At Ayar Labs, optical chips borne of MIT research 

Ayar Labs wants to make computing more energy efficient by using light to move data between chips.

By Dylan Walsh  |  November 3, 2017


Ayar Labs is developing optical chips that could transfer data 10 times faster than anything on the market.

Why It Matters

A breakthrough in optical chips could mean massive energy and cost savings in data centers.

In 1965, Gordon Moore noticed that computer chips were roughly doubling in power every year, following an exponential curve. The trend held with such reliability that the observation was codified into Moore’s law. “But the speed at which we take data in and out of chips has instead followed linear growth,” said Alexandra Wright-Gladstein, a 2015 graduate of MIT Sloan. This expanding gap between the speed of individual computer chips and the speed with which they can communicate creates a bottleneck. “Right now, processors in big data centers spend the majority of their time idling, simply waiting for data to come or go.”

Wright-Gladstein enrolled in business school seeking a solution to this drag on economic and environmental efficiency. She graduated as the cofounder and CEO of Ayar Labs, a company born around the use of light, instead of electricity, to transfer data between chips. The technology was invented by Rajeev Ram in MIT’s Research Laboratory of Electronics, with colleagues at Berkeley and University of Colorado Boulder. Ayar Labs expects its first product to be 10 times faster than anything currently on the market.

For more than a decade, big companies like Intel and IBM have been chasing an optical technology for this kind of data transfer. Their efforts, however, always assumed that such a technology required tinkering with the composition of silicon: pure silicon could be used for computer processors, but some admixture was needed for the optical device. This entailed building two separate chips.

“But maybe because we were a small research group across a few universities and we didn’t have the resources that Intel had we ended up questioning this assumption,” said Wright-Gladstein. “I think most of the industry figured it was a stupid idea, but we tried anyway, and after a decade of research we’ve created a new class of optical chips.”

Though a few years out from its first product, Ayar Labs raised $2.5 million in its first round of funding and has attracted interest from a range of large companies that provide supercomputing or data center networking services. The challenge of recruiting these customers was somewhat unconventional: “We had to get customers to think about what their tech can do in a different way,” said Wright-Gladstein. For decades, industry has watched input-output capacity double every 3-10 years. In that sense, Ayar is not simply introducing a new product, but a new frame of reference — a device that has potential to increase speed exponentially instead of linearly. “We’ve had to find the individuals within big organizations who are excited about thinking this way, about thinking differently,” Wright-Gladstein said. “Once we do find those people it’s really fun.”

But what excites Wright-Gladstein most is the company’s unforeseeable future. “I like to draw a comparison to the inductor in the radio,” she said. It was common knowledge in the 1990s that inductors should not be made of silicon. When a professor at Berkeley did make one of silicon, he laid the groundwork for today’s cellphones. “We’ve created a new platform for manufacturing optics and electronics together, for the first time, on the same chip,” she said. “In the same way that the first person to make an inductor in silicon likely had no idea he was enabling the wireless revolution, we can’t imagine today what this technology will enable in the future.”