Silicon speakers arrive: xMEMS Cypress and Sycamore explained

Something genuinely historic happened at the junction of September 2025 and CES 2026, and it did not involve a new flagship turntable or a reference amplifier with a six-figure price tag. It happened at the microscopic scale — literally inside a silicon wafer — and it has the potential to rewrite the engineering assumptions that have governed every loudspeaker and headphone driver made in the past century and a half. xMEMS, the California-based firm that has spent years accumulating a portfolio of more than 300 granted patents on its piezoMEMS platform, announced two milestones in quick succession: first, that its Cypress driver had reached mass-production readiness, and second, that a new device called Sycamore was being unveiled ahead of CES 2026 as the world's first silicon full-range loudspeaker. Both announcements deserve careful unpacking, because the implications stretch well beyond TWS earbuds and into questions about what a speaker driver even is.

Why the conventional driver is still a 19th-century mechanism

Before getting into what xMEMS has done, it helps to appreciate what everyone else is still doing. The dynamic driver — a voice coil suspended in a magnetic gap, attached to a cone or dome diaphragm — was patented by Rice and Kellogg at General Electric in 1925, and the fundamental operating principle has not changed since. You pass a varying electrical current through the coil, the coil experiences a force inside the magnetic field, the diaphragm attached to it moves, and air is displaced. Simple, elegant, and remarkably difficult to perfect.

The problems are well understood to anyone who has spent serious time with high-end audio. A moving voice coil has mass. That mass resists acceleration at high frequencies, which is why tweeters need to be small and light. The suspension — the spider and surround — introduces non-linearities that worsen with excursion. The motor system requires tight mechanical tolerances. The whole assembly is manufactured by hand or semi-automated processes with significant unit-to-unit variation. And at the system level, the driver needs a crossover to divide the frequency range, introducing phase shifts and insertion loss at every junction.

Planar magnetic and electrostatic transducers address some of these issues — distributed force across a large, thin diaphragm, no voice coil, lower moving mass — but they come with their own compromises: sensitivity, impedance characteristics, and enclosure requirements that make them largely impractical for portable applications. If you want to understand some of those electrical trade-offs, our Impedance glossary entry is a good starting point, and the relationship between driver efficiency and system output is covered in our Sensitivity explainer.

xMEMS is doing something fundamentally different from all of the above. Its drivers are fabricated on silicon wafers using the same photolithographic processes used to make microprocessors and MEMS microphones. There is no voice coil, no permanent magnet, no spider, no surround in the conventional sense. The actuator is piezoelectric: apply a voltage to the silicon structure and it flexes. Reverse the voltage and it flexes the other way. Do this at audio frequencies and you have a loudspeaker. Because the entire device is produced in a semiconductor fab, dimensional tolerances are measured in microns rather than millimetres, and unit-to-unit consistency approaches what we expect from integrated circuits rather than from hand-assembled driver baskets.

Cypress: the first MEMS driver that can actually power ANC earbuds

xMEMS announced on 10 September 2025 that Cypress had reached mass-production readiness, with customer shipments expected to ramp through 2026. That timing matters: reaching mass-production readiness means a device has cleared the manufacturing yield, reliability, and supply chain hurdles that separate a working prototype from something a major consumer electronics brand can design into a product and ship in volume.

The headline specification for Cypress is remarkable: it delivers over 140 dB SPL at frequencies down to 20 Hz. To contextualise that figure, xMEMS states this represents roughly 40 times the low-frequency output of prior xMEMS designs. That 40× figure is in terms of acoustic pressure, which translates to a 32 dB improvement — a very large step in any acoustic engineering context.

Why does 140 dB SPL at 20 Hz matter specifically for ANC earbuds? Active noise cancellation works by generating an anti-noise signal — a phase-inverted replica of the incoming noise — and summing it with the incident sound field at the ear. The more accurate and powerful that anti-noise signal, the deeper the cancellation. Low-frequency noise — aircraft cabin rumble, air conditioning, train vibration — is particularly difficult to cancel because it requires significant diaphragm excursion, and MEMS drivers have historically been limited in their low-frequency capability. Prior xMEMS designs were competent in the midrange and treble but could not generate the low-frequency acoustic pressure needed to operate as a primary driver in an ANC system. Cypress clears that bar.

Cypress pairs with the Alta-S driver ASIC — a companion chip that handles the drive signal processing. The complete package occupies just 46 cubic millimetres and weighs 98 milligrams. For comparison, a standard AAA battery is approximately 11,500 cubic millimetres. The entire driver and amplifier system for one ear fits in a volume you could balance on your thumbnail.

The operating principle behind Cypress is what xMEMS calls sound-from-ultrasound modulation. Rather than directly driving the diaphragm at audio frequencies, the driver operates at an ultrasonic carrier frequency that is then modulated to produce the desired audio signal. This approach sidesteps some of the mechanical resonance constraints that limit conventional direct-radiating piezo designs at low frequencies, allowing the device to achieve useful output well into the bass octaves despite its microscopic dimensions. The precise implementation is covered by xMEMS's extensive patent portfolio — the company cites more than 300 granted patents on its piezoMEMS platform — so independent verification of the specific mechanism awaits detailed technical publications and third-party measurement.

Sycamore: a silicon full-range loudspeaker at 1.28 mm

If Cypress is an evolution of what xMEMS has been building towards — a MEMS driver capable of replacing the dynamic driver in a TWS earbud — then Sycamore is a more conceptually radical announcement. Unveiled on 9 December 2025 ahead of CES 2026, Sycamore is described as the world's first silicon full-range loudspeaker.

The specifications are almost difficult to accept as real: 1.28 mm thick, 150 milligrams, and capable of covering the full audio frequency range without a crossover. xMEMS states it is up to 90% lighter than conventional speakers of comparable acoustic function. A 150 mg driver. To put that in perspective, a single US dollar bill weighs approximately one gram — Sycamore weighs about 15% of that.

The full-range capability is the most technically significant claim. Every conventional loudspeaker system above a certain performance threshold uses multiple drivers — a woofer, a midrange, a tweeter — because no single dynamic driver can efficiently cover the full 20 Hz to 20 kHz bandwidth with acceptable linearity. The crossover network that divides the signal between those drivers is simultaneously a necessary evil and one of the most critical design elements in any loudspeaker, influencing phase coherence, power handling, and cost. If a single silicon die can genuinely cover the full audio band, the implications for system design — in earphones, hearing aids, augmented reality headsets, and conceivably in near-field monitoring — are profound. For those curious about how crossover design shapes what you hear from a conventional standmount, our KEF LS50 Meta review (check price) discusses Uni-Q co-axial topology as one approach to the phase-coherence problem, and our guide to the best standmount speakers touches on how driver integration affects the listening experience.

How piezoMEMS actually makes sound

Let us step back and explain the underlying physics for readers who are less familiar with MEMS technology, because the operating principles are genuinely different enough from dynamic drivers to warrant careful explanation.

MEMS stands for Micro-Electro-Mechanical Systems. It is a fabrication technology — not a single material or mechanism — that uses semiconductor manufacturing processes to create miniaturised mechanical structures on silicon. MEMS sensors are everywhere: the accelerometer in your smartphone, the pressure sensor in your car's tyre monitoring system, the microphone in your laptop's lid. The microphone application is particularly relevant because MEMS microphones are the inverse of what xMEMS is doing — they convert sound pressure into an electrical signal, while a MEMS speaker converts an electrical signal into sound pressure.

xMEMS uses the piezoelectric effect as its actuation mechanism. Certain materials — including specially processed silicon structures — deform mechanically when a voltage is applied across them. This is the same principle used in quartz crystal oscillators, ultrasonic transducers, and piezoelectric buzzers. The challenge in using piezoelectric actuation for high-quality audio is achieving sufficient excursion (diaphragm displacement) across a wide frequency range with low distortion. Conventional piezo buzzers sound harsh and limited precisely because their resonant behaviour dominates the response and they lack the excursion for meaningful bass output.

xMEMS's approach — particularly the sound-from-ultrasound modulation used in Cypress — addresses the excursion limitation by shifting the operating frequency of the actuator to the ultrasonic range and then using signal modulation to extract audio-frequency output. Think of it as analogous to how an AM radio transmitter works: a high-frequency carrier wave is modulated by a low-frequency audio signal, and a receiver demodulates it to recover the audio. In the acoustic domain, the ear and the air itself act as the demodulator, but the details of how xMEMS achieves this in a sealed silicon package without generating audible artefacts from the ultrasonic carrier are non-trivial and represent the core of their intellectual property.

Because the entire driver is fabricated on silicon using photolithographic processes, the consistency achievable across a production run is dramatically better than any hand-assembled dynamic driver. In high-end audio we talk frequently about driver matching — the process of measuring and pairing individual drivers that measure similarly, to ensure left-right balance in a stereo system. With semiconductor fabrication tolerances, driver matching in the traditional sense becomes largely unnecessary. This has obvious implications for quality control and for the performance of stereo and multichannel systems.

What this means for the headphone and earbud market

The immediate commercial application for both Cypress and Sycamore is the TWS earbud and headphone market. The competitive pressure in this category is intense, and the differentiation between products increasingly comes down to ANC performance, battery life, and acoustic tuning rather than driver technology — because driver technology has been largely static. A MEMS driver that can match or exceed the bass output of a conventional dynamic driver, in a package that is orders of magnitude smaller and lighter, with better unit-to-unit consistency and potentially lower power consumption, gives earbud designers a genuinely new toolkit.

For listeners who prefer open-back headphones for their soundstage and imaging characteristics — and if you are unsure about the trade-offs, our Open-Back vs Closed-Back explainer is worth reading — the implications of MEMS drivers are somewhat more distant. The acoustic advantages of an open-back design like the Sennheiser HD 660S2 (check price) are primarily about the interaction between the driver and the listening environment, and those interactions are not fundamentally changed by swapping the driver technology. But MEMS fabrication tolerances could eventually mean more consistent stereo imaging in any headphone format.

For ANC specifically, the 140 dB SPL figure for Cypress is not a listening level — it is a measure of the available acoustic headroom for the anti-noise system to work with. More headroom means the ANC processor has more range with which to generate the cancellation signal, which should translate to deeper and more consistent noise reduction across the bass frequencies where ANC has traditionally been weakest.

The longer view: silicon in every speaker?

It would be premature to declare the dynamic driver obsolete. The technology has had more than a century of refinement, there is an enormous installed base of manufacturing expertise, and the cost of silicon fabrication — while falling — is not trivial. The companies making high-end moving-coil drivers, from boutique Scandinavian specialists to the vertically integrated giants, have accumulated decades of acoustic and mechanical knowledge that does not become irrelevant overnight.

But the trajectory is clear. xMEMS has demonstrated, at least on paper, that silicon-based acoustic transducers can now compete with dynamic drivers across the full audio frequency range. The 300+ granted patents on their piezoMEMS platform represent a significant defensive moat, and the pairing of Cypress with a dedicated ASIC — the Alta-S — suggests the company is thinking in terms of complete systems rather than just components. The 46 cubic millimetre package size means these drivers will appear first in applications where space is at an absolute premium: TWS earbuds, hearing aids, augmented reality glasses, bone-conduction supplements.

Whether the technology ultimately scales to the ear-cup headphone or the bookshelf speaker remains to be seen. The acoustic requirements of a 40 mm headphone driver are quite different from those of a 1.28 mm MEMS die, and the bass extension needed for a satisfying full-range loudspeaker at realistic listening distances is a different problem again. But the Sycamore announcement — a full-range silicon loudspeaker, however small — is a proof of concept that the technology can cover the full bandwidth. That is an important threshold to have crossed.

For the Australian market, the practical takeaway in 2026 is to watch TWS product launches carefully for any mention of xMEMS driver technology. If Cypress reaches earbuds in volume shipments through this year as expected, we should start to see products in the market by the time the Australian summer rolls around — and the ANC performance of those products will be the acid test of whether the real-world results match the silicon specifications. We will be measuring and listening when they arrive.

In the meantime, the best conventional in-ear and over-ear options remain dynamic and planar magnetic designs. But for the first time in a very long time, there is a genuinely new driver technology on the horizon that is not simply an incremental refinement of a century-old mechanism. That is worth paying attention to.