Cadence Raises Guidance to 19% Growth as Intel Partnership Deepens and Agentic AI Moves From Pilot to Production
Q2 2026 earnings call, July 27, 2026
Cadence Design Systems delivered a quarter that management repeatedly described as its broadest in years, prompting the company to raise its full-year revenue growth guidance to 19% alongside higher profitability targets. Second-quarter revenue reached $1.584 billion, up 24% year-over-year, with every product group posting double-digit growth and bookings driving a record backlog of $8.1 billion. CEO Anirudh Devgan called it "the highest raise we ever had" for a single quarter, and CFO John Wall was explicit that the strength was not concentrated in one customer or product line but spread across core EDA, IP, hardware and system design and analysis.
Intel relationship normalizes after a decade of distance
The most notable strategic development was a newly expanded, multiyear engagement with Intel tied to its 14A process node. Devgan was unusually candid about the history behind this deal, calling the lack of a closer Intel relationship "a 10- or 20-year-old problem" that has now been resolved under Lip-Bu Tan's leadership. The agreement spans design IP, agentic EDA and DTCO co-optimization for next-generation HPC and mobile designs, and management indicated engagement now extends across multiple parts of Intel, including its product groups. Devgan characterized the deal as "more a normalization of our relationship with Intel, like we work with all the other household names," suggesting Cadence views this as closing a long-standing competitive gap versus its positioning at TSMC. Management confirmed the Intel revenue is entirely incremental to prior arrangements, though most of the benefit is still to come rather than reflected in this year's numbers. A similar dynamic played out with Samsung Foundry on 2-nanometer and 3D-IC technology, and with Rapidus, which is integrating Cadence's InnoStack agent into its own AI design offering.
Agentic AI push shows early monetization but no step change yet
Cadence now has four "Super Agent" products live in the market — AuraStack for PCB and advanced packaging, ChipStack for verification, ViraStack for analog and custom design, and InnoStack for advanced-node SoC design — collectively representing roughly six months of customer deployment since their Q1 launch. ChipStack has more than 20 customer engagements with production deployments across multiple chip designs, and a version co-developed with Nvidia reportedly compressed a five-week RTL verification cycle to under a day, a 40x improvement. ViraStack has more than 25 engagements showing 2x to 10x productivity gains. Devgan reiterated his long-standing "Three-Layer Cake" framework — compute and data at the base, physically accurate simulation engines in the middle, and AI agents and orchestration on top — arguing that agentic adoption expands Cadence's addressable market rather than cannibalizing existing tool usage, since agents "call our underlying physically accurate engines more often." Importantly, Wall pushed back on the idea that this is already driving the raise in a major way, telling analysts "we're still not assuming a sudden step function in our guidance," even as engagement, evaluations and pilots continue to accelerate.
Asked directly about the recent case of the Kimi open-source model autonomously invoking EDA tools to complete a chip design task, Devgan downplayed the threat to Cadence's moat, noting the design in question involved "a small block" on a node roughly 20 to 30 times slower than current leading-edge frequencies. He argued the episode actually validates his three-layer thesis rather than undermining it, since even a trivial design required proprietary EDA infrastructure to execute. On the fear of an end-to-end LLM generating tape-out-ready GDSII files without commercial EDA tools, Devgan was dismissive, predicting instead that competition will concentrate at the orchestration layer among large language models rather than commoditizing the physically accurate engines Cadence controls: "The ground truth will prevail."
IP business becomes a standout growth engine
The IP segment grew more than 40% year-over-year, a result management attributed mostly to organic execution rather than acquisitions. Devgan pointed to three structural drivers: materially improved power-performance-area quality of Cadence's IP at leading nodes, which is winning competitive slots the company "would not participate in" two years ago; a deliberate focus on high-value interface and memory IP tied to AI and HPC workloads, including PCIe, UCIe, HBM and LPDDR6; and an increasingly diversified foundry ecosystem spanning TSMC, Intel, Samsung and Rapidus. Both the IP and system design and analysis businesses are now approaching roughly $1 billion in annual run rate each, giving Cadence what Devgan called "a lot of strength in our portfolio to engage with our customers." Wall cautioned that IP revenue can be lumpy quarter to quarter and should not be annualized off a single strong print.
Hardware remains supply-constrained, not demand-constrained
The hardware business, anchored by Palladium Z3 and Protium X3, posted another record quarter with 12 new customer logos and a competitive win against a major AI infrastructure provider. Wall described hardware as "still remains supply constrained by customer demand rather than demand constrained," with the company building systems as fast as possible to work through backlog. Management reaffirmed expectations for 2026 to be another record hardware year, driven primarily by hyperscalers and leading semiconductor companies using hardware-assisted verification as what Devgan called "a strategic capacity layer" for increasingly complex AI silicon.
Margins dip slightly in second half on deliberate investment
Full-year guidance now calls for revenue of $6.260 billion to $6.340 billion, non-GAAP operating margin of 43.75% to 44.75%, and non-GAAP EPS of $8.05 to $8.15, with operating cash flow of approximately $2 billion. Wall flagged that second-half margins will run slightly below first-half levels, not due to any deterioration in the underlying model but because of targeted investment in integrating Hexagon's design and engineering business and in ramping the Intel relationship. He said the goal is to improve profitability heading into 2027, noting "organic incremental margins remain very attractive" and that acquisition-related and IP profitability should continue improving next year. System design and analysis revenue, which includes the Hexagon D&E integration, grew 37% year-over-year, with Hexagon contributing roughly 4 points to overall recurring revenue growth of 24%.
On bookings, Wall noted this year falls in a structurally lower point of Cadence's roughly three-year renewal cycle, which makes the first-half bookings strength — up an implied 55% year-over-year by one analyst's estimate — more notable, driven largely by add-on business rather than renewals. Devgan summarized the setup for the back half of the year as resting on three pillars: an improved macro environment across both AI-native and traditional semiconductor customers, a competitive position he described as the strongest in the company's history, and the still-early Agentic AI opportunity layered on top.