For decades, the MEMS and sensors industry competed on specifications: smaller, cheaper, more accurate. That race is not over, but it is no longer the whole game. Industry leaders explained the shift during the 2026 MEMS & Sensors Technical Congress (MSTC) at SEMI headquarters in Milpitas, California, and here are seven themes that are transforming the industry.
1. Sensors Become the Context Layer for AI
AI is only as useful as its picture of the physical world. The winning sensor platforms will turn raw signals into context about a user, a device and an environment, all without burning through power. Qualcomm's Ana Londergan described a sub-milliwatt, always-on sensing hub that fuses data without waking the CPU, GPU or NPU, making proactive personal AI possible within tight power, latency and privacy limits. Realistic training data matters too: S. Siddharth of STMicroelectronics showed a Sim2Real IMU model that adds real-world noise and drift to simulation data, reaching more than 97% correlation with physical sensors. The advantage is shifting from capturing signals to interpreting them.
2. Physical AI Is the Next Growth Engine
Humanoid robots turn sensing into a volume story. Uma Krishnamoorthy of TDK InvenSense noted that a single humanoid may carry more than 100 sensors and pointed to a potential market approaching 400 million sensors a year by about 2035. The obstacles are just as clear: reliability, standards, cybersecurity, functional safety and public trust. Whoever solves them alongside the hardware will define the category.
3. Touch Is the Hardest Frontier
Robots can see far better than they can feel. Sondra Hellstrom of Bosch Sensortec explained that reliable manipulation depends on grip force, slip and shear, and a capable robotic hand may need 80 to 200 pressure sensors. Soft gripping materials drift and wear, and touch data is much harder to simulate or transfer than vision data. Synchronized multimodal sensing, accurate error models and self-calibration will become real differentiators.
4. MEMS Is Expanding Beyond Sensing
AI creates physical problems, and MEMS is well placed to solve them. Curtis Ray of Omnitron Sensors showed MEMS micromirrors forming reconfigurable optical paths in AI data centers, a market expected to exceed $8 billion by 2030. Vincent Destefanis of CEA-Leti presented a photonics demonstrator with 12 to 18 nanoseconds of latency. At the edge, Kevin Hughes of xMEMS tackled heat, with a piezoMEMS microcooler that lowered AI-glasses case temperature by about 13°C. Optical routing, cooling and control are all open territory.
5. Trust Comes From Co-Designing Hardware and Analytics
Algorithms cannot rescue weak measurements. Radislav Potyrailo of GE Vernova described gas sensing that pairs multi-response transducers, controlled excitation, ASIC readout and edge analytics, delivering 20 to 100 times greater sensitivity than conventional resistance measurement. Hossain Fahad of Serinus Labs showed a chemical-sensitive transistor that detects trace hydrogen from early lithium-ion battery failure, giving more than 20 minutes of extra warning over a thermal-conductivity sensor. The best intelligent systems start with hardware designed to expose useful physical information.
6. BioMEMS Borrows Semiconductor Discipline
Biology increasingly depends on dimensions that wafer-scale manufacturing already controls. Bryan Rice of Exokeryx presented a chip that isolates extracellular vesicles using electrically controlled capture and release, with no labels needed. A 20-nanometer change in hydrogel thickness could drop recovery to the noise floor, so process control, down to roughly one-nanometer metrology, becomes part of the product. Jochen Hoffman of Bosch Research applied the same logic to point-of-care diagnostics: miniaturization, automation and repeatability.
7. The Dataset May Be the New Moat
Sensors create data, data improves algorithms, and deployed algorithms generate more data. Shasha Jumbe of Level 42 AI presented the Vibrome, which separates cardiac, respiratory and other signals from skin vibrations; tuberculosis validation across 2,583 participants in five countries reported an externally validated AUC of 0.926. Manny Singh of SensTek Diagnostics described a reusable reader with disposable cartridges, growing from hardware into recurring revenue, licensing and data. Companies that own this loop may capture more value than component suppliers alone.
The Bottom Line
Individual sensor performance remains essential, but it is no longer enough. Winning products will combine differentiated transduction with sensor fusion, contextual intelligence, simulation, self-calibration, functional safety, advanced packaging and scalable manufacturing.
That points to real collaboration opportunities for the SEMI MEMS & Sensors Industry Group (MSIG) community: multimodal sensor modules, tactile sensing and training data, calibration and diagnostics, safety frameworks, and the path from prototype to high-volume production.
The industry is moving from building sensors that measure the world to building intelligent systems that understand it and act on that understanding.
