SemiWiki profiles Jerry Chen, who left Apple and Inphi to build an ultra-low-power edge AI chip startup in Taipei
SemiWiki's CEO interview with Jerry Chen traces his path from IC design at Intel and Apple to founding Upbeat Technology in 2021, where the UP201/UP301 RISC-V MCU and UPM01 MEMS sensor are now shipping.

SemiWiki has published a CEO interview with Jerry Chen, founder and chief executive of Upbeat Technology, a Taipei-based fabless startup that pairs ultra-low-power RISC-V microcontrollers with MEMS vibration sensors for edge AI applications[1]. The interview traces a career arc that runs from IC design benches at major Silicon Valley firms to a bet on near-threshold computing as the key to always-on intelligence at the edge.
From Intel's Itanium to founding a startup
Chen received his Master of Science in ECE from the University of Texas at Austin and began working as an IC design engineer in Silicon Valley in 1995, with stints at Intel, Lucent, Apple, and Procket Networks[1]. At Intel he worked on the Itanium processor; he later spent time at Apple before moving into technical marketing. He served as Director of Marketing and VP Marketing at Inphi/Marvell and Genesys Logic, accumulating expertise across CPU architecture, high-speed SerDes, and image processing.
Those experiences led to the founding of Upbeat Technology, with Chen raising between $6 million and $7 million to start the company in 2021. The company is headquartered in Taipei, Taiwan. The team blends expertise in AI, MEMS, and IC design, with members drawn from Intel, Sony, Inphi (Marvell), TSMC, and ITRI.
A two-product architecture built around near-threshold computing
Upbeat's core IP is a power-reduction technique called near-threshold computing (NTC), which drops the SoC supply voltage toward the transistor threshold to slash dynamic power. The UP201/UP301 family MCU integrates two SiFive Essential IP cores and two AI accelerators designed by Upbeat, with its patented Error Detection and Correction (EDAC) architecture enabling reliable operation at those reduced voltages.
The combination of near-threshold operation, EDAC resilience, and SRAM innovation delivers energy efficiency at 16.8 µW/MHz/DMIPS (Dhrystone), scaling up to 400 MHz or 717 DMIPS as a production-ready platform.
On the sensing side, Upbeat's UPM01 and UPM02 high-bandwidth MEMS vibration sensors feature a flat frequency response from 5 Hz to 12 kHz and a signal-to-noise ratio of up to 68 dB. Target applications span several markets:
- Smart factories and overhead cranes, where the UP201 MCU combined with the UPM01 vibration sensor continuously monitors machine vibration patterns in the 20 kHz range, with the company claiming a reduction in unplanned downtime of over 90%.
- AI wearables, smart glasses, and true wireless stereo earbuds, where bone-conduction sensing isolates the speaker's voice from ambient noise.
- Drones and robotic platforms, where the UPM01 detects ultra-low-level micro-vibrations associated with mechanical imbalance, fatigue, or wear, enabling predictive maintenance and propeller health monitoring.
Shipping timeline and market context
The UPM01 has been in high-volume production and shipping throughout 2025, demonstrating proven performance in real-world smart glasses and TWS applications. Mass production shipments for the UPM01/UPM02 are underway, with the UP201/UP301 ICs scheduled to ship in October 2026.
The timing aligns with a broad market expansion. According to Mordor Intelligence, the global MEMS sensor market is forecast to grow from $20.24 billion in 2026 to $29.08 billion by 2031, driven by pervasive edge-AI deployment, miniaturization in wearables, and surging demand for smart industrial systems.
The full SemiWiki interview covers Chen's views on the competitive landscape for ultra-low-power MCUs and how Upbeat plans to expand its design-win base as the UP201/UP301 moves from engineering samples into volume production later this year.
Written by Electronics Insider's automated desk from the sources above and published automatically. How we work.
Related
Design & EDAETH Zürich and NYU show JEDEC test loops miss worst-case HBM2 power by 23.6%
Researchers at ETH Zürich and NYU tested 36 real HBM2 chips and found JEDEC measurement loops understate peak power by 23.6%, with read ops and data-bus toggling as the main drivers.
15 Oct 2026
Design & EDAFlowAccel's Joel Jorgensen argues chip teams must design their workflows before AI can help them hit tape-out
Joel Jorgensen of FlowAccel argues that AI tools alone won't save a tape-out schedule - teams must first design the cross-team workflows that let those tools compound into real gains.
15 Oct 2026
SemiconductorsETH Zürich and NYU find JEDEC test loops underestimate HBM2 power by up to 35.4%
Researchers at ETH Zürich and NYU tested 36 real HBM2 chips and found JEDEC-specified current loops can underestimate peak power by up to 35.4%, with implications for AI accelerator thermal design.
15 Oct 2026