For his tenth birthday, Tanmay Gupta’s (BS ’24) parents bought him a small telescope. “All you could really see through it was the Moon,” he says. “But you bet I was looking at the moon every night.” It was an unexpectedly prescient gift. Gupta’s obsession with space led him to Caltech and the Jet Propulsion Laboratory. Today, he is working at the intersection of space and optics, as the study of light rapidly advances from the basic system of glass lenses that brought the moon within his reach when he was a child to metalenses—ultra-thin surfaces etched with patterns that focus light—and photonic computing. “There are many problems with working in space that photonic computing solves,” says Gupta, who is in the stealth phase of launching a space-photonics startup as part of Entrepreneurs First, the AI and deep tech incubator. “I have no doubt photonics is the future.”
Let’s start simple: What is photonics?
Photonics is the new study of light. Optics—which a lot of people are more familiar with—is the old study of light. It’s our glass lenses, our camera lenses, things like that. Photonics is the next thing. It’s fiber optic Wi-Fi, sending data via light instead of old copper cables. It’s the metalens in the new iPad Pro that is nanometers thick and does the same job as the previous physical lens in the Face ID module. Cool stuff like that. The question I’ve been diving into is, how can we use light to do computation? There are a lot of advantages to using photons—that’s why it’s called photonics—instead of electrons or electronics to do math.
What are some of the advantages of photonic computing?
The idea of using light to do computing is not entirely new. There are a couple of interesting papers from NASA about it in the 1970s, but the world moved toward digital computing instead. Now we’re hitting a bottleneck with that approach. We’ve reached the end of Moore’s Law, and doing things like matrix multiplication—the math behind AI training—is very energy-intensive and slow. So we’re seeing a renaissance in optical—or photonic—computing. When you use light, there are no scaling energy costs and no resistive heating. Light also provides lower latency, so you can move more data around more quickly.
How does photonics fit into the future of space technology?
A lot of new satellites are using optical communications links because they are faster, have higher bandwidths, and are more secure, so it makes sense to use optical computing, too, to avoid the conversion between the electronic and optical domains. Any conversions increase latency. Small, low-Earth-orbit (LEO) constellation satellites are also limited in their compute power because powerful traditional electronic chips need too much energy and create too much heat. To power the chips and dissipate that heat into space, companies would have to make the satellites much larger. That’s expensive. Consider Earth observation satellites collecting terabytes of data. That is too much data to send down to Earth quickly. Companies want to process the data onboard in real time and send down the insights: there’s a fire in the Amazon at these coordinates or this ship is illegally moving from North Korea to China in violation of international sanctions. Photonics can help do that.
What does the computing field look like in 10 years?
There’s still a lot of research and development to be done in photonics, but it’s finally starting to move into commercialization. Over the next 10 years, I think the world will move toward using photonics in many applications. But that doesn’t mean the end of electronic computing. I see both of these technologies running in parallel, in different use cases.
For a long time we’ve moved toward convenience in technology—one tool that can do everything—but we’ve sacrificed performance. Now we’re demanding performance we can’t get from convenience, so I think we are moving toward a world where things are specialized again.