CRUS — Cirrus Logic Inc.

Cirrus Logic has supplied custom audio and haptic silicon to Apple for fifteen years and is now winning new design slots in Face ID sensing — a signal that the world's most powerful technology company is deepening, not dissolving, its reliance on Cirrus. At 12.6 times enterprise value to free cash flow — with $1.2 billion in net cash, zero debt, and $635 million in annual FCF — the stock prices in near-zero growth for a business whose newest Apple design win hasn't even appeared in revenue yet. Compelling at the current price.

---

The semiconductor sector in 2026 is a study in bifurcation. AI infrastructure chips command premium multiples and attract daily investor attention. The companies supplying the chips that go into the billion smartphones shipped every year have been systematically de-rated. The consensus narrative holds that the smartphone market is ex-growth, that any company with 89% revenue concentration in a single customer is one product cycle away from catastrophe, and that fabless suppliers without AI exposure deserve discount multiples. This narrative has driven smartphone-adjacent semiconductor stocks to valuations that assume stagnation — even as several of them quietly compounded free cash flow through the same period when the consensus declared them stranded assets.

Cirrus Logic sits at the center of this misperception. The company makes the chips that process audio in every iPhone: the codecs that digitize voice for Siri and Face ID authentication, the boosted amplifiers that drive speakers louder and cleaner than physics suggests possible in a thin aluminum frame, the haptic drivers that make the iPhone's touchscreen feel like it has actual mechanical buttons. None of this is glamorous semiconductor work. There are no AI accelerators, no data center exposure, no cutting-edge logic node ambitions. What there is, quietly, is a business generating 32% free cash flow margins, holding $1.2 billion in net cash against zero debt, and serving a customer that has chosen them for fifteen consecutive product cycles.

The smartphone audio chip market is structurally bifurcated between a commodity segment — basic codecs selling for cents per unit to Chinese smartphone manufacturers — and a premium segment where design requirements are complex enough that only a handful of companies can meet them. Realtek dominates commodity PC and Android phone audio with solutions priced 15–30% below Cirrus Logic's comparable offerings. Texas Instruments and Analog Devices sell into professional audio and industrial markets with broader portfolios. These are different markets serving different customers with different performance requirements. The relevant question for Cirrus is not whether it can compete with Realtek on price — it cannot and does not try — but whether the technical complexity of premium smartphone audio has risen high enough to sustain the differentiation gap that keeps Cirrus at Apple's side.

The answer from the evidence of the past decade is yes, and the gap is widening. When Apple moved from basic audio codecs to codecs with embedded DSP for beamforming, noise cancellation, and voice biometrics — capabilities that require years of algorithm development, not merely silicon design — the addressable supplier pool collapsed from dozens of vendors to essentially one. Cirrus Logic's 22-nanometer smart codec represents approximately four to five years of R&D advantage over the nearest alternative. Qualcomm integrates audio into its Snapdragon system-on-chip, which is compelling for Android OEMs but absent from iPhones where Apple's custom silicon leaves no room for Qualcomm's audio stack. Realtek makes codecs adequate for laptops and budget phones but has never demonstrated the algorithm depth or process node capability required for premium smartphone deployment. The overall audio IC market is large — estimated at $32–36 billion globally in 2024, growing toward $62 billion by 2033 — but Cirrus Logic competes in the narrow, high-value tier where performance is the purchase decision, not price.

Cirrus Logic is a fabless semiconductor company headquartered in Austin, Texas, with 1,660 employees and no manufacturing of its own. Chips are designed in-house and fabricated by TSMC and GlobalFoundries. The business has two product segments: audio ICs, which include smart codecs and boosted amplifiers (57% of Q4 fiscal 2026 revenue), and high-performance mixed-signal, or HPMS, which includes haptic drivers, camera controllers, and power management ICs (43% of Q4 fiscal 2026 revenue). Apple accounts for approximately 89–90% of annual revenue. The remaining 10–11% comes from PC OEMs, automotive customers, and professional audio applications.

The company earns money in a single way: Apple pays a per-unit price for each chip that goes into an iPhone or other Apple device. That price has increased over time as the complexity of the chips has grown. A basic codec in 2014 was worth a few dollars of content per device; a smart codec plus boosted amplifier plus haptic driver in 2026 is worth approximately $4.20–$4.30 per iPhone. The business carries no recurring revenue, no subscription model, no annuity stream — it is a pure component supplier whose fortunes are tied to how many iPhones Apple ships and how much silicon Apple elects to put in each one. What makes this more interesting than it sounds is that Cirrus has expanded content per device in virtually every cycle. The HPMS segment — haptic drivers, camera controllers, power ICs — grew from roughly a third of revenue in fiscal 2022 to 43% in fiscal 2026. Every time Apple added a feature requiring analog precision — better haptics, faster face recognition, improved camera — it returned to Cirrus for the enabling silicon.

The moat is narrower than the language of "deep Apple relationship" implies but harder to break than the Apple concentration statistics suggest. The technical differentiation is specific. Cirrus's Halo Core DSP, embedded in its haptic drivers, implements closed-loop control of the linear resonant actuator: it monitors the actuator in real time, adapts waveforms to maintain consistent haptic feel as battery voltage drops, and responds in four milliseconds. No competitor has commercially deployed a closed-loop haptic architecture in a mass-market smartphone. Open-loop alternatives from Texas Instruments and others exist and are cheaper; they cannot replicate the consistency of feel that Apple requires for its Taptic Engine experience. The switching cost here is not merely financial — it is technical. A competitor seeking to displace Cirrus would need years of algorithm development, integration work, and Apple qualification before a single chip shipped.

The audio codec story is similar. The CS42L92 family supports 384kHz sample rates, multi-channel beamforming to six meters, wind noise reduction, stereo ambient-noise cancellation, and simultaneous voice and media processing — all while meeting Apple's extremely tight power budgets. The integration of these capabilities into a single 22-nanometer die, codesigned with Apple's software stack over years of collaboration, creates a switching cost that cannot be expressed in dollar terms alone. Cirrus holds 3,900-plus patents in audio DSP and mixed-signal design, representing not merely legal protection but accumulated engineering judgment that is genuinely difficult to replicate through acquisition or internal development.

Company	Gross Margin	Manufacturing Model	Primary Market
Cirrus Logic	52.8%	Fabless (TSMC, GF)	Premium smartphone audio, haptics
Texas Instruments	58.0%	Vertically integrated	Broad industrial, consumer
Analog Devices	54.7%	Mixed (fab + fabless)	Industrial, professional audio
Realtek	50.0%	Fabless	Commodity PC and Android audio

Cirrus's 52.8% gross margin is not the highest in the table — TI and ADI benefit from broader portfolios and, in TI's case, its own fabs amortized across more products. But the margin has been stable through cycles. When revenue dipped from $1.90 billion in fiscal 2023 to $1.79 billion in fiscal 2024 due to weak iPhone demand, gross margins did not compress. A supplier with negotiating power does not absorb revenue declines in its margins. Cirrus did not.

The bear case on the moat rests on Apple's documented history of in-sourcing key components. In 2018, Apple acquired Dialog Semiconductor's power management business for $600 million, eliminating Dialog's primary revenue source within two years. In 2017, Apple announced it would phase out Imagination Technologies' GPU IP; Imagination's stock collapsed approximately 70% on the announcement. Both substitutions followed the same pattern: Apple identified a technically defined problem, acquired the IP and engineers to solve it internally, and replaced the supplier. The question is whether audio DSP is analogous to power management ICs or GPU IP — well-defined problems amenable to internalization — or whether it is a different category of problem. Power management is electrical engineering against a set of defined specifications. GPU IP is mathematical architecture that Apple could recreate once it understood the approach. Audio DSP with beamforming, noise cancellation, and haptic closed-loop control requires years of tuning against real acoustic environments, microphone arrays, and actuator physics that vary device by device, ear by ear. The 3,900-plus patents Cirrus holds represent not merely legal barriers but engineering judgment accumulated through thousands of real-world deployments. And the Face ID design win — Apple formally confirming Cirrus as a strategic supplier for sensing functions in early 2026, with a committed $400 million manufacturing investment alongside Cirrus at GlobalFoundries' Malta, New York facility — is not the behavior of a company planning to phase out the relationship. It is the behavior of a company building it.

Cirrus Logic reported $2.00 billion in revenue for fiscal year 2026 (ending March 2026), a record, up 5% from $1.90 billion in fiscal 2025. GAAP net income was approximately $408 million, yielding GAAP EPS of $7.85. Non-GAAP EPS was $9.26 — the reconciling difference of approximately $1.41 per share, or roughly $73 million in aggregate, consists primarily of stock-based compensation and amortization of acquired intangibles. Stock-based compensation is a real cost of retaining the engineering talent that creates the moat and should not be dismissed when assessing normalized earnings. Operating free cash flow was $635.8 million in fiscal 2026, up 46% from $421.6 million in fiscal 2025. Capital expenditure was $14.8 million — a fabless model requires almost no physical capital. The balance sheet at fiscal year-end shows $1.2 billion in cash and investments and zero debt. Inventory stood at $240.9 million, down from $299.1 million the prior year, suggesting inventory management has normalized after prior cycle excesses.

R&D spending in fiscal 2025 was $434.7 million — approximately 22% of revenue devoted entirely to maintaining and extending the technical lead. This is high for a component supplier and is not optional. A company that stopped investing in audio DSP algorithm development would see its advantage erode within three to five years as competitors caught up on the technical specifications, if not on the accumulated engineering intuition. The R&D line is better understood as the ongoing cost of the moat than as a discretionary expense subject to margin optimization.

John Forsyth became CEO in January 2021, having joined Cirrus through the 2014 Wolfson Microelectronics acquisition. His tenure has been characterized by consistent execution on the content-expansion story within Apple and the early stages of diversification into PCs and automotive. Share count has declined from approximately 59 million in fiscal 2022 to 51 million by May 2026, a 14% reduction achieved through $500 million buyback programs authorized in August 2022, March 2025, and May 2026. In fiscal 2026, $280 million was deployed in actual repurchases against $635.8 million in free cash flow — a 44% return-of-capital ratio with no debt to service. The 2021 acquisition of Lion Semiconductor for $335 million brought fast-charging power management IP that now drives a meaningful portion of the HPMS segment, and the integration appears to have been productive. Insider ownership is low at 1.16%, and trailing twelve-month activity shows twelve insider sells and zero insider buys — not an encouraging ownership profile, though consistent with semiconductor companies where equity compensation is high and insiders manage concentration risk through systematic sales rather than conviction accumulation.

The growth runway is where the analysis becomes most important and most uncomfortable simultaneously.

Fiscal Year	Revenue ($B)	Apple % of Revenue	Non-Apple Revenue ($M)
FY2022	$1.78	79%	$374
FY2023	$1.90	82%	$342
FY2024	$1.79	87%	$233
FY2025	$1.90	89%	$209
FY2026	$2.00	~90%	~$200

The table makes a point that management's investor presentations do not. Total revenue grew from $1.78 billion to $2.00 billion between fiscal 2022 and fiscal 2026 — an 12% increase over four years. Over the same period, non-Apple revenue fell from $374 million to approximately $200 million, a decline of nearly 47% in absolute dollars. Apple concentration moved from 79% to approximately 90%. The diversification strategy exists, has real products behind it, and is generating actual revenue — but it has not kept pace with the erosion of the non-Apple base. This is the decisive gap between management's narrative and what the numbers show, and it is the single most important fact for anyone underwriting the Apple relationship as durable.

What drove the non-Apple erosion is itself instructive. The company lost audio content in the Android smartphone market as Chinese OEMs gravitated toward cheaper alternatives and Qualcomm integrated audio functions into its SoC for mid-range devices. The professional audio and industrial segments have been growing but from a small base. The PC story, meanwhile, is only now showing up in meaningful revenue: fiscal 2025 PC revenue was in the low single-digit millions; fiscal 2026 grew to approximately $40 million; management guided for another doubling in fiscal 2027. The company has secured design wins across 150-plus laptop SKUs, holds approximately 75% win rate in the new SDCA audio architecture that is transitioning the laptop market to higher-performance audio, and targets a $1.2 billion serviceable PC market by 2029. At approximately $40 million in current PC revenue, Cirrus has captured roughly 3% of that addressable market — 97% remains unaddressed, and the design win pipeline suggests the architecture transition is proceeding faster than the revenue curve currently shows.

The Face ID win changes the structure of the runway argument more than any single data point. Apple, in March 2026, committed $400 million alongside Cirrus to establish U.S.-based mixed-signal semiconductor manufacturing at GlobalFoundries' Malta, New York facility — specifically for Face ID sensing and other Apple product categories. The smart power IC for Face ID carries an estimated ASP of approximately $2 per unit. At 90 million Pro iPhone units, the Face ID win represents approximately $180 million in incremental annual revenue when it ramps, expected in fiscal 2028–2029. That incremental revenue would represent a 9% increase on current fiscal 2026 levels — from a single new design win, in a product category Cirrus did not participate in twelve months ago. Cirrus has captured approximately 27% of its current $7.4 billion serviceable addressable market. The remaining 73% — across PC, automotive ($1.7 billion SAM today, growing to $2.2 billion by 2030), consumer electronics, and the Face ID ramp — is either nascent or only beginning to show up in reported results.

At $173.74 per share as of May 11, 2026, Cirrus Logic carries a market capitalization of approximately $8.75 billion and an enterprise value of approximately $8.0 billion after subtracting $1.2 billion in net cash. Against fiscal 2026 free cash flow of $635.8 million, the EV/FCF multiple is 12.6 times. Excluding the net cash, the operating business trades at approximately 10.7 times free cash flow. Against GAAP EPS of $7.85, the GAAP P/E is 22.1 times. Against non-GAAP EPS of $9.26 — where the reconciling items are primarily stock-based compensation — the non-GAAP P/E is 18.8 times.

The question of whether this business earns high returns on unleveraged tangible capital is answered quickly by the fabless structure: with $14.8 million in annual capex against $635.8 million in free cash flow, the tangible capital intensity is near zero. Working capital (principally inventory of $241 million plus receivables) constitutes the effective tangible asset base. The returns on that base are exceptional. This is a business whose value lives in its IP, its algorithms, and its customer relationships — none of which appear on the balance sheet. The $1.2 billion cash balance and zero debt position mean the enterprise value captures only the operating business at 10.7 times FCF; the cash is a separate asset.

The most intelligent bear argues that the face ID win is evidence not of a deepening relationship but of a deliberate Apple strategy to keep Cirrus just dependent enough on new design wins that the supplier can never afford to walk away, while Apple continues building internal audio competency that will eventually displace the legacy position. The answer to this objection is that it requires Apple to have been executing this strategy for fifteen years against a supplier who has simultaneously won new positions in haptics, camera power management, and now sensing — while also making a $400 million capital commitment to co-manufacture with Cirrus in the United States. That is an elaborate and expensive strategy for a company whose prior supplier replacements (Dialog, Imagination) were completed in under three years. The simpler explanation is that Cirrus's technical depth in audio DSP is genuinely difficult to replicate, that Apple has tried the internal option on simpler problems and found it works there, and that audio algorithms require a different kind of institutional knowledge than power management circuits. The Face ID win is better read as Apple choosing a proven DSP partner for a new sensing problem than as a Trojan horse for eventual displacement.

What would change this conclusion? Two things could alter it negatively: Apple announcing internal development of audio codec capabilities — not a rumor, but an engineering hire pattern or supplier notification analogous to what preceded the Dialog acquisition — or a sustained multi-year decline in iPhone unit volumes that erodes the revenue base faster than PC and automotive diversification can offset it. The second risk is real but partially hedged by the content-per-device expansion story: if Apple sells fewer iPhones but puts more Cirrus silicon in each one, the revenue impact can be muted. The first risk is the existential scenario, and the Face ID investment argues against it for at least the next three to five years.

For the conclusion to turn more optimistic, the PC doubling in fiscal 2027 needs to arrive as guided, the Face ID ramp needs to proceed on the fiscal 2028–2029 timeline, and non-Apple revenue in absolute dollars needs to stabilize and then reverse the five-year declining trend shown in the table above. None of these require heroic assumptions. They require the company to execute what is already in the design win pipeline.

At 10.7 times operating FCF with $1.2 billion in undeployed cash, the market is pricing Cirrus Logic as though the Face ID win will not happen, the PC ramp will stall, and Apple will eventually do what it did to Dialog. The evidence for each of these outcomes is weaker than the market multiple implies. A business generating 32% free cash flow margins, holding net cash equal to 14% of its market cap, expanding into new Apple product categories, and trading at 10.7 times ex-cash operating FCF is not priced for what it actually is. It is priced for something worse.