A 1980s semiconductor fab in Austin, Texas is undergoing a renovation. Texas Electronics InstituteTIE), as it is now known, is preparing to become the world's only advanced packaging plant dedicated to 3D heterogeneous integration (3DHI)—the stacking of chips made from multiple materials, both silicon and non-silicon.
Fab is the underlying infrastructure DARPA's Next Generation Microelectronics Manufacturing (NGMM) program.. “NGMM is focused on revolutionizing microelectronics through 3D heterogeneous integration,” said Michael Holmesmanaging director of the program.
Placing two or more silicon chips in a single package makes them act as if they were a single integrated circuit. It's already putting some of the the most modern processors in the world. But DARPA predicts that silicon-on-silicon stacking will result in no more than a 30-fold increase in performance over what is possible with 2D integration. On the contrary, if you do it using a mixture of materials…gallium nitride, silicon carbideand other semiconductors—could deliver 100-fold gains, Holmes told engineers and other stakeholders at the program's after-party, the NGMM Summit, late last month.
The new plant will ensure that these unusual multilayer chips are prototyped and manufactured in the United States. Startups, and there were many of them at the launch event, are looking for a place to prototype and start producing ideas that are too weird for anything else – and hopefully work around from laboratory to factory the valley of death that claims the lives of many hardware startups.
The state of Texas will contribute $552 million to maintain the factory and its programs, and DARPA will contribute the remaining $840 million. Once the NGMM's five-year mission is completed, the factory is expected to become a self-sustaining business. “We are, frankly, a startup,” said TIE’s CEO. Dwayne LaBrake. “We have more room to take off than a typical startup, but we have to be self-sufficient.”
Launch of 3DHI Fab
Getting to this point will require a lot of work, but the TIE foundry is already off to a flying start. During the inspection of the object, IEEE spectrum saw a few chip production and testing tools at various stages of installation, and met with several engineers and technicians who started work over the past three months. TIE expects all of the plant's tools to be ready by the first quarter of 2026.
The ability of foundry customers to use them in a predictable production process is as important as the tools themselves. This is something that is particularly difficult to develop, TIE representatives explained. At the most basic level, non-silicon wafers are often not the same size. And they have different mechanical properties, meaning they expand and contract depending on temperature at different rates. However, much of the factory's work will involve connecting these chips together with micrometer precision.
The first step in this is the development of what is called a process design kit and an assembly design kit. The first sets rules that limit semiconductor design at the factory. The last one, the assembly design kit, is the real crux of things as it defines the rules for 3D assembly and other advanced packaging.
TIE will then refine these with three 3DHI projects that NGMM calls exemplary. These are a phased array radar, an infrared imager called a focal plane array, and a compact power converter. Piloting them into production “gives us an initial blueprint … a path to huge innovation in the broader application space,” Holmes said.
These three very different products symbolize how the plant will have to operate once it comes online. Executives described it as a “high variety, low volume” foundry, meaning it will have to do a lot of different things well, but it won't make a lot of any one thing.
This is the opposite of most silicon foundries. A large silicon fab has to run many similar test wafers to catch errors. But TIE can't do that, so instead it relies on artificial intelligence developed by an Austin startup. Sandbox Semiconductor– to help predict the outcome of changes in its processes.
Along the way, the NGMM will provide a number of research opportunities. “What we have with NGMM is a very rare opportunity,” said Ted MoiseProfessor at UT Dallas and IEEE Fellow. With NGMM, universities plan to work on new thermal conductivity films, microfluidic cooling technology, understanding failure mechanisms in complex packaging, and more.
“NGMM is a strange program for DARPA,” admitted Whitney Masondirector of the agency's microsystems technology department. “We are not in the habit of raising enterprises engaged in production.”
But the city's unofficial motto is “Keep Austin Weird,” so maybe NGMM and TIE would be a perfect fit.
Articles from your site
Related articles on the Internet
