The first xMEMS speaker did not roar. It buzzed. Specifically, it sounded enough like a mosquito that any normal product manager might have reached for bug spray, not a bottle. Joseph Jiang and his team opened champagne. In that thin, awkward note they heard proof: a piece of silicon, animated by a piezoelectric film, had moved enough air to make sound. The volume was trivial. The consequence was not.
Jiang had spent decades around chips, sensors and audio before arriving at this moment. He understood that semiconductor inventions live two lives. The first happens in equations and laboratory apparatus. The second, much less romantic life happens in yield charts, test fixtures, packaging lines and customer prototypes. A chip that sings once is an experiment. A chip that sings tens of thousands of times, consistently and at a competitive cost, is a business.
The distinction explains both Jiang’s career and xMEMS, the Santa Clara company he started with Jemm Liang in 2017 and formally launched in 2018. Jiang is an electrical engineer by education, with degrees from National Chiao Tung University and San José State University. Yet his professional education came from the seams of the semiconductor trade: technical marketing at Cadence, sales at Applied Materials, startup leadership at Mobilic Technology, motion processing at InvenSense, smartphone noise suppression at Fortemedia and intelligent audio at Knowles. He learned where an invention meets a purchase order.
01 / The betA speaker with no coil, cone or polite respect for precedent
Traditional miniature speakers rely on a coil and a moving diaphragm. xMEMS proposed to put actuation and membrane into a monolithic silicon structure, fabricated with semiconductor methods. The pitch was smaller size, repeatable manufacturing, fast mechanical response and resistance to shock, dust and water. The practical difficulty was equally plain: microphones only need to detect tiny pressure changes, while speakers must push enough air for a human ear to notice. Silicon is disciplined. Bass is not.
The opening appeared when advances in piezoelectric materials made a new architecture conceivable. Jiang and his co-founders placed their bet before a complete supply chain was waiting for them. They bought blank wafers, worked with Japanese partners to add piezoelectric material and ran experiments in rented laboratory space. Once a chip made that famous insect noise, they carried the evidence to TSMC and argued that xMEMS deserved access to piezoelectric wafers.
Winning a foundry relationship was only the beginning. The team tested tens of thousands of dies. It designed test machinery, commissioned equipment makers to build it, then operated the gear itself. When a speaker failed to shine inside an ordinary earbud enclosure, engineers learned 3D printing and made their own earbuds. Customers liked the prototypes enough to use them as development references. A new component had dragged a new little ecosystem into existence behind it.
The unglamorous curve that made the idea real
Reported by Jiang after tens of thousands of dies and a production ramp that exceeded the original schedule by more than a year.
02 / The operatorWhat a commercial career notices that a laboratory can miss
Jiang’s contribution is easiest to see as translation. At InvenSense, he led parts of the motion-processing business and contributed to the company’s 2011 public offering. At Fortemedia, he helped win noise-suppression programs with major smartphone makers. At Knowles, he worked on intelligent-audio software. The path moved through motion, microphones, algorithms and customers before landing on the speaker itself. By the time he founded xMEMS, he had watched new component categories earn their way into consumer devices.
He also knew that technical novelty cannot demand infinite patience from buyers. The first Montara-based products occupied expensive corners of personal audio. Cowell aimed lower in cost and broader in true wireless earbuds. Cypress attacked the low-frequency problem with what xMEMS calls Sound from Ultrasound: an ultrasonic carrier is modulated, then demodulates in air into audible sound. Sycamore extended the approach into a one-millimeter-thin, near-field loudspeaker intended for open-fit wearables, glasses and other narrow spaces.
The product names are arboreal, but the strategy is geological: add layers, compress them, make the stack useful. Jiang does not present xMEMS as a lone inventor’s workshop. He repeatedly credits the engineering group, many of them Taiwanese semiconductor veterans. He jokes that a handful selected at random could form a band because so many are musicians or serious music enthusiasts. For an audio company, listening is a technical competency with an excellent soundtrack.
03 / The cultureProduct first. Fun, with permission, a close second
Jiang calls himself a businessperson, which makes his account of xMEMS culture more interesting. He has described the company’s order of operations as product first and fun second. The fun is not compulsory cheer. It is permission to make something strange after the committed work is safe.
One group wanted to assemble a display using sixteen tiny speakers. The idea was expensive and had no obvious customer. Jiang made company materials available, with the understanding that the experiment happen outside core work. Another experienced engineer joined after asking to stop managing a large team and simply make an earphone. Jiang gave him room. Engineers who requested money to develop new test equipment often received a quick yes, and particularly good results could earn bonuses.
This generosity had a hard commercial edge. If xMEMS was redefining the earbud, Jiang reasoned, outsiders would not yet know how to finish the job. The company had to explore the missing steps itself, then turn useful discoveries into reference designs, production methods and, ideally, industry norms. Play was a search function. Speed was a competitive one. He likened product judgment to chess: experience tells you when to move, though not every move proves correct.
One practiced motion, two stubborn problems
04 / The second actAfter teaching silicon to sing, teach it to breathe
The most revealing development at xMEMS is not another speaker. It is a cooling chip. Compact electronics are acquiring more local compute while keeping their thin frames and limited air volume. Heat becomes the tax on every clever feature. Conventional fans bring thickness, bearings, noise and vibration. A piezoMEMS membrane already knows how to move quickly and reliably. The company’s next question was whether it could move air for thermal management rather than music.
That question produced µCooling, a family of solid-state active cooling devices aimed at phones, smart glasses, SSDs and other cramped electronics. In 2026, xMEMS introduced a smaller model for smart glasses and sharpened its attention on AI wearables, where comfort, battery life, audio and heat all compete for the same millimeters. The move from speakers to cooling looks broad only from the product shelf. From the wafer, it is adjacent.
This is the strategic pattern worth borrowing. A deep-tech company can spend years earning one difficult physical capability. Its first market pays for the education. The second market tests whether that education has become a platform. xMEMS learned to deposit, shape, actuate, package and manufacture piezoelectric membranes. Audio was the first answer. Airflow is the next.
05 / The aspirationShip the difficult thing, then make room for another
In 2023, Jiang spoke plainly about what he wanted next: focus on products, move the company toward public-market readiness, let employees share more fully in the value of their work, then keep making interesting things in MEMS. By October 2025, a $21 million Series D had supplied fresh capital for mass production, customer partnerships and research. The company reported more than 250 granted patents at the time. The laboratory buzz had accumulated a considerable paper trail.
Yet the founder’s useful lesson is not found in the patent count or financing headline. It sits in the gap between 18 months and more than 30. That gap contains the test machines nobody planned to invent, the earbud enclosure nobody else knew to print, the wafers that did not yield and the repeated argument that a speaker could be made differently. Hardware makes optimism submit receipts.
Jiang’s style places curiosity beside those receipts. Let the engineers build the sixteen-speaker contraption. Hire the manager who would rather return to the bench. Celebrate the mosquito, because morale also needs a prototype. Then return to the line and make the next die work.
Silicon Valley prefers its revolutions loud. This one began at the edge of hearing. Its next act may be quieter still: a millimeter-thin device moving cool air behind the frame of a pair of glasses. Joseph Jiang has built a career around noticing such small motions before the market does. The trick is not merely to hear them. It is to manufacture them.