# How Chips Use On-Chip Piezoelectric Fans for Active Cooling

An episode of [The Hardware Podcast with Fexingo: Chips, Devices, and Electronics Engineering Conversations](https://invisiblenetwork.ai/podcast/1896799982), 6 Jul 2026, 13 min.

Audio: https://audio.fexingo.com/business/the-hardware-podcast/episode-0094.mp3

Episode 94 of The Hardware Podcast dives into the latest innovation in chip-level thermal management: on-chip piezoelectric fans. These tiny vibrating blades, just millimeters wide, can generate targeted airflow over hot spots without the bulk and noise of traditional fans. Lucas explains how researchers at Purdue and a startup called Frore Systems are using piezoelectric materials—the same stuff in inkjet printers and ultrasonic sensors—to create fans that consume under 10 milliwatts yet reduce hotspot temperatures by over 10 degrees Celsius. Luna questions the reliability of moving parts on a chip and asks whether this could finally replace thermal paste in thin laptops. The hosts also touch on how this technology might scale to data centers, where fan arrays could cool server racks with zero moving bearings. A concrete look at how a century-old principle is being miniaturized to solve a modern heat problem, plus a brief moment on how listener support keeps this kind of deep-dive coverage ad-free. #PiezoelectricFan #ChipCooling #ThermalManagement #ActiveCooling #FroreSystems #PurdueUniversity #Semiconductor #MEMS #HeatDissipation #FanTechnology #DataCenterCooling #ThinLaptop #Noctua #Silicon #ElectronicsEngineering #Technology #FexingoBusiness #BusinessPodcast Keep every episode free: buymeacoffee.com/fexingo

## Transcript

Lucas: Alright, I want to start today with a number: ten degrees Celsius. That's the temperature drop researchers at Purdue recently reported by sticking a tiny vibrating blade onto a running chip.
Luna: A vibrating blade. Like a microscopic fan made of a tuning fork?
Lucas: Exactly. It's called a piezoelectric fan. The blade itself is a strip of flexible material — often Mylar or metal — bonded to a piezoelectric actuator. When you apply a voltage, the actuator bends, and the blade whips back and forth at its resonant frequency, typically between 60 and 200 hertz.
Luna: And that generates enough airflow to actually cool a chip? These things are millimeters wide.
Lucas: That's what surprised me too. The key is that they're placed directly over the hotspot. A traditional fan moves air over the entire heatsink, which is fine, but a piezoelectric fan targets a specific millimeter-scale region. In the Purdue experiments, they cut hotspot temperature from about seventy-five degrees down to sixty-five — a ten-degree drop — using a fan that consumes less than ten milliwatts.
Luna: Ten milliwatts. Compare that to a typical laptop fan that pulls maybe one to two watts. So this is two orders of magnitude less power for local cooling.
Lucas: Right. And the noise profile is completely different. A spinning fan has bearings, turbulence, that whirring sound. A piezoelectric fan just hums at a low frequency — if you're at sixty hertz, you barely hear it. Some designs are virtually silent.
Luna: But the obvious question: moving parts on a chip. The blade is flexing millions of cycles per hour. Does it fatigue and snap?
Lucas: That's been the big reliability question since researchers first played with these things in the 1970s. The answer from modern materials is surprisingly good. The typical blade is designed to operate at resonance with a stress point well below the fatigue limit of the material. Companies like Frore Systems — that's a startup out of San Jose — claim lifetimes exceeding ten years of continuous operation.
Luna: Frore Systems. I've seen their press around 'AirJet' — that's a solid-state cooling module, right? Not a flapping blade?
Lucas: Frore's product is actually a different approach — they use a vibrating membrane to push air through microchannels, not a cantilever blade. But the physics is similar: piezoelectric actuation, no bearings, no fan blades. They've gotten a lot of attention because they're targeting ultra-thin laptops where you can't fit a traditional fan at all.
Luna: So we have two parallel tracks: the cantilever blade approach from academia and the membrane approach from Frore. Which one is more likely to end up in a consumer device first?
Lucas: I'd bet on the membrane approach for phones and tablets, because it's truly flat. The blade needs a few millimeters of clearance to flap. But for chips with exposed hotspots — say, a power amplifier on a 5G base station — a blade array could be mounted directly onto the package. There's actually a company called Piezo Sonic, spun out of the University of Tokyo, that's making arrays of these fans that look like a tiny comb.
Luna: A comb of vibrating blades. That sounds like it could cool multiple hotspots independently. You could have one blade aimed at the CPU, another at the GPU, each running at its own frequency.
Lucas: Precisely. And that's the real advantage over a single fan: spatial control. In a data center, you could imagine an array of these things mounted on a cold plate, each blade targeting a specific row of chips. A study from Georgia Tech showed that using eight piezoelectric fans on a server CPU reduced the required airflow from the main chassis fan by thirty percent, which translates directly to energy savings.
Luna: Let's talk about the material side. The piezoelectric material itself — is it lead zirconate titanate, the usual PZT? Or something more exotic?
Lucas: Most prototypes use PZT, yes. It's cheap, well-characterized, and has a high electromechanical coupling coefficient. But PZT contains lead, which runs into RoHS restrictions in Europe. So researchers are also looking at lead-free alternatives like potassium sodium niobate, or even polyvinylidene fluoride — a flexible polymer that's much easier to deposit on a chip.
Luna: And the fabrication process? Can you integrate a piezoelectric fan into a standard CMOS flow?
Lucas: Not directly. The fan is a separate component that gets attached after the chip is packaged. But there's a fascinating line of research called 'on-chip piezoelectric fans' where the entire actuator and blade are microfabricated using MEMS techniques — essentially etching the blade out of silicon and depositing a thin film of PZT on top. That would let you integrate the fan right onto the die.
Luna: So the fan becomes part of the chip itself. That's a huge leap from a discrete component.
Lucas: It is. And it's still very much in the lab. The challenge is that a mems scale blade generates a tiny airflow — we're talking single-digit milliliters per second. To move enough heat, you'd need hundreds of blades operating in parallel. But if you can fabricate them cheaply as an array, it could replace thermal paste and heatsinks in some applications.
Luna: Let's bring this back to a concrete use case. The latest thin laptops — like the MacBook Air or Dell XPS — use passive cooling with just a heat spreader. But as chips get more powerful, they hit thermal throttle under sustained load. Could a piezoelectric fan allow those laptops to keep performance up without adding bulk?
Lucas: That's exactly the pitch from Frore Systems. Their AirJet module is about two and a half millimeters thick — thin enough to slip under the keyboard. They claim it can dissipate about ten watts in a whisper-quiet manner. If you were to put two of those on a laptop's CPU, you could handle a twenty-five watt thermal load without ever spinning up a noisy fan.
Luna: Twenty-five watts is roughly the TDP of a mid-range laptop chip. So you could have a fanless design that actually runs at full clock speed under load. That changes the whole product segment.
Lucas: Yes. And the reliability argument gets even stronger when you consider that there are no bearings to wear out. The piezoelectric material does degrade over time — it loses polarization — but that's a slow process. Frore says their modules can run for over ten years at sixty degrees Celsius without failure.
Luna: I want to push back a little: ten years is a long time, but consumer electronics are often replaced every three to five. The bigger question is cost. How much does a piezoelectric fan add to the bill of materials compared to a conventional fan?
Lucas: Right now, a lot. The piezo ceramic itself is cheap — maybe a few cents — but the manufacturing process to attach and tune each blade is manual. Frore hasn't disclosed pricing, but industry estimates suggest a module might cost five to ten dollars. A standard laptop fan costs about two to three dollars. So you're looking at a premium of two to five times.
Luna: That premium is acceptable in a thousand dollar plus laptop, but not in a budget Chromebook. So this starts in high-end thin and lights and then hopefully scales down.
Lucas: Exactly the pattern we've seen with other cooling innovations — vapor chambers started in gaming laptops, now they're in mid-range phones. Piezoelectric fans will follow a similar curve if the reliability holds up.
Luna: Let's talk about the audio side. You mentioned the hum at sixty hertz. In a quiet room, that could be annoying. Is there any work on making the frequency inaudible?
Lucas: Some designs push the resonance up above twenty kilohertz — into the ultrasonic range. Then you can't hear it at all. But the trade-off is that higher frequencies reduce the blade's amplitude, which reduces airflow. So you have to find a sweet spot: around one hundred to two hundred hertz is typical, which is a low hum that most people find less annoying than a whining fan.
Luna: And the amplitude itself — how far does the tip of the blade travel? A few millimeters?
Lucas: In typical designs, the tip displacement is about one to three millimeters peak to peak. That's enough to generate a local air velocity of maybe one to two meters per second. For context, a desk fan pushes air at about three meters per second. So it's comparable, just over a much smaller area.
Luna: That's impressive for something that's essentially a strip of plastic wiggling back and forth.
Lucas: It really is. And I think the most exciting part is that this technology is finally moving out of the lab. Purdue has licensed some of its IP to a startup called CoolChip, and Frore announced a partnership with Intel at CES last year. So we might see the first consumer devices with piezoelectric cooling in late 2026 or early 2027.
Luna: Speaking of keeping up with developments like these — and I know we've mentioned this before — the reason we can spend a whole episode drilling into a niche like piezoelectric fans is that this show is listener-supported. No ads, no sponsors dictating topics. A handful of listeners chip in monthly through buy me a coffee dot com slash fexingo, and that literally funds keeping episodes like this one ad-free and deep.
Lucas: Yeah, and we're grateful for that. It means we can cover a topic like MEMS cooling without having to pitch it to an advertiser first. If you find value in these deep dives, that's the way to keep them coming.
Luna: Exactly. So back to the chips — one angle we haven't touched: could this technology scale to data center level cooling? Not just a single chip, but a whole rack?
Lucas: There's work on that too. Researchers at the University of Arkansas designed a 'piezoelectric fan array' that mounts on a server's cold plate. They used thirty-two blades arranged in a grid, each one individually addressable. The idea is that you can modulate the cooling locally based on which cores are hot. In simulations, they reduced the total cooling energy by forty percent compared to a fixed-speed rack fan.
Luna: Forty percent is enormous. Data centers are already under pressure to cut power consumption. If piezoelectric fans can shave off even ten percent of cooling costs, that's millions of dollars at scale.
Lucas: And the reliability argument is even stronger in a data center — you don't want to replace failed fans in a server every three years. Piezoelectric fans have no bearings, so the mean time between failures could be dramatically higher. Some estimates put it at over a hundred thousand hours, compared to about fifty thousand for a good ball-bearing fan.
Luna: But there's a catch: the blades might collect dust. And dust on a vibrating blade changes its resonant frequency. If it drifts off resonance, the airflow drops off a cliff.
Lucas: That's a real concern. In clean-room environments like a data center, dust is minimal, but it's not zero. Some designs incorporate a self-cleaning mode where you briefly drive the blade at a higher amplitude to shake off particles. But that adds complexity. I think for now, the most promising near-term application is in sealed consumer devices — phones, tablets, thin laptops — where dust ingress is low.
Luna: So we're looking at a technology that's proven in principle, has a clear performance advantage, but is still battling cost and reliability questions. Sounds like the typical journey from lab to product.
Lucas: Exactly. And as with many semiconductor technologies, the first adoption will come where the pain point is highest. For chips that are thermally limited — like the latest mobile processors that throttle within seconds — a ten-degree drop could mean sustained peak performance. That's a feature worth paying for.
Luna: Alright, I'm convinced. Let's see if the 2027 MacBook Air has a silent, vibrating blade inside.

Shows are listed from Apple Podcasts' charts and directory, and Podcast Index where it is used. Every show and episode belongs to its publisher and plays from the publisher's own feed; this site hosts none of them.