Quartz To MEMS — SiTime's Moment | Long-term Investing : Thematic

A 100-year-old quartz industry is being disrupted by silicon MEMS — and one listed pure-play stands to capture the shift‍

Every digital device on Earth has a metronome inside it. For a century, that metronome has been a sliver of quartz. That assumption is now being challenged in real time.

‍A century-old, materially-constrained, hand-tuned analogue industry (quartz) is losing its premium segments to a scalable, programmable, silicon-fabricated alternative (MEMS) — at the precise moment when the most demanding new end-markets (AI, satellites, defence, autonomy) need exactly what quartz cannot deliver.

The 30-Second Primer — Clocks & Oscillators‍

Every digital chip in the world — every CPU, GPU, smartphone, satellite radio, ADAS sensor, IoT node — is a vast team of tiny circuits that must operate in synchrony. They need a common heartbeat. That heartbeat is supplied by an oscillator, a small component that produces a precise, stable electrical signal at a fixed frequency. The cleaner and steadier the signal, the more accurately the chip can move and process data; the noisier or wobblier it is, the more errors creep in.

‍For one hundred years, the way humans have generated this heartbeat has been with quartz — a small, machined sliver of natural or synthetic quartz crystal that vibrates at a precise frequency when stimulated with electricity (the piezoelectric effect, discovered by the Curie brothers in 1880). Quartz is everywhere: a modern smartphone contains 8–12 quartz oscillators; an AI server, dozens; a satellite, hundreds. The global market for these components is roughly USD 6.4 billion (oscillators) + USD 1.5 billion (crystals) ≈ $8 billion, growing at low-to-mid single digits, and historically dominated by a handful of Japanese and Taiwanese specialists.

‍This century-old quartz monopoly is now being broken. MEMS oscillators — Micro-Electro-Mechanical Systems built on silicon wafers using the same fabrication processes as modern semiconductor chips — are taking share in the most demanding applications, and the high-value end of the timing market is reshaping faster than the headline numbers suggest.

The Top-Down Thesis‍

• Quartz is hitting hard physical limits.

  A quartz oscillator's accuracy degrades with temperature, vibration, shock, age, and acceleration — properties of the rock itself, not of the design. Each crystal is hand-cut and hand-tuned, with a 6-month build cycle from raw quartz to finished part. ~80% of high-grade quartz blanks come from Japan, and the top four producers operated at 92% utilisation with 26-week lead times through 2025. Capacity expansion is glacial: Epson's JPY 15bn Suwa expansion adds just 15% by late 2027.

• The new end-markets are exactly where quartz is weakest.

  AI training clusters need ultra-low-jitter clocks at 100°C+ rack inlet temperatures — quartz frequency drifts with heat. LEO satellites (Starlink, OneWeb, Kuiper) need oscillators that survive launch g-forces and decade-long radiation exposure — quartz is fragile. Defence and aerospace systems demand sub-ppb stability over -55°C to +125°C — quartz needs power-hungry ovens to deliver this. Autonomous vehicles require AEC-Q100 Grade 0 reliability over 15-year design lives in shock-and-vibration-rich environments — quartz cracks. These markets are growing 15–25% per year while traditional consumer quartz volumes are flat-to-declining.

• MEMS solves what quartz physically cannot.

  A MEMS oscillator etches a microscopic vibrating structure directly onto a silicon wafer, paired with a temperature-sensing compensation IC in the same package. Result: ~50x better shock resistance, ~10x better phase noise at high temperatures, ~30x reliability improvement, and — critically — frequency is programmable rather than physically fixed at fabrication. One MEMS part can replace 5–10 distinct quartz SKUs through software configuration.

• The economics are now silicon economics.

  MEMS oscillators are fabricated in standard semiconductor foundries (TSMC, Bosch). Quartz needs purpose-built crystal-cutting and tuning lines with USD 80–120m capex per cleanroom and ~12-week lead times even at full ramp. When AI / satellite / EV demand surges, MEMS can scale; quartz cannot. This was vividly demonstrated in 2024–2025: SiTime grew revenue 41% (FY24) then 50%+ (FY25E) while NDK, TXC, Epson grew low-to-mid single digits.

• Design-in moats run both ways.

  The same multi-year qualification cycle that historically protected quartz incumbents now protects the MEMS challenger once it wins a socket. SiTime's Communications-Enterprise-Datacenter segment grew 137% YoY in Q2 2025 as AI design-wins from 2022–2023 ramp into production. Quartz incumbents are not losing customers — they are losing the marginal new socket, which is where the entire industry's growth is concentrated.

• The category is consolidating.

  Microchip closed its USD 850m acquisition of Renesas' timing division (Jan 2026) to bolt on aerospace and defence OCXO capability. Kyocera opened a JPY 25bn crystal plant (Dec 2025) focused on automotive. Rakon won a NZD 45m OCXO contract for a 120-satellite LEO constellation (Nov 2025). The industry is rearming around the new end-markets — but the only listed pure-play on the MEMS disruption remains SiTime.

Why The Physical Limits Of Quartz Are Now Binding‍

None of quartz's physical limitations are new — the technology has been mature for decades. What is new is the demand stack that has emerged in the 2020s — AI infrastructure, LEO satellite constellations, autonomous vehicles, defence modernisation, edge IoT — that simultaneously demands all of higher temperature stability, better shock resistance, smaller form factors, programmability, and supply-chain diversification. Quartz can deliver any one of these. It cannot deliver all five.

Enter MEMS — Etching The Tuning Fork Into Silicon

The concept is elegant. Instead of cutting a slab of quartz and mounting it in a metal can, a MEMS oscillator uses standard semiconductor photolithography to etch a microscopic vibrating structure — typically a free-hanging silicon resonator a few hundred microns across — directly into a silicon wafer. A separate silicon die holds the analog driver and temperature-compensation circuitry; the two dies are stacked in a standard QFN/DFN package. The technical innovation that made this commercially viable was a vacuum-sealed wafer-bonded cavity protecting the silicon resonator from contamination and pressure variation — pioneered by SiTime (founded 2005, IPO 2019) and protected by a substantial patent portfolio. Earlier MEMS attempts (Sand 9, Discera) failed to achieve required long-term stability; SiTime cracked it and has since extended into TCXO, OCXO, automotive (AEC-Q100), and ultra-low-jitter network timing variants.

The result is a product that fits into existing customer designs (same footprint, voltage, signalling) while delivering measurably superior performance on every metric that matters for the new end-markets: 50x better g-sensitivity, 10x better phase noise at high temperatures, 30x improvement in field reliability, fully programmable frequencies, and the ability to scale supply through standard semiconductor foundry capacity.

Addressing The Price Question — Is The Premium A Roadblock?

‍The most common pushback on the MEMS thesis is price. The premium is real but varies materially by product tier, and at the system level it usually inverts entirely in the segments that matter for this thesis.

• The commodity end is irrelevant.

  The 3–5x premium only exists at the lowest tier (SPXOs in microwaves, TV remotes, basic IoT). SiTime is not playing there; quartz keeps that segment forever. The premium gap that matters — at TCXO and above, where SiTime competes — is only 1.5–2x at the unit level.

• SKU consolidation collapses the unit-cost gap.

  One programmable MEMS part replaces 5–10 distinct quartz cuts. When inventory carrying cost, qualification cost, and procurement complexity are factored in, total cost of ownership often favours MEMS at TCXO ASP parity. Cisco, Nvidia, and Juniper have publicly cited SKU consolidation as a primary adoption driver.

• At the system level the oscillator is rounding error.

  In a $300k AI server rack, the timing component is a fraction of a percent of BOM. A quartz drift event that causes a packet retransmission wastes thousands of dollars of GPU time — that economic asymmetry overwhelms any unit-cost premium.

• Where price IS still a barrier:

  the high-end space/defence OCXO segment ($500–$5,000 ASP), where quartz retains an absolute performance advantage and MEMS hasn't yet matched the spec. The constraint here is performance, not price — and the runway to close it is 3–5 years.

Where The Demand Comes From — Five End-Markets, Each An Independent Tailwind‍

• (1) AI servers & hyperscale data centres.

  Nvidia GB200/GB300/Rubin platforms, AWS/Google/Meta ASIC clusters, and 800G/1.6T optical networking all demand ultra-low-jitter clocking under high-temperature operation. SiTime's CED (Communications, Enterprise, Datacenter) segment grew 137% YoY in Q2 2025 alone. PCIe Gen 5/6/7 retimers, NVLink switches, and PAM4 SerDes all need sub-100 femtosecond jitter that quartz physically struggles to deliver at 100°C+ junction temperatures.

• (2) Satellites & LEO constellations.

  Starlink, OneWeb, Kuiper, Iridium NEXT, Galileo, and a dozen national defence satellite programmes are launching at unprecedented cadence. Each satellite needs multiple ultra-stable oscillators for navigation, payload synchronisation, and inter-satellite optical links. Rakon's RK409 series is in ~200 spacecraft including Galileo, Jason, and HTV; SiTime's Endura family meets MIL-PRF-55310 with 50x lower acceleration sensitivity than quartz, opening up the new-space (small-sat, mass-deployment) segment that legacy space-qualified quartz cannot price-compete for.

• (3) Defence & aerospace.

  Radar, electronic warfare, secure comms, missile guidance, GPS/GNSS denial-resilient platforms, and the Golden Dome / IRIS² / GIDS-style programmes all require precision timing at MIL-STD reliability levels. Microchip's $850m acquisition of Renesas' timing division (Jan 2026) was explicitly motivated by aerospace & defence OCXO positioning. High-margin, long-cycle, programme-budget-funded demand that doesn't move with the consumer cycle.

• (4) Automotive ADAS & EV.

  Zonal vehicle architectures, ADAS sensor fusion, and EV battery management systems all need AEC-Q100-grade timing that survives -40°C to +125°C across a 15-year design life with shock and vibration. MEMS oscillators are now winning Grade 0 automotive designs that previously could only use quartz, with only six vendors clearing the spec as of 2025. SiTime's Cascade and Symphonic automotive families target the post-2026 vehicle architecture pivot.

• (5) Industrial IoT & edge.

  14 billion+ industrial IoT endpoints deployed by 2025, each needing low-power, long-life timing references that survive battery-powered operation across factory-floor environments. SiTime's Cascade family delivers ±20 ppm accuracy at 32 kHz while consuming 1.2 μA — extending node battery life to seven years where quartz alternatives struggle past three.

Market Sizing — Where The Money Is Today, And Where It's Migrating

How To Position — Two Tiers, One Recommendation

‍The investable universe for this thematic is fundamentally narrow, by design rather than oversight — the technology shift is real but the market is small (~$8bn) and the supplier base is concentrated. We group the public-market opportunity into two tiers. Tier 1 (MEMS disruptor) is the LC investment recommendation; Tier 2 (high-end quartz incumbents) is included for awareness only — these are defensive incumbents with credible niches, not active recommendations.

‍LC RECOMMENDATION — Tier 1 only. SiTime (SITM US) is the only listed pure-play on the MEMS-versus-quartz technology shift and the sole name we are recommending as an active expression of this thesis. Sizing should reflect single-stock concentration risk.

Tier 1 · Recommendation — The MEMS Disruptor: One Listed Pure-Play

‍SiTime is the only credible listed pure-play on the MEMS-versus-quartz technology shift. Some MEMS technology is also being developed in-house at large diversified semis (Microchip, NXP, Murata) but in no case is it a meaningful slice of group revenue, and we do not consider these viable expressions of the thematic.‍

Tier 2 · For Information Only — Quartz Incumbents: Defending The Premium Segments

Tier 2 names retain qualified-supplier status across legacy automotive, telecom, and broadcast applications and continue to grow at low-to-mid single digits. These are NOT investment recommendations. They are listed here so readers understand the full competitive landscape and the incumbents being structurally challenged. The risk to these names is not displacement of the installed base but loss of incremental new-socket wins in AI / space / defence — i.e. the segments where the entire market's growth is concentrated.

Data sources. Figures sourced from company filings, Mordor Intelligence, Markets & Markets, IBE Electronics, Verified Market Reports, and channel checks. Estimates of market size and segment share vary materially across third-party research providers; ranges are flagged where this is the case.

Please refer to our full disclaimer for important disclosures and regulatory information.