SiTime Corporation (NASDAQ: SITM) — A Category-Defining Timing Franchise Priced for an Uninterrupted AI Decade

Independent equity research · Report date: 2026-06-20 · Price: $729.04 (2026-06-18 close)

⚡ Claude’s Take

This block is the author’s own independent opinion and general information only — not investment advice. The analysis that follows takes no position, sets no price target, and carries no recommendation; the single subjective view is fenced into this block.

Verdict: AVOID at $729 · not-a-short · accumulate-on-deep-weakness ~$300–450 · fair-value zone ~$400–550 (≈18–24x forward pro-forma EV/sales, ≈25–32x forward non-GAAP EPS). Conviction: medium.

SiTime is a genuinely good business — the company that created the MEMS precision-timing category and now owns it, with a real demand-side qualification moat, 53–65% gross margins, and a legitimate, multi-year AI-datacenter content tailwind that is showing up right now in the numbers (CED revenue +158%, eighth straight triple-digit-growth quarter, FY26 guided to ≥80% growth). None of that is in dispute. The problem is entirely price and pricing of risk. At $729 the stock trades at roughly 48–56x trailing and ~28–31x forward EV/sales — the richest revenue multiple in the entire semiconductor comp set, two-to-three times the closest profitable analog (Monolithic Power) and above ARM — for a business that has booked a GAAP net loss in every year but two, whose entire reported “profitability” and “free cash flow” exist only before charging stock-based compensation running at ~32% of revenue, which has just flipped from net cash to net debt to fund a $1.5B acquisition, and which fell −49% in a single year (2023) the last time its end-demand digested. The framing is not value, not quality, and not a falling knife — it is a crowded, beta-2.67 AI-thematic momentum melt-up (y1 +220%, a −78% five-year max drawdown on the record), into which insiders and the largest holder have sold ~$379M year-to-date with zero open-market buys. You are being asked to underwrite flawless multi-year execution, an SBC normalization that has not begun, clean integration of a levered carve-out, and the persistence of a ~30–50x revenue multiple — with no margin of safety on any of them.

The tag: “the heartbeat of AI, priced as if the cycle has been repealed.” This is a great company I would happily own far lower; at $729 the risk/reward is asymmetric to the downside, but I will not short a thin-float, real-franchise momentum name with a sponsor backstop and live squeeze risk. Flips bullish on a de-rate into the low-$300s–$400s with CED growth still intact, or two-to-three quarters proving GAAP-positive economics net of SBC with the multiple already reset. Flips bearish (toward outright avoid/short-watch) on a CED sequential roll-over or a 2023-style guide cut while the stock is still priced for perfection.

📈 Stock Price Action — Five-Year Event Map

Factual price history — no recommendation, no price target. Price moves are Fact (AZI five-year price series); attributed causes are Interpretation.

The arc. SiTime is a full semiconductor-cycle round-trip overlaid with a parabolic AI re-rating. From a $13 IPO (November 2019) it ran to a $292.50 year-end-2021 close, crashed to $101.60 (year-end 2022) in the chip downturn, bottomed near a 52-week intraday low of $186.49 (2025-08-01), then exploded to an all-time intraday high of $901.81 (2026-05-11). It now trades at $729.04 (2026-06-18), roughly −19% off that high, against a 52-week range of ~$186 to ~$902. The stock has nearly quadrupled in twelve months and is one of the most volatile names in its sector (beta 2.67).

Cycle narrative. (1) From the $13 IPO, MEMS-oscillator adoption and the 2021 up-cycle drove a ~22x advance into the risk-appetite peak. (2) 2022 erased two-thirds of it as the broad semi downturn, destock, and rate-driven de-rating hit high-multiple chips — the lived precedent for how violently SiTime reprices. (3) 2023–24 was trough-and-recovery: FY23 revenue fell −49%, the stock based, and the December-2023 Aura deal seeded the clock business. (4) The ~$186 August-2025 low was the launchpad before the AI leg. (5) On 2026-02-04 a strong Q4/FY25 print plus the Renesas-deal announcement gapped the stock ~18%. (6) The 2026-05-06 Q1-26 report (revenue +88%, CED +158%, raised guide) drove a single-day ~28% surge. (7) The $901.81 high on 2026-05-11 capped the melt-up, with HSR clearance removing the last deal condition. (8) The pullback to ~$729 tracked the $1.35B convertible’s marketing/close, rising bond yields, and sector-wide chop — recent news flow is dominated by macro “semis up/down” items, consistent with the stock now trading on the thematic tape rather than idiosyncratic news.

1. Executive Summary

SiTime is the inventor and dominant supplier of MEMS (silicon) precision timing — the resonators, oscillators (XO/TCXO/Super-TCXO/OCXO), and clock ICs that synchronize electronic systems — displacing a fragmented, decades-old quartz-crystal incumbency. It is fabless: MEMS resonator die come essentially from a single foundry (Bosch), analog ICs from TSMC/UMC on legacy nodes, and back-end from OSATs. The company created the category and holds an estimated ~90%+ of MEMS timing, but that is a low-single-digit slice of an ~$11B total timing market still dominated by quartz; management targets a ~$4B “precision timing” served market.

The investment tension is sharp. The business is real and improving; the equity is priced for perfection. Revenue has compounded violently and cyclically — $116M (2020) → $284M (2022 peak) → $144M (2023, −49%) → $327M (2025, +61%) — and is now inflecting hard on AI-datacenter timing content: Q1-2026 revenue was $113.6M (+88% YoY), the Communications/Enterprise/Datacenter (CED) segment grew +158% (its eighth consecutive triple-digit quarter and now two-thirds of revenue), and management raised FY26 guidance to ≥80% growth with a target of sustainable 65% gross margin / 30% operating margin from Q2-2026. The AI content thesis is genuine: inference infrastructure carries 2–4x the timing content of training, 1.6T optical modules and co-packaged optics multiply content, and aerospace/LEO-satellite sockets are a credible new high-ASP vertical.

But the quality of the reported economics is the crux. SiTime has posted a GAAP net loss every year except 2021 and 2022. FY25 GAAP net loss was −$42.9M on a −20.5% operating margin, and the entire loss is stock-based compensation: $103.5M, ~32% of revenue. The “$35M of free cash flow” is an SBC add-back artifact (SBC alone is ~3x it); real cash generation net of the cost of dilution is negative, and the share count has risen ~63% in five years with no offsetting buyback. The balance sheet — a $808M net-cash fortress at year-end 2025 — has just flipped to net debt via a $1.35B 0% convertible (due 2031, conversion ~$1,040) raised to fund the ~$1.5B-cash-plus-4.13M-share acquisition of Renesas’s timing business, a lower-margin clock carve-out whose standalone economics are undisclosed.

At $729 the equity capitalizes at ~$19B (pro-forma EV ~$22–23B). On the only workable lens — EV/sales, since there is no GAAP profit — that is ~48–56x trailing and ~28–31x forward, the richest in the comp set, 2–3x Monolithic Power and above ARM. The price embeds sustained 30%+ growth for years to a multi-billion revenue base, a margin/SBC normalization that has not begun, clean Renesas accretion, and a persistently extreme multiple. Insiders and largest holder MegaChips have sold ~$379M year-to-date with zero open-market purchases. The factor profile is a crowded, beta-2.67, highly idiosyncratic AI-thematic momentum vehicle — not value, not quality, not a falling knife. This memo takes no position; it argues that the business is a high-quality, genuinely-moated franchise riding a real-but-cyclical AI content wave, and that the price discounts the optimistic version of that future with no margin of safety. The decisive open question is whether the 2025–26 CED surge is a permanent secular baseline or a cyclical AI-capex peak — the same question 2022 answered badly.

2. Business Overview

What SiTime makes. SiTime designs and sells precision timing components — the reference signals that keep electronic systems synchronized. The portfolio spans the timing stack:

1. Oscillators — the historical core and still the bulk of revenue. A MEMS resonator die plus an analog/mixed-signal IC in one package, generating a stable clock. The family runs from simple XOs and µPower 32 kHz devices to high-end TCXO / Super-TCXO (temperature-compensated), OCXO (oven-controlled), VCXO, and DCXO parts. The flagship Elite and new Elite 2 Super-TCXO families target AI infrastructure (Elite 2 claimed “up to 3x better synchronization” than Elite). (FACT — FY25 10-K; Q1-26 call.)
2. Clock ICs — network synchronizers, jitter cleaners, clock generators, buffers. Entered via the December 2023 Aura Semiconductor asset purchase (~$148M fixed + up to $120M earnouts); a relatively small portfolio (“around 50 clocks”). (FACT — FY25 10-K; Q1-26 call.)
3. Standalone MEMS resonators — the Titan platform (launched Sept-2025), claimed to improve resonator performance “by 100x” and enable semiconductor-level packaging. (FACT — FY25 10-K.)
4. Synchronization software — the TimeFabric suite, IEEE-1588 + holdover, sold embedded within products, not separately (no standalone software revenue line). (FACT — FY25 10-K.)

As of 12/31/25 SiTime had shipped 40,000+ unique part numbers into 400+ applications. (FACT.)

MEMS vs. quartz — the source of differentiation. A quartz oscillator pairs a mechanically-cut quartz resonator with a separate clock IC in a ceramic/metal package. SiTime instead fabricates its resonator in silicon using proprietary MEMS processes (MEMS First®, EpiSeal®, TempFlat®), then integrates it with its own analog IC — both die are silicon, so it can package, integrate, and software-tune frequency like any chip. The 10-K claims concrete physical advantages: wider frequency range, far better resilience to vibration/shock/temperature (“up to 50x better acceleration sensitivity under vibration”), programmability, smaller size, lower power. (FACT for the claims; magnitude is management INTERPRETATION.) Critically, SiTime designs all four blocks in-house (MEMS + analog + integration + software); quartz vendors typically outsource the analog/packaging and clock-IC vendors depend on third-party quartz references. This vertical, all-silicon design is the genuine technical distinction.

Revenue model. Product sales, not royalty. There is no material licensing stream; in fact SiTime historically paid Bosch a MEMS royalty, an obligation that expired March 31, 2024 (freeing dual-sourcing). Parts are batch-produced then custom-programmed to a customer’s spec, enabling weeks-long lead times and tens of thousands of SKUs from a common platform. (FACT — FY25 10-K.)

Segmentation. SiTime reports one segment but discloses three end markets (Q1-26):

(FACT — Q1-26 call.)

Customer & supplier concentration. SiTime sells primarily through distributors. FY25 >10% distributors: Arrow 26%, Pernas Electronics 25%, Quantek 13%; product through Pernas/Quantek largely ends up in Apple, making Apple the largest end customer at ~17% of FY25 revenue (22% FY24, 21% FY23) — trending down as CED diversifies. Manufacturing is fabless and concentrated: MEMS die from Bosch (sole/primary, 10-year agreement, initial term through Feb-2027, ~3 months safety stock; Teledyne second source), analog ICs from TSMC/UMC on legacy 180/150/130nm, back-end via OSATs. (FACT — FY25 10-K.)

The MegaChips history. SiTime was acquired by Japan’s MegaChips in 2016, then IPO’d November 2019. MegaChips was both controlling shareholder and an Asian distributor channel; it sold down its stake in 2021 secondaries and continued selling into 2026 (below). It is no longer a >10% customer; the distributor relationship is wound down — historical context, not a live dependency. (FACT — 10-K history.)

Recurring vs. non-recurring. No contractual recurring revenue (no subscriptions/royalties), but the economics behave semi-recurringly: a 6-month-to-3-year design cycle wins a socket that then ships for a 10-year+ product life. (FACT/INTERPRETATION.)

Verdict. A genuinely differentiated, vertically-designed timing franchise that created a category and is riding an AI-datacenter inflection — but concentrated on every axis: single dominant product line (oscillators), customer (Apple ~17%, three distributors ~64%), supplier (Bosch), and now end-market (CED two-thirds). The Renesas deal is an attempt to build the clock leg into a real second franchise.

3. Industry Dynamics

Structure and size. Timing is unavoidable in essentially all electronics. Management sizes the total timing TAM at ~$11B, within which it targets a ~$4B “precision timing” SAM — the high-performance high-end. Historically the resonator/oscillator base is quartz: a 50±year-old, “largely unchanged,” fragmented, commodity-leaning technology. The CEO’s own 2010 framing — “a highly fragmented commodity-oriented timing industry” — is telling: a commodity, fragmented industry is, in Greenwald terms, the default low-return structure. (FACT — Q1-26 call; FY25 10-K.)

The incumbents being displaced. The 10-K names: TXC, Daishinku (KDS), Kyocera, Seiko Epson, Nihon Dempa (NDK), Rakon, Murata, Abracon, Microchip, Diodes, Skyworks, Texas Instruments, and Renesas. Two groups: (a) quartz crystal/oscillator houses (TXC, NDK, KDS, Kyocera, Epson, Rakon, Abracon) — large, mostly Asian scale manufacturers of a maturing technology competing on price/volume; (b) clock-IC / mixed-signal players (TI, Skyworks, Microchip, Renesas, Diodes) whose clock products depend on a third-party quartz reference. The 10-K is candid that “many of our competitors are substantially larger, have greater financial, technical, marketing, distribution resources, significantly better brand recognition and broader product offerings.” SiTime is the small disruptor, not the scale player. (FACT — FY25 10-K.)

Profit pools and intensity. The quartz base earns commodity economics; the differentiated high-performance niche SiTime created earns 50–65% gross margins. Competitive intensity is rising on two fronts — quartz incumbents pushing up-market, and clock-IC consolidation. The 10-K warns competition will “increase and intensify… as internal resources of large OEMs grow,” risking “price pressure, reduced gross margins, and loss of market share.” (FACT.)

Switching costs and barriers. Real but uneven. A timing part is qualified into a system over a 6-month-to-3-year design cycle, then ships for up to 10 years; re-qualifying a substitute “could take several quarters or longer.” That is genuine demand-side switching cost at the socket level. The barrier to being SiTime is high (proprietary Bosch-co-developed MEMS process, 147 issued US patents expiring 2026–2043, six MEMS generations of process learning). But the barrier to competing in timing generally is low — quartz is mature and widely available from dozens of suppliers. (FACT.)

AI-datacenter as the structural driver. The bull case concentrates here, and the demand evidence is concrete (Q1-26 call): inference on newer XPUs needs 2–4x more timing content than training, and GPU utilization (targeting 50–60%) depends on synchronization, driving high-ASP Super-TCXOs; 1.6T optical modules ramp in 2026 with higher-priced oscillators than 800G; co-packaged optics carry “up to 3x” content; AI-telecom (5G RAN, fixed wireless) “3x” content; LEO satellites “up to $2,000 of SiTime content per satellite,” 7,000–10,000 launches over three years. This is a content-per-system tailwind, not pure unit growth. (FACT for the claims; durability is INTERPRETATION.)

Marathon capital-cycle read. Skepticism is warranted. The AI-timing pool is now visibly attractive and every adjacent player is pointing capacity at the same sockets — quartz incumbents moving up-market, clock-IC vendors consolidating. Management concedes it “shares” optical-module sockets with quartz suppliers and is “not fully penetrated.” SiTime is playing the offensive side of the cycle — consolidating the clock leg via M&A (Aura 2023; Renesas 2026) rather than building merchant capacity, the smarter cycle move, but now debt-funded. A “white hot” order book during an industry-wide capex spike is precisely the late-cycle signal Marathon flags — and the 2022 peak → 2023 −49% sequence is the standing precedent for treating a hot timing order book as permanent. (INTERPRETATION.)

Verdict: mixed — structurally good for the qualified high-end incumbent, structurally average/commodity for the broad market. SiTime occupies the attractive, high-margin, qualification-barriered precision niche it created (50–65% gross margins), but the broad timing market is mature, fragmented, and price-competitive, and the high-end niche is drawing a competitive-supply response. The demand is real; the current level of growth is elevated by an AI capex cycle that will eventually digest and invite supply.

4. Competitive Position

Name the mechanism (Greenwald). SiTime’s advantage is bifurcated and narrow — a supply-side / proprietary-technology barrier plus a demand-side customer-captivity (switching-cost) advantage, with no economies-of-scale advantage (SiTime is the small player). Neither element is as wide as “we created the category” implies.

1. Supply-side / proprietary technology — real, but a single point of dependence. SiTime is effectively the only scaled all-silicon MEMS timing supplier. The barrier is the Bosch-co-developed proprietary MEMS process, 147 patents, six MEMS generations, and in-house ownership of all four blocks. A new entrant cannot easily replicate this — a genuine supply-side barrier in MEMS timing. But the same fact is a concentration risk: the MEMS die comes essentially from one foundry. The Bosch relationship is the moat and the largest single operational vulnerability (Bosch can decline POs; initial term Feb-2027; ~3 months safety stock). The royalty expiry (3/24) and Teledyne second-source modestly de-risk it, but Bosch dependency caps how “fortress-like” the moat is. (FACT + INTERPRETATION.)

2. Demand-side captivity — real at the high end, contestable in the mainstream. Once qualified, SiTime ships for a 10-year+ product life, and requalifying a substitute “could take several quarters or longer.” In mission-critical sockets (AI datacenter where timing failure corrupts data / shuts down networks; aero/defense; automotive ADAS) switching cost is amplified by uptime/qualification risk — a genuine Greenwald captivity advantage, the durable part of the moat. However, management concedes it “shares” sockets with quartz suppliers who win on price, so in the price-sensitive mainstream the captivity is weak. The moat is concentrated in the highest-performance sockets. (FACT + INTERPRETATION.)

3. Economies of scale — absent. SiTime is far smaller than TI, Microchip, Murata, Renesas, Kyocera, Epson. No cost-scale advantage; the 10-K admits third-party manufacturers “serve customers larger than us… which may decrease our relative importance and negotiating leverage.” The fabless model gives flexibility, not a cost moat. (FACT.)

Share-stability test. SiTime holds ~90%+ of MEMS timing — but MEMS is a small slice of the ~$11B total TAM still dominated by quartz; its $327M FY25 revenue is low-single-digit % of the total and only ~8% of even its own $4B SAM. The relevant test is whether it is taking durable, stable share from quartz across the broader market. The record is improving but not yet proven stable: it gained in the 2021–22 shortage (capacity when quartz couldn’t supply — partly cyclical), gave some back in 2023, and is gaining again on AI. Share is trending up structurally in the high end, but the company is too early in penetration and too cyclical to call share “stable” in the Greenwald sense. (INTERPRETATION.)

The financial test (decisive). Would economics deteriorate without the moat? Partially, and only now emerging. Gross margin clearly passes: 49.9% (2020) → 64.5% (2022) → 53.6% GAAP (2025) → 64.5% non-GAAP (Q1-26). A 50–65% hardware gross margin is strong evidence of differentiation; quartz commodity parts do not earn it. Operating-level returns FAIL through 2024: GAAP net losses in every year except 2021/22, and even then ROIC was only ~8.6% (2021) / ~2.4% (2022). The moat-driven gross margin was, until recently, entirely consumed by heavy R&D + SBC — the company was investing the gross-margin dollars to build the franchise, not harvesting excess returns. The inflection is real but unproven for durability: Q1-26 hit 28% non-GAAP operating margin (GAAP operating income just turned positive); management targets sustainable 65%/30% from Q2-26. If sustained, that finally converts the moat into excess returns — but it is one inflection against a six-year record of sub-WACC consolidated returns, at a cyclical AI peak. (FACT + INTERPRETATION.)

Verdict: a real but narrow and only-recently-monetized moat — durable demand-side captivity + supply-side technology barrier at the high end, contestable in the mainstream, no scale advantage, single-supplier dependency. The decisive open question, handed to Valuation: is the 2025–26 CED surge a permanent baseline (moat finally harvesting) or a cyclical AI-capex peak (the way 2022 was)?

5. Growth History and Forward Opportunities

The revenue record: small base, cyclical detonation, AI inflection. $116M (2020) → $219M (2021) → $284M (2022 peak) → $144M (2023, −49%) → $203M (2024, +41%) → $327M (2025, +61%); Q1-26 $113.6M (+88%); FY26 guide ≥80% (~$590M); Q2-26 $140–150M. (FACT.) The 2022→2025 CAGR is only ~5% — the company round-tripped a full cycle and is exiting at a higher run-rate. The 2023 collapse is the most important fact here: a −49% one-year decline is the signature of a high-beta component supplier whose end-markets destock hard, not a fortress secular compounder. The current melt-up cannot be read without that precedent. (INTERPRETATION.)

Segment composition — one vertical carries everything. CED is two-thirds of revenue growing +158%; the historical mobile/consumer core (the very segment that drove the 2020–21 boom and 2023 bust) is now flat-to-down. SiTime has swapped one concentration (Apple/mobile) for another (AI datacenter) — higher-quality in margin/content terms, but not diversification, and it ties the near-term trajectory directly to the AI-datacenter capex cycle. (INTERPRETATION.)

Organic vs. acquired. Growth through 2025 was overwhelmingly organic (design-win/content-driven), bracketed by two bolt-ons: Aura (Dec-2023, clocks) and the transformational Renesas timing business (announced Feb-2026), which adds a clock-IC franchise and cross-sell, financed by the $1.35B 0% convert. Strategic logic: pair SiTime’s MEMS resonators/oscillators with a clock-IC portfolio to sell a complete timing-tree solution. Execution unproven; it roughly doubles the addressable portfolio but adds first material leverage and dilution. (FACT + INTERPRETATION.)

Forward opportunity — content-per-system. The right lens for a component supplier (durable value comes from content/share gains that survive the cycle): AI inference 2–4x content; 1.6T optical + co-packaged optics; AI-telecom 3x; Titan resonators ~$400M funnel; aero/defense + LEO satellites (~$2,000/satellite, 7,000–10,000 launches) — a genuinely new, high-ASP, less-cyclical vertical; SAM framing $4B-in-$11B implies a long share-gain runway vs. quartz. (Fact for the claims; treat as management hypothesis pending external validation.)

Verdict — mixed; the distinction matters more than the headline rate. There is a real, durable secular thread (MEMS-over-quartz share gain, rising content per box, the aero/defense/LEO vertical) that should persist across cycles and shows up in expanding margins — genuinely high-quality. But the current +88%/+158% rate is not separable from the AI-datacenter capex spike, CED is two-thirds of revenue, and the 2023 −49% precedent proves how fast end-demand reverses. Secular content growth is real and durable; the level and rate of the current print are cycle-flattered — investable on the secular thread, dangerous if priced (as it is) as if the peak rate were the baseline.

6. Financial Quality

Revenue: hyper-growth around a brutal air-pocket. Covered above; the 2023 −49% is the cautionary fact, and the +88% Q1-26 print laps a depressed comparable. The durable run-rate is the open question, not the trailing growth rate.

Gross margin: a genuinely good franchise margin. ~50% (2020) → 53.6% FY25 GAAP → 59.0% Q1-26 GAAP / 64.5% non-GAAP, management targeting 65%. A 53–59% GAAP gross margin on a fabless model, rising with volume and mix, is the cleanest evidence of a real product advantage and the part of the story that survives skepticism. (FACT — 10-K; Q1-26 10-Q.)

The GAAP-vs-non-GAAP chasm — not cosmetic. SiTime has reported a GAAP net loss every year except 2021/2022.

(FACT — 10-K Statements of Operations & Cash Flows, Note 8; Q1-26 10-Q.)

The single line that explains the entire GAAP loss is SBC: $103.5M in FY25, ~32% of revenue. The −$67.0M GAAP operating loss is smaller than SBC — the company would be GAAP-operating-profitable if it paid people in cash, but it pays in stock, so the cost is borne by shareholders through dilution rather than the P&L. Non-GAAP flips the story (Q1-26 non-GAAP operating margin ~28%, EPS $1.44). SBC is a real economic cost, not a “non-cash” courtesy: it transferred ~$100M/year of value to employees, and share count rose ~63% in five years; $111M of unrecognized SBC remains. The “operating leverage” narrative is partly real at the gross-margin line and partly a non-GAAP illusion at the operating line. (FACT + INTERPRETATION.)

“Free cash flow” is an SBC add-back artifact. FY25 operating cash flow $87.2M − capex $52.0M ≈ $35M “FCF” — but built from a −$42.9M net loss plus $103.5M of SBC (and $40.2M D&A). SBC alone is ~3x reported FCF. Treating SBC as the cost it is, real economic FCF net of dilution is roughly −$68M. Q1-26 is starker: OCF $31.2M against SBC $30.8M — the entire quarter’s operating cash flow ≈ the stock handed out. On a true owner-cash basis the company does not yet self-fund. (INTERPRETATION — analytic adjustment.)

Capex: fabless but no longer light. $8.9M (2023) → $36.2M (2024) → $52.0M (2025), ~16% of revenue — investment in test/manufacturing equipment for the AI ramp; further pressures real FCF. (FACT.)

Returns and balance sheet. GAAP ROIC/ROE are negative (net losses on $1.16B equity) — on a GAAP basis SiTime has not yet earned a return on capital; the quality lives in gross margin, not yet in returns. The balance sheet was a fortress until May 2026: cash + short-term investments $808M, no funded debt; total equity $1,156M; goodwill $87M + intangibles $147M (Aura), tangible book ~$922M (~$35/sh). That net-cash story ended with the $1.35B convertible — post-Renesas the company swings to a modestly net-debt position, a material change the trailing balance sheet does not yet show. (FACT — 12/31/25 balance sheet; 8-K 5/22/26.)

Verdict — high-quality top line, low-quality cash conversion once SBC is costed. A genuinely good gross-margin franchise (53–59% and rising) attached to a cyclical revenue base and a GAAP-loss bottom line. Economics improve with scale at the gross-margin line, but reported operating leverage and “FCF” are substantially a non-GAAP/SBC add-back illusion: SBC at ~32% of revenue is the entire GAAP loss, real FCF net of SBC is negative, and capex is rising.

7. Capital Allocation

Reinvestment — R&D-heavy, appropriately. R&D ran $97.6M / $106.9M / $118.9M (2023–25) — ~36% of FY25 revenue. For a differentiated component company in a design-win business, plowing roughly a third of revenue into R&D is the right instinct and the engine of the gross-margin advantage. The most defensible piece of the record. (FACT.)

Equity issuance — well-timed raises, no offsetting buyback. SiTime funds itself by selling stock, opportunistically into strength: a June 2025 follow-on of 2,012,500 shares at $200 = ~$387M net (the stock later traded above $700 — good timing), plus an ATM (263,400 sh at avg $251 = $64M in FY25) to fund RSU tax withholding. Total FY25 common-stock proceeds $468.7M. Against this there is no share-repurchase program while share count climbed 16.1M (2020) → 26.3M (2025). The company issues; it does not retire — defensible for a growth ramp, but per-share value continually leaks through dilution. (FACT — 10-K MD&A; 8-Ks.)

The Renesas bet — the defining event. On 2026-02-04 SiTime agreed to buy Renesas’s timing business for ~$1.5B cash + 4,130,644 shares (stock leg collared $308.67 floor / $417.61 ceiling); HSR cleared 5/8/26. Financed by a $1.35B 0% convertible due 2031 (closed 5/22/26): zero coupon, conversion ~$1,040.47 (~50% premium to the $693.65 reference), not callable before June-2029. The trade-off is stark: SiTime gets effectively free five-year money but (a) converts a $808M net-cash balance sheet to net debt after the $1.5B outflow, (b) adds ~4.13M shares, and © creates ~1.3M shares of conversion overhang above ~$1,040. The “0%” is not free — the cost is embedded equity optionality handed to noteholders at a 50% premium. Whether brilliant (cheap capital to consolidate a fragmented industry) or top-of-cycle empire-building depends on whether Renesas and the AI ramp justify a ~$1.5B+ outlay at a record share price; the target’s standalone revenue/margin is undisclosed pre-close. (FACT + INTERPRETATION.)

Compensation — Marathon mis-incentive. The proxy tilts the wrong way: annual bonus 70% weighted to GAAP revenue-growth goals + 30% individual (FY25 paid at the 200% maximum, settled in RSUs); LTI is RSUs + PRSUs with relative TSR vs. the Philadelphia Semiconductor Index as the only performance metric. No ROIC, no per-share, no margin, no FCF metric anywhere. Cash/operating pay rewards revenue growth — the classic Marathon “size” mis-incentive that encourages chasing top line and large acquisitions over capital efficiency; the lone aligned element is the rel-TSR PRSU. CEO Rajesh Vashist FY25 total comp $14.26M (heavily equity-weighted); CFO Beth Howe $3.08M; insiders ~1.4% of shares. The 2026 say-on-pay passed but with a notable ~18% against. (FACT — DEF 14A.)

Verdict — a capable growth allocator, not yet a per-share-value allocator. Credits: R&D intensity, the well-timed $200 raise. Debits: the Renesas/$1.35B-convert combination is a bold, cycle-top, leverage-and-dilution-adding bet of unproven merit; no buyback against chronic dilution; comp keyed to revenue growth with no ROIC/per-share metric. The balance sheet’s financial character has just changed materially from net cash to net debt.

8. Changes and Headwinds — Last Two Years

The last 24 months transformed SiTime from a recovering small-cap MEMS-oscillator vendor into a leveraged, AI-levered, M&A-driven timing platform. Six changes stand out:

1. Aura acquisition / move into clocks (Dec-2023, ~$148M + earnouts). Closed at the cyclical trough; gave SiTime a clock-IC capability and the strategic seed of the later Renesas logic. (FACT.)
2. The AI-datacenter inflection (2024–2026). The decisive operational change: CED from a minor line to two-thirds of revenue on eight consecutive triple-digit quarters. What re-rated the stock and what the valuation now underwrites — simultaneously the thesis and the largest concentration risk. (FACT + INTERPRETATION.)
3. The June-2025 equity raise (2.0125M sh at $200 = ~$387M). Opportunistic, accretive to the balance sheet, raised ahead of the deal — but the first signal management would fund expansion with paper. (FACT.)
4. The Renesas deal + $1.35B convertible (Feb–May 2026). Roughly doubles the timing portfolio, adds clock cross-sell, introduces first material leverage (offset by zero coupon / high strike), and layers convert dilution on top of deal-share and raise dilution. The thesis now depends on integrating a carve-out — the riskiest kind of deal. (FACT.)
5. Governance / ownership: MegaChips sell-down + insider tape. Former parent and largest holder MegaChips sold ~$312M into the spike; total insider selling YTD ~$379M with zero open-market purchases. Say-on-pay passed (~18% against). The tape is unambiguously distributive. (FACT.)
6. The gross-margin / operating-leverage inflection. Management targets sustainable 65% GM / 30% operating margin from Q2-2026 — the genuine quality improvement of the period. (FACT.)

Headwinds. Cyclicality (the 2023 −49% is two years old, not ancient; AI-capex is the most procyclical demand pool in tech — and CEO Vashist himself said in Q3-2022, “In 40 years of industry experience, I have not seen a downturn as rapid or as deep as this one”); customer/vertical concentration (Apple ~17%, CED two-thirds); integration + leverage; valuation; and a new ~$1,040 conversion overhang plus continued sponsor distribution.

Verdict — operationally stronger, risk-profile materially higher. The business is unambiguously better than two years ago — bigger, higher-margin, more strategically complete, with a real AI/aero content engine. But the equity thesis is more fragile: the changes concentrated the revenue base into the single most cyclical end-market, added first-time leverage and serial dilution, and arrived alongside heavy insider/sponsor selling and zero insider buying. The business strengthened; the margin of safety weakened.

9. Risk Analysis

Narrative. The dominant risk is the price itself (#1): at the richest revenue multiple in its peer set and the 99th percentile of its own history, even strong execution can produce flat-to-negative returns if the multiple normalizes toward profitable peers. Cyclicality (#2) has lived precedent — the −49% 2023 collapse is direct evidence the demand is cyclical, and the book is now more concentrated in the AI-capex-sensitive socket; bulls and bears disagree only on whether 2025–26 is a floor or a peak. SBC (#3) is the quality issue the non-GAAP framing obscures. Bosch (#5) is a genuine single-point-of-failure tail. Taiwan/China (#11) is the low-probability, catastrophic-impact tail common to fabless semis but acute given TSMC dependence. Catastrophic/total-loss risk is low (solvent, asset-light, liquid, 0% convert maturing 2031) — but a 50–65% drawdown requires no business failure at all, only a normalization of the AI cycle and the multiple together.

10. Valuation Discussion (Embedded Expectations)

Methodological note: SiTime cannot be valued on earnings. GAAP net loss in every year but 2021–22, FY25 net loss −$42.9M on a −20.5% operating margin, the entire loss is SBC ($103.5M, ~32% of revenue). Reported “$35M FCF” is an SBC add-back artifact; TTM EBITDA is a rounding-error ~$5M (a nonsensical ~1,600x). There is no meaningful trailing P/E, clean EV/EBITDA, or FCF yield. EV/sales is the only workable lens, supplemented by forward non-GAAP P/E on the company’s own guidance and a reverse-DCF. That fact pattern is itself a finding: investors pay one of the highest revenue multiples in semiconductors for a business whose bottom line and cash conversion do not yet exist after costing labor.

Building a clean pro-forma EV. Stand-alone (pre-close): 26.3M shares × $729 = ~$19.2B equity, less ~$808M pre-deal net cash = ~$18.4B EV; recomputed at the actual $729 close, that is ~48x Q1-26 TTM revenue (~$380M) and ~56x FY25 revenue ($326.7M) (the ROIC feed’s ~21–24x uses a stale ~$315 price). Pro-forma (post-close): pre-deal cash $808M + $1.35B convert − $1.5B Renesas cash ≈ −$658M residual (i.e., net-cash flips to net-debt ~$0.5B); adding 4.13M Renesas shares (~$3.0B) lifts equity to ~$22.2B on ~30.4M shares, and pro-forma EV ≈ $22–23B. The debt-free net-cash story that underwrote part of the multiple is over. (FACT for inputs; pro-forma is INTERPRETATION.)

Renesas revenue — ASSUMPTION / OPEN QUESTION. Standalone economics undisclosed pre-close. The Renesas/IDT timing-and-clock franchise was historically a ~$300–400M line in total; the carve-out is a subset. We assume ~$150–250M (midpoint ~$200M), structurally lower-margin and slower-growth than SiTime’s core. Combined full-year pro-forma revenue ≈ ~$590M SiTime FY26E + ~$200M Renesas ≈ ~$790M, putting pro-forma EV/forward-sales at ~28–29x. Revisit at close.

Comparables.

Forward non-GAAP P/E: Q2-26 guide non-GAAP EPS $1.85–2.00 annualized ≈ $7.40–8.00 → ~91–99x at $729.

The read. SiTime is the most expensive name in the entire comp set on revenue — ~2–3x the richest profitable analog (MPWR) and above ARM. Even on forward standalone sales (~$590M → ~31x) or pro-forma combined (~$790M → ~28x), it is ~2x ARM and well above the analog group, despite GAAP losses where ADI/TXN/MPWR earn 25–49% operating margins. The bull retort — earlier in the S-curve, higher growth, deserves a duration premium — is the same argument behind MPWR’s and ARM’s premiums; the discomfort is paying it on top of a +220% twelve-month re-rating, on a business that demonstrably collapsed −49% in 2023.

Reverse-DCF / what the price requires. To justify a ~$20B EV at a defensible terminal ~3% free-cash yield, SiTime must eventually generate ~$600M of sustainable FCF. At a charitable mature 25% FCF-to-sales margin (it has never produced positive FCF net of SBC), that implies ~$2.4B of revenue — ~7x FY25 and ~4x FY26E. Stacked, the embedded expectation is: sustained ~30–40% growth for 5+ years to a ~$1.5–2.5B base; conversion to genuine 25–30% operating margins net of SBC; SBC declining as a share of revenue; clean Renesas accretion; and the ~30–50x multiple persisting. Every one has to land. There is no margin of safety.

Scenarios (3-year, illustrative).

Bear is the lived 2022→2023 path (revenue −49% as the prior end-market digested), reapplied to a book now ~67% CED/AI. Base assumes the secular thesis is real but decelerates, growth offsets multiple de-rating, Renesas adds a lower-multiple layer — roughly a wash. Bull requires the category-monopoly narrative fully true and the market continuing to pay ~30–40x revenue. Valuation verdict: no margin of safety — EV/sales at ~48–56x trailing / ~28–31x forward, 2–3x the richest profitable analog; the asymmetry favors the seller at $729. (No price target.)

11. Variant Perception

Consensus belief. SiTime is the category-defining near-monopoly in MEMS precision timing — an emerging “timing standard” — riding a structural multi-year AI-datacenter content cycle (content rising 2–4x per system), with an operating-leverage inflection finally arriving (non-GAAP GM ~64%, op margin ~28%, “65%/30% sustainable”) and Renesas consolidating the clock field. The +220% twelve-month move and +158% CED growth are taken as confirmation of a durable secular premium.

Strongest bull case. Timing is a content-growth story, not a unit story: every AI server, optical module, switch, and satellite needs more and higher-performance timing, and SiTime owns the only scaled all-silicon MEMS platform with qualification lock-in. The 2022→2023 collapse was a consumer/mobile inventory correction; the current book is AI-infrastructure — a structurally more durable demand pool. As revenue scales, fixed R&D and SBC leverage down, GAAP follows non-GAAP into clear profitability, and Renesas adds clock breadth for a full timing-tree solution. A small-base category monopoly in a secularly-growing niche can compound 30–40% for years and grow into even this multiple.

Strongest bear case. You are paying ~48–56x trailing EV/sales — the richest in the comp set — for a business that (1) has never produced positive FCF net of SBC and runs a GAAP loss because SBC is ~32% of revenue; (2) fell −49% the last time end-demand digested and is now more concentrated in the most AI-capex-sensitive socket; (3) just flipped net-cash → net-debt via a $1.35B convert for a ~$1.5B deal whose target economics it won’t disclose; (4) shows a net-distribution insider tape (~$379M sold YTD, zero buys, MegaChips $311.8M); and (5) trades as a beta-2.67 momentum vehicle (specific vol 71%, no value/quality/growth factor support) whose +319%-annualized six-month melt-up is a flow phenomenon. If AI capex air-pockets, revenue and multiple compress together — a 50–65% drawdown with the business otherwise intact.

Factor read (the tape as evidence). FactorsToday places SITM as a crowded, high-beta (2.67), highly idiosyncratic (R² ~0.35–0.39, specific vol 71%) AI-thematic momentum name — anti-LowVol (−1.29), Semis +0.66, Tech +0.42, with no Value, Quality, Growth, or even clean Momentum style loading (L1-sparse-zeroed), and a factor-similar peer basket of QTUM/IGPT/QCLN/SIMO/SKYT (an AI/quantum/tech ETF complex). Track record annualized y1 +220% / m6 +319%, against a y5 max drawdown −78%. This is not a falling knife, not value, not quality — a narrative/flow vehicle whose 2.67 beta amplifies the AI trade both ways. That is precisely where consensus is most offsides: the price reflects positioning and momentum as much as fundamentals, so a regime shift hits before any fundamental data confirms it. The thin float cuts both ways — amplifying squeeze risk on the upside and air-pocket risk on the downside.

The 3–5 assumptions that matter most: (1) CED durability — permanent secular baseline or cyclical AI-capex peak? (the crux); (2) SBC normalization — does it fall enough that per-share economics materialize?; (3) Renesas accretion — does an undisclosed-economics, lower-margin clock business add value or dilute the franchise?; (4) Multiple persistence — does the market keep paying ~30–50x revenue, or de-rate toward profitable peers (~15–24x)?; (5) Competitive response — do larger rivals and consolidating clock-IC players erode the high end (Marathon)?

Falsification. Bull is falsified if CED growth rolls over (sequential declines, distributor inventory build, a 2023-style guide cut), if SBC stays ~30% of revenue as growth slows, or if Renesas integration stumbles. Bear is falsified if CED sustains 2–3 more quarters with widening backlog, GAAP turns durably positive net of SBC, Renesas proves cleanly accretive, and design-win disclosures confirm content gains rather than just AI-capex tailwind.

12. Fact vs. Interpretation

13. Open Questions

1. Renesas economics — standalone revenue, gross/operating margin, growth rate, and the implied acquisition multiple (undisclosed pre-close). The single biggest valuation unknown.
2. CED durability — is the AI-datacenter content surge a permanent baseline, or will it digest like 2022→2023? Watch sequential CED, distributor inventory weeks, and book-to-bill.
3. SBC trajectory — does SBC fall meaningfully as a share of revenue as growth scales, or does dilution permanently eat the operating leverage?
4. MegaChips intent — is the 400k-share sale the start of a full exit of a ~13% strategic holder? Watch 13G/A and Form 144 flow.
5. GAAP-margin reconciliation — the filed FY25 GAAP operating margin is −20.5% (−$67.0M / $326.7M); aggregator feeds differ by the classification of acquisition costs. Use the filing.
6. Post-close pro-forma capital structure — exact net-debt, share count, and conversion overhang once the deal closes and the 10-Q reflects it.

14. What Must Be True

Bull case — what must be true: (a) the AI-datacenter timing content cycle is secular, not a capex peak — CED sustains growth and widening backlog for several more quarters; (b) operating leverage is real net of SBC, with SBC falling as a share of revenue so GAAP/per-share economics finally materialize; © Renesas integrates cleanly and proves accretive, justifying the ~$1.5B+ outlay and the new leverage; (d) the market continues to pay a premium revenue multiple as the company scales toward a multi-billion revenue base. Falsification test: a single 2023-style guide cut or two quarters of CED sequential decline / distributor inventory build, or SBC stuck near 30% of revenue as growth decelerates, breaks the bull thesis.

Bear case — what must be true: (a) the 2025–26 surge is a cyclical AI-capex peak that digests as 2022 did, with revenue and the multiple compressing together; (b) ~48–56x trailing EV/sales is unsustainable and de-rates toward profitable peers; © SBC dilution and the new convert leave per-share value creation chronically leaking; (d) competitive/capital-cycle pressure erodes the high-end sockets. Falsification test: CED sustains 2–3 more quarters of growth with a widening order book, GAAP turns durably positive net of SBC, Renesas proves cleanly accretive, and design-win disclosures confirm structural content gains — which would validate the secular read and defend the premium.

15. Source Appendix

See the separate Source Appendix (SITM_source_appendix.md) for the full list of primary and secondary sources, and the Diligence Questionnaire (SITM_diligence_appendix.md) for the standard diligence answers. Principal sources: SiTime FY2025 Form 10-K (filed 2026-02-11); Q1-2026 Form 10-Q and earnings call (2026-05-07); 8-Ks dated 2026-02-04 (Renesas APA), 2026-05-08 (HSR), 2026-05-22 ($1.35B convertible), and 2025-06-26 ($200 follow-on); DEF 14A (2026 proxy); the SEC Form 4 corpus (2026 YTD); ROIC.ai fundamentals; AZI price series and valuation-index percentiles; FactorsToday factor model; and SiTime historical earnings transcripts (2022–2023).

This memo discusses valuation only as embedded expectations and scenarios. It contains no buy/sell recommendation and no price target. The single subjective view appears in the clearly-labeled “Claude’s Take” block at the top.

APPENDIX A — Standard Diligence Questionnaire

Supplemental to the research memo. Report date 2026-06-20. Fact/Interpretation/Assumption labels applied where material.

General

What thoughtful questions have other investors asked about this company? The dominant debates: (1) Is the AI-datacenter timing content surge secular or a cyclical capex peak (the 2022→2023 −49% precedent looms over every model)? (2) When — and whether — does GAAP profitability arrive net of ~32%-of-revenue stock-based comp? (3) Is the Renesas acquisition value-creating or top-of-cycle empire-building, and what does the carve-out actually earn? (4) Can a ~50x EV/sales multiple persist? (5) Does Bosch single-sourcing the MEMS die deserve a risk discount? (6) What does the wall of insider/MegaChips selling signal?

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? Interpretation: near a cyclical high in growth rate (+88% YoY laps a depressed comp), though revenue dollars are at an all-time high and arguably still ramping. The end-market (AI datacenter, ~67% of revenue) is the most procyclical demand pool in tech.

Driven by external environment or internal actions? Both: internal design-win execution and content gains (durable) atop an external AI-capex boom (cyclical). The two are hard to disentangle, which is the crux of the thesis.

How stable are revenues? Fact: Highly unstable historically — $284M (2022) → $144M (2023, −49%) → $327M (2025). This is a cyclical component supplier, not a smooth annuity.

Outlook for products/services? Fact (guidance): FY26 ≥80% growth, Q2-26 $140–150M, target 65% GM / 30% operating margin. Content-per-system tailwinds (AI inference 2–4x, 1.6T optical, LEO satellites) support a multi-year secular thread.

How big will the market be? Fact (management): ~$4B precision-timing SAM within an ~$11B total timing TAM; growing, global, with a long share-gain runway vs. quartz.

Business Quality & Competitive Moat

Industry more or less competitive? More — quartz incumbents pushing up-market and clock-IC players (TI, Microchip, Renesas-pre-deal) consolidating; SiTime’s own 10-K warns of intensifying competition.

How profitable is the business (ROIC, ROE)? Fact: GAAP ROIC/ROE negative (net losses); even in 2021/2022 ROIC was only ~8.6% / ~2.4%. Gross margin (53–65%) is excellent; returns on capital are not yet positive on a GAAP basis. The Q1-26 operating inflection is one quarter, unproven for durability.

How profitable is the industry — competitors, barriers? Bifurcated: commodity quartz earns low returns; the precision niche SiTime created earns 50–65% gross margins. Barriers to being SiTime are high (proprietary MEMS, 147 patents); barriers to competing in timing generally are low.

Easily understood? Yes — a differentiated component (silicon MEMS timing) displacing a legacy one (quartz), sold into electronics.

Undermined by foreign low-cost labor? Partially — quartz incumbents are largely Asian low-cost manufacturers; SiTime competes on performance/reliability, not price, and shares price-sensitive sockets with them.

Do brands matter? Modestly — “SiTime / precision timing” carries weight in high-reliability design-ins, but the moat is qualification lock-in and MEMS technology, not brand.

Nature of competition? Performance, reliability, supply assurance, and qualification lock-in at the high end; price in the mainstream.

Customers’ switching costs? Real at the socket level — 6-month-to-3-year qualification, 10-year+ product life, requalification “several quarters or longer,” amplified by uptime risk in mission-critical applications.

Financial Condition & Balance Sheet

Assets not fully on the balance sheet? The proprietary MEMS process IP and design-win backlog are not capitalized; the moat is largely off-balance-sheet intangible.

Off-balance-sheet liabilities? Operating leases; the ~1.3M-share conversion overhang above ~$1,040 on the new convert is a contingent dilution claim.

How conservative is the accounting? Reasonable — Fact: no restatements; one-time acquisition costs disclosed; the key issue is not aggressiveness but the GAAP-vs-non-GAAP gap (SBC) and management’s emphasis on the non-GAAP framing. Management has repeatedly stated “the only difference between GAAP and non-GAAP is stock-based compensation.”

How CapEx-hungry? Rising — $9M (2023) → $52M (2025), ~16% of revenue; “fabless” but no longer capex-light as it invests for the AI ramp.

Capital Allocation & Management

FCF generation and use; philosophy? Interpretation: real FCF net of SBC is negative; the company is reinvesting (R&D ~36% of revenue) and funding via equity/convert, not returning cash. No dividend, no buyback.

Significant acquisitions recently? Yes — Aura (Dec-2023, ~$148M + earnouts, clocks) and the transformational Renesas timing business (Feb-2026, ~$1.5B cash + 4.13M shares).

Buying back shares? No — share count rose ~63% in five years; no repurchase program.

Issuing shares to insiders? Yes — heavily: SBC ~$103.5M FY25 (~32% of revenue), $111M unrecognized; plus the $200 follow-on and ATM.

Compensation policy? Fact: annual bonus 70% revenue-growth-weighted (paid at 200% max in FY25), LTI on relative TSR vs. SOX; no ROIC/per-share/margin/FCF metric — a Marathon “size” mis-incentive. CEO Vashist FY25 comp $14.26M.

Motivations of management? Founder-CEO (Vashist) driving category creation and aggressive growth/M&A; equity-heavy comp aligns form with the share price but the metric mix rewards revenue scale over capital efficiency.

Valuation & Market Data

ADR, MLP, or K-1? No — a US C-corp common stock (1099), single share class.

Dividend policy? None.

How profitable? GAAP-unprofitable; non-GAAP-profitable only because SBC (~32% of revenue) is added back.

Net income diverging from cash from operations? Yes — net loss −$42.9M vs. OCF +$87.2M FY25, the gap driven by ~$103.5M SBC and D&A; OCF flatters the true owner-cash picture.

Risks & Downside

Factors that would cause the stock to decline? Multiple compression from a 99th-percentile valuation; an AI-capex digestion / 2023-style guide cut; SBC dilution; a Renesas integration stumble; a Bosch supply disruption; continued insider/MegaChips distribution; a regime shift in the AI-thematic trade (beta 2.67).

Risk of catastrophic loss? Low operationally — solvent, asset-light, liquid, 0% convert maturing 2031, real franchise. A 50–65% drawdown requires no business failure, only normalization of the cycle and multiple together.

Chance of total loss? Very low — net-cash-to-modest-net-debt, real revenue and IP, no near-term solvency risk.

Recent News & Events

Has the business environment changed recently? Yes, materially — the AI-datacenter inflection (CED +158%), the Renesas deal + $1.35B convert (net-cash → net-debt), the June-2025 equity raise, and heavy insider/MegaChips selling into a +220% twelve-month run.

Significant acquisitions? Renesas timing business (pending close) and Aura (2023).

Change in accounting policies? None identified.

Recent changes — new markets, facilities, management? New AI-datacenter and aerospace/LEO verticals; clock-IC expansion via Aura/Renesas; CFO transition Art Chadwick → Beth Howe (late 2023).

APPENDIX B — Source Appendix

Report date 2026-06-20. Primary sources prioritized. All financial figures reconciled to SEC filings.

Primary — SEC filings (SiTime Corporation, CIK 0001451809)

• Form 10-K, FY2025 (filed 2026-02-11) — business description, MEMS/quartz technology, segments, customer/distributor concentration (Arrow 26%, Pernas 25%, Quantek 13%; Apple ~17%), Bosch/TSMC/UMC manufacturing, competition, risk factors, Aura acquisition, statements of operations/cash flow/balance sheet, SBC (Note 8).
• Form 10-K, FY2021–FY2024 (filed 2022-02-25, 2023-02-27, 2024-02-26, 2025-02-14) — multi-year revenue/margin history, MegaChips ownership history, Bosch royalty expiry (3/2024).
• Form 10-Q, Q1-2026 (filed 2026-05-07) — Q1-26 revenue $113.6M, segment detail, SBC $30.8M, balance sheet.
• Q1-2026 earnings call transcript (2026-05-07) — CED +158%, FY26 ≥80% guide, Q2-26 $140–150M / non-GAAP EPS $1.85–2.00, 65%/30% margin target, AI content drivers, Renesas commentary; CEO Rajesh Vashist, CFO Beth Howe.
• Form 8-K (2026-02-04) — Renesas timing-business Asset Purchase Agreement (~$1.5B cash + 4,130,644 shares; collar $308.67/$417.61).
• Form 8-K (2026-05-08 / 2026-05-11) — HSR waiting-period expiration.
• Form 8-K (2026-05-22) — $1.35B 0% Convertible Senior Notes due 2031 (conversion ~$1,040.47, ~50% premium).
• Form 8-K (2025-06-26) — follow-on offering 2,012,500 shares at $200.00 (~$387M net).
• DEF 14A (2026 proxy) — compensation structure (70% revenue-growth bonus, rel-TSR PRSU vs. SOX, no ROIC/per-share metric); CEO comp $14.26M, CFO $3.08M; insider ownership ~1.4%; say-on-pay vote.
• Form 4 / Form 144 corpus (2026 YTD) — ~$379M insider sales, zero open-market buys; MegaChips 400,000 shares ~$311.8M; CEO Vashist ~$43.8M (10b5-1).

Primary — historical transcripts (SiTime quarterly earnings calls)

• SiTime quarterly earnings call transcripts, 2022–2023 (publicly available) — cyclicality precedent; Q3-2022: CEO Vashist, “In 40 years of industry experience, I have not seen a downturn as rapid or as deep as this one”; trough quarters (Q2-23 $27.7M); confirmation that GAAP/non-GAAP differ only by SBC; CFO transition Chadwick → Howe.

Quantitative data services

• ROIC.ai — income statement, balance sheet, cash flow, profitability/credit ratios, enterprise value, valuation multiples (multi-year). Third-party aggregated; reconciled to filings.
• AZI price series & valuation-index percentiles — five-year OHLCV (split/dividend-adjusted), 21/50/200-EMA, beta; own-history P/S 99.3rd / P/B 86.8th / composite 93rd percentile.
• FactorsToday factor model — ElasticNet factor loadings (Market beta 2.06–2.67, anti-LowVol −1.29, Semis +0.66; no value/quality/growth style loading), leaderboard (y1 +220%, y5 max drawdown −78%), specific vol 71%, related stocks (QTUM/IGPT/QCLN/SIMO/SKYT).

Comparable-company context

• Public financial data and filings for Monolithic Power (MPWR), Analog Devices (ADI), Microchip (MCHP), Texas Instruments (TXN), Silicon Labs (SLAB), and ARM Holdings (ARM) — used for EV/sales, EV/EBITDA, gross/operating margin, and growth benchmarking.

Notes on source treatment

• Management commentary (calls, releases, investor framing) is treated as hypothesis, validated against filings and external data.
• ROIC/AZI/FactorsToday are third-party aggregated data; SEC filings are primary. Where they differ on a material figure (e.g., the GAAP operating margin, affected by acquisition-cost classification), the filing governs.
• Renesas-business standalone revenue/margin is undisclosed pre-close; the ~$150–250M estimate is an explicit assumption flagged in the memo.