-28 nine firms raised or reiterated (Citi $610, BofA $515→$600, Stifel $600, Needham $580, Rosenblatt $575) — and the stock fell 8.6% that day. BofA has round-tripped Buy → Underperform → Neutral → Buy → $600 inside nine months. The only bear of record is BNP Paribas/Exane (Underperform since 2025-10-02, target $450 — still above spot).

Read this as a warning, not a comfort. There is no bearish sell-side anchor below $384 and short interest is negligible (5.61M shares, 2.86 days to cover). The $564 mean target has little credibility as a floor, and there is no downgrade cycle left to absorb a bad print.

🚩 Insiders: zero dip-buying

A portfolio-specific passage was removed from the public build.

Yahoo's "41,603 shares purchased / 25 transactions" is misleading — those are RSU vestings and director grants at $0.00. Every voluntary transaction in the last 12 months was a sale, including the CEO taking $6.7M off the table at $459 six weeks before the stock hit $373. Likely 10b5-1 and not a red flag standing alone, but there is zero insider buying at $384 while management is publicly telling the market the IP business has bottomed.

Institutional (3/31/26 filings): 93.6% institutional. BlackRock 8.29% (+4.8%), Vanguard 6.50%, State Street 4.65%, FMR 4.37% (+11.7%), Nvidia Corp 2.52% / 4.82M shares (~$1.85B, unchanged), Goldman 1.64% (+296%), Morgan Stanley 2.05% (−14.4%).

Live overhangs

Overhang Status Severity
Securities class action Filed 2025-10-31, N.D. Cal. §10(b)/10b-5, naming Ghazi (CEO) and Glaser (CFO) personally. Class period 2024-12-04 → 2025-09-09. Alleges concealment of AI-customer customization destroying IP profitability. Motion-to-dismiss status unverified. Medium-High
Elliott Management Stake revealed 2026-03-22/23 ("multibillion"); cooperation agreement 2026-05-27; Jesse Cohn on the board. Pushing software monetization, growth, margin expansion. Exact stake % undisclosed. Positive catalyst
Restructuring ~2,000 jobs / ~10% of combined workforce; $300–350M pre-tax charge; runs through end of FY2027. Headcount 28,330 → 26,380. Medium (execution)
Product EOL Reuters 2026-07-07: discontinuing EES and FDC fab process-control software; ~10 chipmakers notified incl. Samsung, SK Hynix, Kioxia, Qorvo. Samsung confirms no production impact. Low financially, medium optically
China Controls lifted 07/2025, demand impaired; 16% → 10% of revenue. Renewed escalation is a live tail risk. Medium
CEO/CFO/accounting No changes, no restatement, no impairment, no auditor issue. Only flag is the GAAP/non-GAAP spread. None / Low

6. Debate Round — Resolving the Tensions

Tension 1 — Sentiment says "26x vs Cadence's 42x is a 38% discount on a duopoly." Moat says "organic growth was 2.8% and non-GAAP EPS is inflated by ~10%-of-revenue SBC." Resolution: Measure both on cash. EV/FY26E FCF is 40.9x for SNPS vs ~48x for Cadence — a ~15–18% discount, not 38%. The P/E gap is largely a leverage-and-amortization artifact. Sentiment's framing overstates the bargain by roughly half. Moat wins this exchange, but with a concession: on normalized FY28 FCF the gap widens back to ~27x vs ~43x, so the bull case survives if — and only if — the FCF bridge is real.

Tension 2 — Fundamentals says ROIC collapsed from 21% to 5%. Sentiment says the deal is tracking, synergies half-captured, $3.5B of debt retired. Resolution: Both are true and they answer different questions. Integration execution is genuinely fine. The price paid is the problem: incremental ROIC on Ansys is 3.0% today and only 5.5% assuming the full $1B synergy target — against ~5.3% cost of debt and ~9% WACC. Good execution of a mediocre deal. Fundamentals wins on substance; Sentiment is right that this is not a deteriorating situation.

Tension 3 — Management says IP "bottomed in Q1." Moat says Cadence IP grew 40% while SNPS IP fell 6%. Resolution: Not resolvable from public data — and that is the point. Q2's +12% sequential and 24.4% margin are consistent with a bottom. The 46pt growth gap versus Cadence is consistent with permanent share loss. These are indistinguishable until at least two more quarters of data. This is the single largest unresolved question in the thesis, and it gets its first real test on 2026-08-26. Treat it as a named risk factor, not a settled call.

Tension 4 — Guidance raised three times vs. the stock down 23% since April. Resolution: Sentiment's diagnosis is correct and important: this is a credibility discount, not an earnings discount. But a credibility discount is not automatically an opportunity — it is a real cost of capital that persists until management re-earns trust, and my valuation work says the discount currently applied is roughly appropriate, not excessive.


7. Verdict

🟡 WATCH — Conviction 6.0 / 10

Do not initiate at $383.82. Not because the business is bad — the core EDA franchise is a 9/10 moat that will be here in 2036 — but because the framework's first question has a clear answer: the market has already priced what's good, and priced the problems roughly correctly too. Weighted fair value $371 against a $383.82 price is not a margin of safety; it is a coin flip with $10.8B of debt attached.

What the bulls have right: a genuinely wide moat protected by switching costs and foundry certification that capital cannot shortcut; an $11B backlog with 83% recurring revenue; FY26 guidance raised three times and never cut; $3.5B of debt retired in three quarters; synergies half-captured; buybacks restarted; an activist on the board with a September Investor Day to attach numbers to; and AI design complexity as a real multi-year demand tailwind.

What the bears have right, and it is more: FCF per share is flat over four years despite $35B of M&A, $10B of debt, and 26% more shares. ROIC went from ~21% to ~5%, and the Ansys deal earns roughly its cost of debt even assuming every promised synergy lands. Organic growth was 2.8% in FY25 — during the biggest design boom in history. Design IP is losing to Cadence by 46 points of growth. SBC runs ~10% of revenue against buybacks a third that size. Management's credibility is broken, the CEO and CFO are personally named in a securities class action, and neither of them bought a share at $384 — the CEO sold $6.7M at $459 six weeks before the low.

The honest summary: this is a good business trading at a fair price, not a great business trading cheap. The framework reserves "value" for the latter.

Why this is a close call

Two things could make me wrong quickly. First, FY26 FCF of $2.0B is a genuine trough — restructuring cash and integration costs roll off through FY27, and if management hits its stated mid-30s% unlevered FCF margin, normalized FCF is $3.0–3.2B and the stock is at 27x normalized EV/FCF versus Cadence at 43x. That is a real gap. Second, there are two dated catalysts inside nine weeks, which is an unusually clean setup — I am not asking you to wait indefinitely, I am asking you to wait 29 days.

Entry framework

Zone Price Rationale
No-buy Above $370 Above weighted fair value. No margin of safety.
🟡 Starter (only after 8/26 confirms) $340–370 Requires Design IP margin >28% and sequential revenue growth on 8/26.
Accumulate $310–335 15–20% below the normalized-FCF base case of $392. This is where the math works without needing to trust guidance.
💚 Strong Below $280 Below the bear case on every method except Graham.
✂️ Trim $520+ Approaching the bull case; would imply 31x FY27 non-GAAP EPS.

⚠️ The Watchlist's $440–480 add zone is wrong and should be retired. It was set on a pre-Ansys view of the company — before $10B of debt, 26% share dilution, and a 16pt ROIC collapse. Corrected to $310–335.

The two dates that matter

Date Event What to watch
2026-08-26 Q3 FY26 earnings Design IP: sequential revenue growth AND margin above 28%. Management staked its restored credibility on "sequential improvements throughout the second half." If IP is flat or down, credibility is gone a second time — and with the lowest street target at $414.77, there is no analyst cushion below $384. Also: core EDA organic growth vs Cadence's ~18%, and FY26 FCF tracking to $2.0B.
2026-09-30 Investor Day (NYC) The venue where the Elliott thesis either gets numbers or doesn't. Watch for: an explicit net-leverage target, an FCF-conversion target, a buyback commitment sized against SBC, and FY27 Ansys revenue-synergy milestones. Absence of a capital-return framework here would be a meaningful negative.

Break triggers (what would kill the thesis entirely)

  • Design IP revenue declines sequentially for two more quarters, or margin fails to hold 24%.
  • Core EDA organic growth falls below 7% while Cadence holds mid-teens — confirms share loss in the moated segment, not just IP.
  • FY26 FCF comes in below $1.8B, or FY27 guidance implies FCF conversion under 75% of non-GAAP net income.
  • Goodwill impairment on Ansys, or the $1B synergy target being walked back.
  • Buybacks fail to reach ~$1.0B/yr run-rate by FY27 while SBC stays near 10% of revenue.
  • Renewed China export controls.

A portfolio-specific passage was removed from the public build.

8. Data Gaps & Unverified Claims

Flagged explicitly per the framework's skepticism default:

  • FY2025 10-K could not be retrieved — SEC.gov returns HTTP 403 to WebFetch and serves bot-detection to agent-browser. Segment data is from Synopsys's Q2 FY26 Financial Supplement and May 2026 investor deck (primary, but not the 10-K). Not verified: verbatim competition section, formal customer-concentration disclosure, risk-factor language.
  • 8yr FCF series unavailable — Yahoo carries 4 annual periods. CAGRs are 3yr. roic.ai would close this gap.
  • "Intel" as the major foundry customer behind the FY25 IP miss is analyst inference. Synopsys has never named them.
  • Market-share figures conflict irreconcilably across sources (31/30/13%, 38/36%, 46/35%). Only "Big 3 ≈ 75–90%" is safe. Tool-level shares (Design Compiler ~85%, PrimeTime ~90%, Calibre ~85%) come from a single source (SemiAnalysis), uncorroborated.
  • Elliott's exact stake size is undisclosed.
  • Motion-to-dismiss status in the securities class action is unverified.
  • No published net-leverage target — only qualitative "prioritize deleveraging."
  • Updated FY26 China revenue % from a primary Synopsys source could not be confirmed beyond the Q2 FY26 figure of 10%.
  • Cadence FY26 FCF is my estimate (~$1.9B) for the EV/FCF comparison, not a guided figure.
  • The 2025-11-04 SNPS-specific −6.59% move has no identified catalyst.

Sources

Primary: Synopsys Q2 FY2026 Financial Supplement · Investor Overview, May 2026 · Q4 FY25 release · Q1 FY26 release · Q2 FY26 release · Q2 FY26 call transcript · GlobalFoundries processor-IP divestiture · AMD/Microsoft agentic AI · Export-control lifting statement

Data: Yahoo Finance via .mcp/fin.py (snapshot, statements, news) · StockAnalysis SNPS

Industry & competitive: SemiAnalysis — EDA Market Primer · Silicon Analysts — Foundry Engagement/PDK · Cadence Q2 2026 results · Cadence Q2 call highlights · Siemens completes Altair · Omdia — Arm steps deeper into silicon

Events: Futurum — Q3 FY25 IP weakness · Rijnberk — Synopsys crashed 36% · Bloomberg Law — securities suit · Kessler Topaz case page · CNBC — Elliott stake · Seeking Alpha — 10% workforce cut · Yahoo — shares fall, targets rise · Investing.com — Kimi K3 EDA demo

China/geopolitics: Tom's Hardware — EDA export ban lifted · TrendForce — China revenue at risk · SCMP — Empyrean · Digitimes — China EDA gap

AI/EDA disruption: ChipAgents $74M raise · SemiEngineering — Startup Funding Q1 2026 · eeNews — LibreLane