Synopsys vs Cadence: three strategy comparisons
Synopsys vs Cadence: three strategy comparisons
As of 9 October 2026. Sources are the filings, call transcripts and Investor Day materials in the Research folder's sources/. Short cites: "Q2 FY26 call" means that quarter's earnings call, and "ID" means the Synopsys Investor Day of 30 September 2026. Each section separates facts from interpretation.
Bottom line
| Topic | Synopsys | Cadence | Edge |
|---|---|---|---|
| Agentic tools | Open platform with three ways to buy; partnered with OpenAI; 50+ customer engagements | Four named "super agents" across the flow; ChipStack already in production | Cadence on shipping; Synopsys on how it plans to charge |
| Hardware-software and cross-sell | Record emulation hardware, plus a $400M Ansys revenue-synergy target that starts landing in FY2027 | Record hardware built on its own chips; cross-selling across EDA, IP, hardware and systems is already showing in results | Cadence on proof; Synopsys has the bigger upside if Ansys works |
| Owned processor IP | Sold the processor IP business (ARC) to GlobalFoundries; moving IP money to custom work paid by royalties | Keeps the Tensilica DSPs and added foundation IP through acquisitions | Synopsys's IP plan is clearer but riskier |
None of the three changes the peer-duel call. Item 3 is the most important for the thesis: if Synopsys's royalty model works, IP becomes a long-term growth engine. If it doesn't, Synopsys repeats the 2025 problem of depending on a few customers.
1. Agentic tools and workflows
Facts
| Synopsys | Cadence | |
|---|---|---|
| Products | AgentEngineer, a five-level autonomy roadmap (L1-L5). Agents cover verification, implementation, analog/mixed-signal, manufacturing and Ansys simulation tasks. The full-stack platform is called "Autopilot" (Q1 FY26 call; ID). | ChipStack AI Super Agent (RTL design and verification) launched February 2026. AgentStack, a head-agent framework, plus ViraStack (analog) and InnoStack (digital implementation and signoff) followed in April. AuraStack (PCB and packaging) came in Q2 (Q4 FY25, Q1 and Q2 FY26 calls). |
| Traction | 20 customers evaluating more than 25 agents (Q2 FY26 call). More than 30 active engagements (Q3 FY26 call). "50+ customer engagements" (ID). | ChipStack has more than 20 customer engagements and is "deployed in production across multiple chip designs" (Q2 FY26 call). Endorsed by Qualcomm, NVIDIA, Altera and Tenstorrent (Q4 FY25 call). |
| Partners | NVIDIA and Microsoft. At DAC, demonstrated an autonomous verification agent with NVIDIA, claiming up to 50x faster time to validated RTL. GPT-Synopsys with OpenAI as "preferred partners" (Q3 FY26 call; ID press release). | Google (ChipStack with Gemini), NVIDIA (a "virtual AI design engineer" shown at Computex), Broadcom, MediaTek, and Rapidus adopting InnoStack (Q4 FY25, Q1 and Q2 FY26 calls). |
| Pricing | Three tiers: platform subscription, agent subscription, and tool subscription or consumption. Customers can run Synopsys's stack, put Synopsys agents inside their own platform, or use frontier models. Agents are claimed to make 5-10x more tool calls (ID). | New agentic products are priced as a "virtual engineer" on subscription plus consumption. Base tools stay on subscription: "we don't see AI forcing a wholesale change from subscriptions to consumption" (Q4 FY25 call; Q1 FY26 call). |
| Revenue disclosed | None (ND). | None (ND). Management says the Q2 guidance raise was "broad-based… rather than any single customer or product" (Q2 FY26 call). |
Interpretation
- Same bet. Both companies argue that agents call the base tools far more often than a human engineer does, so license demand grows faster than engineer headcount. If that is right, both gain, and the agent is a way to sell more seats rather than a new profit pool.
- Where they differ. Cadence turned its agents into named products for each stage of the design flow faster, and has one in production. Synopsys has the more open design and the clearer pricing plan, and its OpenAI tie hedges the main risk: that customers' own agents or frontier models capture the orchestration layer and treat EDA tools as interchangeable engines.
- Engagement counts don't compare. Synopsys's 50+ covers its whole portfolio; Cadence's 20+ is for ChipStack alone.
- KPI. Watch for the first disclosed consumption contract or agentic revenue figure, and for production deployments rather than evaluations. A slower rise in design-team headcount than in tool-license demand would confirm the thesis for both.
2. Hardware-software integration and cross-selling
Facts
| Synopsys | Cadence | |
|---|---|---|
| Emulation and prototyping hardware | ZeBu Server 5, ZeBu-200, HAPS-200 and the EP hybrid. Record hardware year in FY25 and a record quarter in Q3 FY26. A "marquee emulation win versus the incumbent at a leading AI HPC customer" (Q4 FY25, Q1 FY26 and Q3 FY26 calls). | Palladium Z3 and Protium X3 ("Dynamic Duo"). More than 30 new hardware customers in 2025; 7 of the top 10 customers bought both. Best quarter ever in Q1 2026 and another record in Q2 (Q4 FY25, Q1 and Q2 FY26 calls). |
| Who makes the silicon | The 10-K cites "a sole supplier for certain hardware products" but does not name it (FY25 10-K). | "We design our own chips, you know, made by TSMC, and we sell full racks" (Q4 FY25 call). |
| Compute hardware | GPU acceleration with NVIDIA, which also bought $2.0B of Synopsys shares in December 2025 (Q3 FY26 10-Q). | Millennium M2000 supercomputer on NVIDIA Blackwell for simulation; Reality Data Center digital twin used at hyperscalers (Q4 FY25 call). |
| Cross-sell mechanism | Ansys. Multiphysics Fusion (thermal analysis built into the chip flow) launched in Q3 FY26. It is "not expected to contribute to EDA growth until 2027." Target: ~$400M revenue-synergy run-rate by FY29, from Multiphysics Fusion plus selling more into systems customers (Q3 FY26 call; ID p.98). | Gradual expansion across the portfolio. The CFO says "Cadence monetizes workload through broad portfolio proliferation across EDA, IP, hardware, and SDA." 25 new digital full-flow customers in 2025. Hexagon D&E (structural analysis) plus BETA CAE and its CFD tools give "finally a full flow in SDA" for physical AI (Q4 FY25 and Q2 FY26 calls). |
| Systems scale | Ansys guided to about $2.98B for FY26 (ID). | System design and analysis about $0.85B in 2025, plus about $0.2B a year from D&E (FY25 10-K, inferred from product mix; Q1 FY26 call). |
Interpretation
- Hardware. Both are riding the same need: AI chips have to be checked against real software before they are manufactured. Cadence's advantage is structural, because it owns the processor inside its emulator. That gives it control of the roadmap and margin. Synopsys depends on an unnamed sole supplier, and the sources don't say who it is.
- Cross-selling. Cadence's integration is already showing up in its numbers (SDA grew 13% in 2025 and 18% in Q1 2026, both helped by acquisitions). Synopsys has the much larger systems franchise, but the cross-sell that justifies the Ansys price is a product launched only last quarter, with no revenue before FY2027.
- Risk. Hardware is up-front revenue, so it is the first line to fall in a pause in AI spending. That matters more for Cadence, where up-front revenue is 22% of the total.
- KPI. For Synopsys, a progress figure on revenue synergies and Multiphysics Fusion named as a driver of FY2027 EDA growth. For Cadence, whether hardware sets another record in the second half of 2026 without the up-front share rising further.
3. Owned processor IP
Facts
- Synopsys sold. The Processor IP Solutions business (ARC, which had moved to RISC-V) went to GlobalFoundries on 1 June 2026. Cash consideration was $443.3M, with a $425.4M pre-tax gain, or $380.5M after $44.9M of divestiture costs. The 10-Q calls it "not material" (Q3 FY26 10-Q). It removed about $40M of revenue from the rest of FY26 (Q2 FY26 call).
- Management's reasoning: "many of our customers are developing their own processor IP, using… our EDA software… our hardware portfolio." Synopsys would rather put investment into interface IP (Q1 FY26 call).
- Where Synopsys IP money goes now. "Factory 2" application-optimized IP is built with selected customers and paid as a license, a customization fee and a royalty. Targets: about $1B by FY30, and royalty revenue that eventually outweighs license fees. The first named deal is with Amazon. The Design IP segment target is 17%+ annual growth for FY26-30 (ID p.29 and p.38; ID press release).
- Cadence kept and added. It keeps Tensilica configurable DSPs for AI, audio, vision and baseband (FY25 10-K). It launched the HiFi iQ DSP, claiming up to 8x AI performance (Q4 FY25 call), and won its "first-ever Tensilica DSP design win with STMicroelectronics" (Q2 FY26 call). In 2025 it bought Arm's Artisan foundation IP and Secure-IC (FY25 10-K). Tensilica revenue is not disclosed (ND).
Interpretation
- Neither competes in general-purpose CPUs any more. Synopsys exited processor cores, a market squeezed between Arm and free RISC-V designs, and keeps its tool revenue from customers who build their own cores. Cadence's Tensilica is a niche DSP franchise, not a CPU bet.
- The real divergence is how they get paid. Synopsys is moving part of IP to custom co-design paid by royalties, which raises the ceiling but ties revenue to a few customers' chip volumes. That is the same kind of exposure that hurt it in 2025, when a single foundry customer drove IP down 8%. Cadence stays with catalog licensing, which is lower risk, and is currently gaining share with it (IP up over 40% in Q2 2026).
- KPI. Synopsys's number of AOIP customers and the first royalty disclosure (FY2027), against Cadence's IP growth staying above 20%.
Not in the sources
- Revenue from agentic products at either company.
- The identity of Synopsys's sole hardware supplier.
- Tensilica revenue.
- Any split of hardware versus software revenue at Synopsys.
