BY ALAN BOYLE on April 16, 2019 at 1:32 pm
It’s taken 20 years, but executives at Seattle-based Nonlinear Materials Corp. are finally putting the pieces in place for what they say could be a revolution in electro-optical processing.
“Everything in tech is about timing,” said Nonlinear Materials CEO Gerard Zytnicki, a Microsoft veteran who’s served as a consultant for a wide range of tech ventures. “And we think that from all perspectives, the timing is right for this technology to basically take off.”
NLM’s technology aims to turbocharge chip processing speeds by taking advantage of optical computing, which manipulates photons of light rather than electrons. That, in turn, could open up new frontiers for a field in which progress seems to be slowing down.
The classic formulation to describe that progress is Moore’s Law — the observation that processing speed tends to double over the course of two years or so. That doubling curve is now leveling out, due to the physical constraints of electronic chips.
“Moore’s Law is not dying, it’s actually dead,” Zytnicki told GeekWire.
He and other NLM executives say switching from electrons to photons would change the equation.
“When you look at the history of the computer business, it has been driven by big jumps in speed of processors, which enable next generations of applications. Great companies have been created when those big jumps have occurred,” said NLM Chairman and President Paul Nye, who has 35 years of experience with technology startups.
In the past, great companies such as Apple, Microsoft and Amazon have all capitalized on the upside of Moore’s Law.
“Now that Moore’s Law has died, the only option is optics,” Nye argued. “People have been waiting for years for optics to make sense. It hasn’t made sense because the materials haven’t been there. But now they are.”
Nye said he expects the computer-chip marketplace to shift rapidly to optics over the next five years.
We’ve heard that before: Back in 2000, researchers at the University of Washington unveiled an “opto-chip” that they said could come into wide commercial use within five years. They expected the chip to speed up processing times by more than an order of magnitude, into the range of hundreds of gigahertz (compared with today’s best electronic performance of around 8.8 gigahertz).
The researchers assumed that they’d be able to shrink down the optical circuitry to mesh with electronics and create smoothly working electro-optical hybrid devices. Unfortunately, it didn’t work out that way.
“Performance improved rapidly over the first few years, and hit a wall around 2007,” said Lewis Johnson, NLM’s chief scientific officer and a research scientist at UW’s Department of Chemistry. “It took a number of years for people to figure out how to integrate even the second-generation materials onto small components on a chip.”
Now NLM and its research partners at UW and other institutions are seeing the light at the end of the plasmonic tunnel. Over the past couple of years, UW researchers have reported a series of advances in the development of electro-optic modulators that can transform electronic signals into optical signals with low signal loss. At the same time, the materials used in optical chips have been improving.
Working in league with UW’s CoMotion innovation hub, researchers like Johnson and electro-optic technology pioneers Larry Dalton and Bruce Robinson joined forces with tech veterans like Zytnicki and Nye to incorporate Nonlinear Materials last year.
NLM operated in stealth mode until last month, when it announced an exclusive deal with UW to license key patents relating to electro-optical materials. Johnson said advances in materials science have boosted the theoretical capabilities for optical computing well beyond what was predicted a couple of decades ago.
“The material itself is capable of potentially 10 to 15 terahertz,” he said. “If anything, the biggest limiting factors with speed are the drive electronics, not the optical components.”
Nye said NLM aims to sell the materials for optical processing to device manufacturers. “We want to be able to show people how to make devices, and in some cases joint-venture with them going into some of these markets,” he said. Johnson said the model would be similar to the way Microsoft built up a wider software ecosystem, or the way ARM created a hardware ecosystem.
Toward that end, NLM has a pilot fabrication facility on the UW campus and is working on a product development kit, or PDK. The company is about halfway through a $1.25 million seed funding round, “mostly with local investors, angels and those kinds of people,” Zytnicki said.
Even though Nye is giving out the standard five-year prediction for commercializing the technology, neither he nor anyone else at NLM expects the rollout to come all at once. Zytnicki said optical computing is more likely to creep slowly into the marketplace, perhaps starting with internet trunk networks, network hardware for data centers and electro-optical connections embedded in computer chips.
Zytnicki said optical computing will eventually find its way into telecommunications, cloud computing and healthcare data processing, as well as military and aerospace applications. But he acknowledged that it’s likely to take significantly more than five years to get to that point.
So what will be the “aha moment” for the optical revolution?
“These are all aha moments, right?” Zytnicki said. “Our first aha moment was, ‘Hey, we signed with the UW.’ The second aha moment was, ‘Hey, we raised half the money we said we were going to raise.’ … The next aha moment is going to be, well, obviously, finishing the round, that’s a big one. Then it’ll be our first contract.”
Meanwhile, Johnson said he and other researchers are preparing for the next set of technical aha moments — on a time frame that’s much shorter than 20 years.
“It’s all happening at once,” he said.