Ayar Labs Gets $2.5M to Go After Optical Chips for Data Centers

Xconomy Boston — 

This startup sits at the intersection of electronics, optics, energy efficiency, and the Internet. It is where more than a decade of academic research meets a multibillion-dollar market. This is the story of Ayar Labs.

The San Francisco startup, which has roots in Cambridge, MA, and Boulder, CO, is trying to solve a fundamental problem in computer systems, starting with data centers—namely, how to make them perform better while also using less energy. Now, the company has raised seed funding from prominent investors (more on that below) as it gears up to build its first product. And its experience sheds light on how a team of diverse experts can take a broad scientific advance and focus it on a particular market.

The story begins in a classroom at MIT in 2014. Alex Wright-Gladstein was getting her MBA at the MIT Sloan School of Management. She had previously worked in energy efficiency and smart-grid systems at EnerNOC and other companies. She says her “passion for the global warming problem” drove her to look for emerging technologies that she could help commercialize to address some of the world’s big challenges.

Wright-Gladstein took a course called Energy Ventures (class number 15.366), where teams work together on energy-related projects, business plans, and startup ideas. She had previously sat down with about 20 professors to talk about energy research projects that might be ripe for commercialization, she says.

One technology stood out. It was co-developed by Rajeev Ram, a professor in MIT’s Research Laboratory of Electronics. Ram and his collaborators at the University of California, Berkeley, and the University of Colorado Boulder had spent about 10 years developing methods for combining high-performance optics and electronics on the same silicon chip. The result: a new kind of chip-to-chip communication using light, but without the need for a custom manufacturing process. (That research was published in Nature last December.)

Wright-Gladstein and others saw a big opportunity in data centers: the potential to boost speed and performance while also saving energy. It boils down to the fact that light can be used to transmit information faster and more efficiently than electrical wires. Yet optical systems have been difficult to integrate with silicon chips, for technical reasons. So, places where lots of computers need to talk to each other fast—see today’s ever-expanding data centers—have inherent bottlenecks. Data centers already use some optics, but they also use (and waste) tons of energy.

To explore possible solutions, Wright-Gladstein and her class team got to know the student researchers on the chip project: Chen Sun, a PhD student at MIT, and Mark Wade, a PhD student at University of Colorado Boulder. They worked out a preliminary business plan and product strategy. In early 2015, Wright-Gladstein convinced Sun and Wade to enter the MIT Clean Energy Prize competition with her. “I said, ‘I’ll do all the work.’ They were busy trying to graduate,” she jokes.

Their team ended up winning two grand prizes that spring, netting $275,000. The money “made it possible to ignore job offers” and focus on the company, says Wright-Gladstein, who is the CEO (pictured above, center). In the meantime, her co-founders Sun (electronics expert, on left) and Wade (optics guru, on right) finished their PhDs and took on the roles of CTO and chief scientist, respectively, at the company.

Prior to the competition, they had named their startup OptiBit. Later they realized the name was already taken by a company in Europe, so they had to change it. They settled on a new name: Ayar Labs, with Ayar pronounced like “IR” for infrared—the part of the electromagnetic spectrum their optical system uses.

The team moved out to San Francisco last summer (much to the chagrin of those worried about brain drain in Boston). They joined two incubator programs, Citris Foundry and Silicon Catalyst, and continued work on their technology and business plan.

Ayar Labs’ scientific co-founders—the company calls them “technical principals”—are MIT’s Ram, Vladimir Stojanovic at UC Berkeley (he was previously a professor at MIT), and Milos Popovic from CU Boulder. (Popovic has just moved to Boston University, so the company has strong ties to the area.)

In late May, Ayar Labs closed a seed financing round of $2.5 million, led by FF Science, which is part of Founders Fund, the venture firm co-led by Peter Thiel. Ayar Labs’ other investors include TechU Angels, led by MIT Sloan alumnus Natanel Barookhian.

Ayar Labs' technology miniaturizes optical and electrical components onto a silicon chip.

Ayar Labs’ technology miniaturizes optical and electrical components onto a silicon chip.

The startup’s first product is slated to be a 400 gigabit per second transceiver, aimed at data centers. The team plans to make it available in 2019, which is when the first data center switches with 400 Gbps ports should appear. (Right now the highest per-port bandwidth available is 100 Gbps.) Ayar Labs is planning a demo of the technology for next year, which will help the company work with data centers and optical switch vendors, Wright-Gladstein says, and land some early customers.

The transceiver will contain a high-performance silicon chip that can send and receive data both optically and electrically, as well as a laser module that supplies light to the chip. It’s different from the team’s Nature demo, which showed optical communication between a microprocessor and memory system. The key to Ayar Labs’ planned product is that the manufacturing process uses standard chip-foundry techniques. (The company has used a facility in New York state.)

But it’s still super early days. Ayar Labs currently has 12 people, including PhD interns. And competition will be fierce. Giants such as Intel and IBM have had huge research efforts in the area, Wright-Gladstein acknowledges, as have some established private companies like Luxtera. There has also been consolidation in the sector, with Aurrion being acquired by Juniper Networks, and Lightwire getting snapped up by Cisco.

What’s more, Internet giants like Facebook, Google, and Amazon are taking hardware matters into their own hands as their data centers deal with exploding demand. It wouldn’t be much of a stretch to see them developing their own optical systems to improve bandwidth. Because of the increasing use of big-data analytics and machine learning, Wright-Gladstein says, enterprises “are just sending more and more data between servers in a data center.” She adds that optical equipment vendors and tech companies are “trying to get these key data centers as their customers.”

It all adds up to an immense opportunity for Ayar Labs and its competitors. Indeed, the field of “silicon photonics”—which brings the advantages of silicon chips to optical devices and networks—has made big strides in the past decade. And its commercial growth seems fastest in the data center market. An example is Acacia Communications, an optical networking company based in the Boston area—and the region’s only tech IPO of 2016. Acacia (NASDAQ: ACIA) uses silicon photonics to make faster and more efficient communication equipment for telecom and Web service providers, in particular to connect between data centers.

Bill Aulet, head of the Martin Trust Center for MIT Entrepreneurship, knows Ayar Labs from its early days, as he is involved with the Energy Ventures class and the Clean Energy Prize. “It is a company with extraordinary potential but some real challenges in the nature of the business, the chip business,” he says. “This is a capital-intensive, long-lead-time business. But they have a strong team and a breakthrough technology and understand that this will take a while.”

For Ayar Labs, the future will come down to hard work, adaptability, and attention to details. Its core enabling technology works, Wright-Gladstein says. “Now we are focused on demonstrating packaging and laser module manufacturing approaches that scale up in volume and down in cost.”

Ultimately, the company’s technology could also be applied in areas such as supercomputers, Lidar and sensing, and healthcare. “It’s a pretty basic innovation, so you can also do a lot of things with it,” she says. “We think data center applications is the way to start.” She adds that in the company’s early days, there was “a lot of debate” over which market to go after first.

Another interesting note: Wright-Gladstein’s original thinking on energy efficiency has taken a bit of a back seat, the way most startups shift from their initial motivations. Over the past year, she says, talking with optical equipment vendors and data center managers has made her team realize that reliability, not efficiency, should be the top priority for their product.

“Energy efficiency is important, but it’s a distant second,” she says. “That made us go back to the drawing board about how we’re building our first product.”