The Phoenix startup that peels chips off their wafers with sound — and hands the expensive part back.
The logo, and the pun: "technology is built on us." Every chip sits on a substrate. Crystal Sonic's whole idea is that the substrate should get up and do it again.
Here is a fact about semiconductors that sounds made up but isn't: to build one thin device, the industry routinely grinds away 90% or more of the wafer it was built on. The wafer is often the single most expensive input - for compound semiconductors like gallium nitride, silicon carbide, and gallium arsenide, the substrate can run up to half the total cost of the finished device. And then most of it becomes powder.
If you think about this the way an accountant would, it is a strange line item. You buy a costly crystalline slab, use a sliver of it, and destroy the rest as a matter of routine. Everyone in the industry knows this. It is simply how thinning and layer-transfer have worked for decades - lapping, grinding, chemical-mechanical planarization. The waste is priced in. It is the tax you pay to get a working device off the wafer.
Crystal Sonic, a deep-tech startup that spun out of an Arizona State University lab in 2018, looked at that tax and asked a slightly annoying question: what if you didn't have to pay it? What if, instead of grinding the wafer down to free the device, you could lift the device off cleanly - and keep the wafer whole enough to use again? Not once. Several times.
Their answer is sound. Literally sound. The company's patented process, Sonic Lift-off, uses precisely tuned acoustic waves to separate a thin device layer from its substrate along a controlled plane. The team calls it an "acoustic scalpel," which is a good name because it captures the two things that matter: it cuts, and it cuts precisely. What's left behind is a smooth surface and a substrate that can go back into the line.
The pitch is not really about sustainability, though it is that too. It's about arithmetic. If the substrate is a big chunk of your cost and you throw it away every cycle, then anything that lets you reuse it goes straight to the bottom line. Here is the problem Crystal Sonic is aiming at, drawn as bars.
→ Reuse the substrate instead of destroying it, and both bars turn from cost into savings. That is the entire business, and it is why the numbers are worth staring at.
Two things, and they're related.
Tuned sound waves separate a thin, finished device layer from the wafer along a controlled plane - no grinding, effectively no kerf waste, and a smooth surface left behind. The acoustic scalpel.
The reclaimed wafer goes back into the line to make the next device. Crystal Sonic is building tools to run this at production scale on 6-inch and 8-inch wafers.
Aimed squarely at pricey compound semiconductors - GaN, SiC, gallium arsenide - the substrates behind electrification, 5G, photonics, sensing, and space solar cells.
There is a nice thing about a technology whose selling point is "you spend less money." Sustainability pitches usually ask a customer to accept a cost in exchange for a cleaner conscience. Crystal Sonic's asks a customer to accept a saving in exchange for slightly rethinking a step they already do. The waste reduction is real - reusing a wafer several times obviously beats making dust - but it rides along on the economics rather than fighting them. That's a comfortable place for a hardware startup to stand.
Wafer thinning and layer transfer are not new problems, and Crystal Sonic is not the first company to notice the waste. The incumbents are grinding and chemical-mechanical planarization, which work but destroy the substrate. There are also cleverer alternatives - laser lift-off, controlled-fracture "spalling," and ion-implant slicing methods like the Smart Cut process associated with Soitec. Each has its own trade-offs in stress, surface quality, and which materials it plays nicely with.
Crystal Sonic's argument is that acoustics hit a useful sweet spot: precise enough to define a clean separation plane, gentle enough to leave a reusable surface, and general enough to work across the expensive compound-semiconductor materials that need it most. That is a claim the company is still proving out at scale, which is exactly what the NASA money and the Lam investment are for - building tools that do this reliably on 6-inch and 8-inch wafers rather than lab coupons.
The timing helps. There is an unusual amount of public attention and public money aimed at American semiconductor manufacturing right now, and a lot of it is specifically interested in cost, resilience, and domestic capacity. A startup that can credibly lower the material bill for GaN, SiC, and gallium arsenide devices is pitching directly into that conversation. Phoenix, meanwhile, has been quietly turning into a real chip town - the "Silicon Desert" - which gives a small ASU spinout a surprisingly deep bench of fabs, talent, and institutional interest within driving distance.
None of this guarantees anything. Deep-tech hardware is slow, capital-hungry, and unforgiving; a technique that works beautifully on a small wafer can misbehave the moment you scale the diameter or change the material stack. But the shape of the bet is clean, which is rare. Crystal Sonic is not asking the industry to want something new. It is asking the industry to keep doing what it already does, minus most of the waste.
The technology started as a finding in Mariana Bertoni's research group at ASU. In 2018 she and her former doctoral student Pablo Guimerá Coll co-founded Crystal Sonic to push it out of the lab. CEO Arno Merkle - a materials scientist with two prior startup exits behind him - runs the commercial side.
It is a small, science-heavy team - around eleven people - still tightly wound around ASU. Crystal Sonic hires ASU interns and student researchers for process development and wafer characterization, and it has worked out of shared innovation spaces in downtown Phoenix. This is the unglamorous middle of deep tech: not the demo-day flash, but the grind of turning a real physics result into a machine other companies can bolt into a fab.
The cap table and grant history read like a who's-who of institutions that care about American semiconductor manufacturing - which is to say, most of the ones that matter right now.
$250K investment via the Enabling Future Semiconductors venture competition, tied to Lam Research's ecosystem.
SBIR funding to scale the tool to larger wafers and adapt it for gallium arsenide solar cells.
Federal research backing from the Department of Energy and the National Science Foundation.
Arizona Commerce Authority, Partnership for Economic Innovation, and Plug and Play's accelerateAZ program.