Chip Fab-in-a-Box Could Democratize Semiconductors

When Mitchell Hsing was a grad student at MIT, he and his classmates were doing cutting-edge things using what was essentially chip-industry cast-off equipment. Just imagine what could get done if everyone had their own small fabs to spin up prototypes or manufacture small batches of chips, he thought. The result is InchFab, a startup selling a $5 million to $15 million shipping-container-size clean-room system that can do just about everything a not-so-advanced fab can. The compact size comes as a result of InchFab using much smaller silicon wafers than today’s multibillion dollar fabs do. Smaller wafers means fab equipment can be much smaller and way cheaper. InchFab now has customers all around the world, especially in places where people are looking to train up a chipmaking workforce ahead of planned new fabs.

Where did the idea for InchFab come from?

Mitchell Hsing: It really came from our own issues that we had trying to get chips made at MIT. MIT is arguably one of the best places in the world to do microfabrication-related research, and we were still using equipment from the 1980s that literally needed a floppy disc just to make it work.

Inchfab really kind of grew out of that frustration. In Martin Schmidt’s group, where I was at the time, we were looking at how we solve the problem of lowering the barrier of entry to microfabrication.

The solution was scaling down the size of the existing chip-processing equipment. As a result of that, you could scale the price of the equipment as well. My cofounders and I looked at how the physics and the chemistry changes when you go from, say, a large, conventional 50-gallon-oil-drum-size plasma vacuum chamber to something that’s the size of a 1-liter soda bottle.

How small are the wafers you designed the equipment to work with?

Hsing: In the very beginning, the project was called “One Inch Fab.” It wasn’t just a code name, there’s some reason behind it: The field of view of a stepper [a common photolithography tool], is about an inch by an inch. But we realized that there are many problems with 1-inch wafers. The first is that you can’t buy them, so you have to cut them out yourself, and there’s lots of problems with that. So then we quickly scaled to two inches, and now we’re actually at four inches [about 100 millimeters].

Is the physics different when you’re dealing with a much smaller substrate?

Hsing: The primary differences are on the plasma-based tools, and it’s not the substrate itself that dictates the changes. It’s the size of the plasma chamber. In plasma-based equipment, you have something called a sheath. It’s basically a layer of plasma along the chamber wall protecting the machine from killing itself. That sheath scales with the surface area of the chamber. When you go smaller, the surface area starts to become more prominent over the volume.

Was anything inherently easier when you shrunk it down?

Hsing: Some things come easier—the equipment that you need on the back end, like vacuum pumps, mass-flow controllers, valves, and things like that. Controlling a smaller volume is easier.

What can InchFab do and what can’t it do compared to a regular fab.

👁 Glass panelling divides a room filled with orange light from an industrial area.
Workforce development is one of the big uses of InchFab’s fab-in-a-box.InchFab

Hsing: Our fab has the same process capabilities any other fab in the world would, including lithography, metrology, dry etch, plasma-enhanced chemical vapor deposition processes, atomic-layer deposition, and wet processes. Basically, you name the process, we likely have the process.

Our primary limitation is lithography, in terms of feature size and speed. Feature size scales with the speed, and the speed is important for manufacturing purposes. We could do half micron feature size using photolithography in some production volume. We could go down even smaller, to tens of nanometers. However, the write times are a lot slower, because you’re looking at electron-beam capabilities or imprint lithography.

When you first told me about this idea years ago, I didn’t believe it was possible. What were some of the early reactions to this effort, and what do you say in response to the naysayers?

Hsing: Oh yeah, we still get naysayers. The primary push back we get is that we need to be on large wafer sizes. The argument is: “If I produce on a larger wafer size, I could make more stuff per wafer. And as a result of that, the price per chip comes down.” That argument holds true if you could fill a fab running 10,000 wafers a month.

But what we have shown is that that doesn’t actually make sense. You need to be producing on a wafer size and a fab throughput that matches the markets you’re serving. That’s ultimately what dictates the price per chip. It’s the capital efficiency and utilization of your fab. Those two things have to match, not your wafer size. That’s what we’ve been able to show through the seven years that we’ve been running. Oftentimes [with our smaller wafers] we can be price competitive with an 8-inch foundry today.

What sort of customer can be price competitive in that scenario?

Hsing: It depends, but in general terms the largest markets that we serve today are industrial, sensing, biomedical, aerospace, and defense. We are branching into other areas right now as well, such as compound semiconductors and the whole slew of products and processes that come with that like power and high-frequency RF applications. And then right beyond that, we’re looking at the quantum and photonic space as well. Anything that requires a custom process flow or where the product volumes are not high to begin with is a good fit for us.

Let me just add another point regarding what our fabs are used for. A large part of our business right now is in workforce development. Nowadays everybody in the world pretty much wants to have some type of domestic semiconductor manufacturing capability. And there’s no better way, no cheaper way, to start it than with something like an InchFab. A lot of these countries are trying to build a large, 8-inch or 12-inch fab, which takes five years to build. But they could get started within the first five years with an InchFab. Even more importantly, once they build that fab, they’ll need a trained workforce to go run it. So we actually supply a fab line with a training course. These courses are modeled after the MIT courses that I took, but I designed them such that the student actually tunes all the recipes themselves. They make all the mistakes themselves. That’s how you learn.

Where do you expect InchFab to be five years from now?

Hsing: Really what we set out to do, is to democratize fabrication—to lower the barrier of entry, to enable everybody to be able to manufacture, or at least to play. I would say that what is limiting innovation on the microscale today is people’s access to microfabrication capabilities. I guarantee you, that if you give people these capabilities, they will figure some [stuff] out that people have never thought of before.