Unlock Premium Content – For just $0.39/day ($12/month) or $0.27/day ($100/year)!
🔗 Portfolio Corner - 🗓️ Monthly Updates (Last Update: 3-May-2026)
When NVIDIA unveiled its latest AI accelerator with 500M+ transistors, or Apple rolled out its 3nm M4, most miss the common thread: neither exists without TSMC… or Cadence Design Systems.
From that vantage point, Cadence sits at one of tech’s most interesting intersections. A $550B semiconductor industry runs on ~$10B of EDA software. If you’re in finance, you live in Excel. If you design chips, you live in Cadence.
What makes this business so compelling is that you literally cannot design advanced chips without sophisticated EDA software. It’s physically impossible to do manually. The entire digital economy rests on semiconductors, and those semiconductors can’t be designed without Cadence’s tools. The company maintains essentially 0% customer churn. Unless a customer goes out of business, they continue using Cadence tools.
Topics I’ll Cover
🔹 Company’s Overview & Culture
🔹 The Business Model
🔹 What Cadence Tools Actually Do
🔹 The Economic Moat
🔹 Three Secular Megatrends Driving Growth
🔹 Capital Allocation
🔹 What Could Go Wrong: Risk Factors
🔹 The AI & Chiplet Optionality
🔹 Outlook, Valuation and DCF Model
🔹 Investment Thesis Summary
Before we dive in, let’s see what else you can find and what’s new at Expanse Stocks.
📰 What’s New?
📚 Articles
🔎 Deep Dive Briefs
⛅ Cloudflare | 👷♂️ Parsons | 𓇲 MPS | 🥼 Medpace #1, Medpace #2 | 🔌 Arista | 🛒☁️ Amazon | 🤖 ASML | 🦎 Topicus | 💡 Lumine | ✈ HEICO | 🧙♂️ CSU Part #1, Part #2, Part #3 | 🤖 Intuitive | 🛒 💳 MELI Part #1, Part #2, Part #3 | 💻 Cadence
💸 General Investing
💎 Hidden Gems Series
✨ Annual Specials – Annual Letters, Investing Philosophy, Industry Write-ups & Top Picks of the Year
📚 Resources for Investors
📢 Latest Stock News → Earnings Digests & other related news
💬 Join My Chat → [🔗 Learn more]
🔐 Paid Subscriber Exclusives
💼 Portfolio Corner – Holdings, valuation models, trades, performance & more!
🤫 Exclusive Sections – From select articles, Deep Dive Briefs & Hidden Gems
👀 Recent Releases!
💸 Software Darwinism and Software Darwinism 2.0 → We talk AI, SaaS, VMS and Where the CSU family stands
💎 Hidden Gems: Sixth Edition → A Primer on TerraVest Industries
👀 Coming Soon
🔎 Deep Dive Brief: Norbit ASA
💎 Hidden Gems: Seventh Edition —> 2 Nordics Niches, early stage.
Company’s Overview & Culture
Founded in 1988 through the merger of two pioneering EDA companies (ECAD and SDA), Cadence has evolved from a fragmented player in a nascent industry into part of a three-legged oligopoly alongside Synopsys and Siemens EDA. Together controlling more than 90% of the global electronic design automation market.
Today, Cadence generates roughly $5.3B in annual revenue and serves every major chip designer, from Apple and NVIDIA to hyperscalers like Google and Amazon, which are increasingly building custom silicon.
Cadence’s story is inseparable from its turnaround under Lip-Bu Tan, who took the helm in 2008 during the company’s darkest hour and the GFC. The stock traded below $3, the management team was wiped out, and many in the industry thought Cadence might not survive.
The problems were structural: destructive pricing where customers would hang salespeople out to dry at quarter-end, product gaps in digital design versus Synopsys, and revenue recognition games prioritizing short-term metrics over long-term value.
What Lip-Bu did next deserves serious study. First, he transitioned to three-year subscription contracts, eliminating quarterly desperation and creating predictable revenue. One of the very first to dare doing this in the early SaaS model transition years. This helped realigned incentives between Cadence and its customers around continuous value delivery rather than transactional relationships.
But the really interesting part started right after that: instead of taking his foot off the gas after stabilizing the business in 2010, he reinvested on the business more, way more. Cadence had always been focused in analog design, but they decided to take on Synopsys head-to-head in digital design, going after the NVIDIAs and Intels of the world as customers.
Speaking of Intel… if there’s one thing that makes me even remotely bullish on this wrecked ship, it’s Lip-Bu Tan. He stepped in as CEO last year, and if there’s anyone capable of turning around a sinking vessel, it’s him.
The R&D Investment Philosophy
With Lip-Bu Tan at the helm, Cadence underwent a 360-degree shift in its capital allocation strategy.
Some context: Cadence reinvests roughly 35% of revenue into R&D, making it one of the most R&D-intensive businesses at scale. For comparison, traditional enterprise software typically spends 15–20%, while Microsoft sits closer to 13–15%.
Why such an extreme commitment? Pure strategic necessity.
Lip-Bu correctly recognized that to stay relevant in semiconductor design, Cadence had to move in lockstep with Moore’s Law, by constantly rebuilding its tools for the next node, including today’s push toward 2nm and lower. Each new process node demands fundamental re-architecture: the physics changes, design rules evolve, and optimization strategies are rewritten. You can’t just iterate on last year’s software and call it progress.
This aggressive reinvestment has paid off handsomely over time through sustained revenue growth and major customer wins. Despite this R&D intensity, Cadence has consistently delivered healthy margins with returns on incremental investment well above the cost of capital. If they weren’t, margins would be compressing, not expanding.
Cadence has also a history of pursuing disciplined M&A, buying add-on capabilities rather than big revenue deals (more on this later). Some of the key strategic acquisitions include:
Tensilica (mid-2010s): High-performance audio IP that’s now expanding into AI accelerators
AWR (2011): RF/microwave design
Sigrity and Celsius: Signal integrity and thermal analysis for system-level design
Various IP companies (since early 2010s): Built the foundation for today’s 15% IP business segment
And what they haven’t done matters too.
Unlike Synopsys, which acquired Ansys for $35B in a painful and arguably overpriced deal, Cadence has stuck to disciplined $50–500M acquisitions. That speaks volumes about management’s philosophy: a clear preference for organic growth, confidence in the core business, and an aversion to integration risk.
More importantly, it reflects valuation discipline. Cadence doesn’t need mega-deals to compound, the underlying engine is already attractive enough.
The Business Model: Understanding the Revenue Architecture
Cadence’s business model transformation represents one of the most successful subscription transitions in enterprise software history, predating Adobe, Autodesk, and others.
The old model (pre-2008) relied on perpetual licenses with unpredictable quarterly swings. Customers would pit vendors against each other, leading to destructive pricing and lumpy revenue.
The new model (post-2010) shifted to multi-year subscription contracts with ratable revenue recognition. Today, roughly 85–90% of revenue is recurring.
That shift transformed the unit economics.
Three-year per-seat EDA commitments reduced volatility. Multi-year agreements improved pricing discipline. Ratable revenue recognition brought predictability. And subscriptions aligned incentives around continuous value delivery rather than one-off transactions.
This flywheel has been spinning for quite some time now, with the inflection point around 2010, clearly visible in Bloomberg Intelligence data below. And despite today’s AI-driven fears around SaaS disruption, this trend is only set to strengthen over the next five years.
What makes this model so enticing is its counter-cyclical nature. The evidence is in the financials. Despite semiconductor industry cycles in 2008-2009, 2015-2016, and 2022-2023, Cadence’s slowest annual revenue growth has been approximately 6%. Remarkable consistency for a company serving a notoriously cyclical industry.
The Three Revenue Segments
Core EDA Tools (~60-65% of revenue, ~$2.8-3.2B annually)
These are the software tools engineers use daily to design chips. Digital IC design, custom/analog design (where Cadence’s Virtuoso platform is the industry standard for 30+ years), verification tools, and system design. Think of this as the Excel of chip design on asteroids: mission-critical, used every day, unwise to replace.
Hardware Platforms (~25-28% of revenue, ~$1.2-1.4B annually)
This segment is becoming increasingly important and represents Cadence’s strongest competitive position. The 🔗 Palladium emulation platform holds approximately 50-55% market share and has become mandatory for advanced node designs (5nm and below) where pure software simulation is insufficient.
Each Palladium system costs $5-15 million, with customers often purchasing multiple units. The hardware segment grows faster than software (15-20% CAGR vs. 10-12%) and has shown strong pricing power due to limited competition and performance criticality.
Intellectual Property (~10-12% of revenue, ~$500-600M annually)
Pre-designed circuit blocks that chip designers integrate into custom chips. Interface IP (PCIe, USB, DDR memory controllers), analog/mixed-signal IP, and processor IP from the Tensilica acquisition.
The IP business operates on dual revenue streams: upfront licensing fees plus ongoing royalties based on chip production volumes. While Cadence holds only 15-20% share (behind ARM and Synopsys), this segment grows 14-16% annually with improving margins as the IP library scales.
The Value Capture Asymmetry
One of the key compelling aspects of Cadence’s business model is the asymmetry between value created and value captured. The semiconductor industry generates $550 billion annually, built on a $10-15 billion EDA market, just 2-3% of total chip industry revenue. Yet EDA tools are non-negotiable.
Designing a leading-edge 3nm chip costs $750 million to $1 billion, with roughly 50-60% tied to design engineering time and EDA software. A chip design failure can wipe out hundreds of millions in investment and delay time-to-market by 12-24 months.
In that context, paying Cadence $2-5 million annually for best-in-class tools isn’t a meaningful expense, it’s insurance.
As one NVIDIA data center expert put it:
“EDA is a low tariff extractor on the growth of the semiconductor industry. It’s not a pain point for customers. They provide something super valuable, but it’s small enough that it doesn’t create resistance.”
This value asymmetry translates into modest but improving pricing power. Following the destructive pricing practices of the 2000s, Cadence has achieved low single-digit annual ASP growth while maintaining gross margins of 88-90%, among the highest in enterprise software.
What Cadence Tools Actually Do
Before diving into the economic moat, the growth story, and the capital allocation, it’s worth pausing to understand better what Cadence’s tools actually do day-to-day. I think it’s essential, because the depth of the competitive advantage becomes obvious once you understand the design flow.
Start with an Idea, Craft the Physical Layout
Modern chips contain billions of transistors. Each bundle of transistors that specifies a particular operation is known as a block. A medium-sized block can contain anywhere from tens of thousands to 1 million transistors. Engineers use a hardware description language to manually write what these blocks are supposed to do. But, figuring out how to physically arrange millions of transistors and route the microscopic wires between them, that’s where EDA software comes in.
The EDA tools decide where to lay out the transistors and how to route the wires, while accounting for factors like heat dissipation, signal speed, or potential signal loss. This step is about taking an “abstract chip architecture” and turning it into a “physical layout”. It is one of the many critical functions of EDA software.
But not every part of the chip can be automatically generated. Certain analog components such as power delivery ICs that distribute voltage, or memory cells for data storage, or even high-speed I/O interfaces that move data in and out of the chip. All these applications require manual design at the transistor level. The reason is that small variations in these components can break the entire chip. This is where Cadence’s Virtuoso and Spectre tools come in, allowing engineers to hand-craft and simulate these analog and custom blocks with precision that automated tools can’t yet match.
The Verification Process
Getting the initial physical layout done is like drawing up a gameplan that’s never been run in a real game. Once you start testing it, all sorts of issues surface that need to be fixed. This is where Cadence adds another layer of enormous value.
The checking process has four distinct phases, each addressing a different challenge:
Simulation and Verification: The problem here addresses the following question, “does the chip actually behave as intended”? 🔗 Cadence’s Xcelium, Spectre, and Jasper Formal Verification tools run the chip through thousands of scenarios to validate its logic before a single physical unit is manufactured.
Sign-Off based on Timing, Power, and Reliability: This problem is about figuring out whether the chip will still work under worst-case real-world conditions. A chip can behave correctly in simulation but fail if signals arrive too late, power delivery collapses, or heat degrades performance over time. 🔗 Cadence’s Tempus and Voltus tools analyze signal timing, power spikes, voltage drop, and reliability to catch these failure modes.
Manufacturability: This is all about answering a fundamental question for foundry customers: “can you actually build this design at scale or not”? A chip can be logically correct and physically stable but still impossible to manufacture reliably and at scale at advanced nodes. 🔗 Cadence’s Pegasus and PVS tools verify that the layout follows the strict manufacturing rules set by foundries like TSMC or Samsung.
Packaging Validation: The last problem, “does this chip still work once it’s packaged and placed into a real system”? The silicon doesn’t exist in a vacuum, it has to communicate with the outside world and survive heat, power, and signal integrity challenges beyond the chip itself. 🔗 Cadence’s Allegro, Sigrity, and Clarity tools simulate packaging, circuit boards, and system-level physics to validate the final product.
The Criticality of Real-World Simulation
Here’s what’s critical to understand: theoretical chip designs almost always have errors. The question isn’t whether problems exist, but whether you find them before or after fabrication. Finding them after costs hundreds of millions of dollars and months of delay. Sometimes it can torpedoe the entire project.
Engineers run simulations repeatedly to uncover three categories of failure. Timing failures occur when wires are too long, transistors slow down due to heat, or multiple blocks compete for power; causing the chip to read wrong values, run slower, or fail entirely. Power spikes happen when too many transistors turn on simultaneously or power delivery is inadequate, triggering shutdowns or data corruption. Signal interference occurs when wires are packed too closely together and electrical current distorts a neighboring signal, causing unpredictable bugs and random failures.
Cadence’s full stack of design and simulation tools addresses all of these failure modes across the entire design flow.
The thing is, this end-to-end coverage, from initial concept through physical layout, verification, manufacturability, and packaging, is precisely what makes switching away from Cadence so difficult once you’re embedded in the flow.
The Economic Moat: Switching Is Irrational
Cadence has two primary sources of moat: