Member Profile: Etched

Gavin Uberti, CEO, & Chris Zhu, CTO, co-founders of Etched 

In March 2023, Uberti and Zhu were still taking math and computer science classes at Harvard. A month later, they closed $5.36 million in a seed round led by Primary Venture Partners, dropped out of school, and moved to the Bay Area to make a more cost-effective chip for generative AI. They share lessons learned about running a company over the past few weeks and how their Private Office at CANOPY Menlo Park is positioning them for early-stage success. 

What was it like to establish an AI company when you were still in school?

We were formally incorporated last October. We were both studying math and computer science, taking classes on how to operate models at scale and how to network computers efficiently. I did more reading about the field on my own and got hooked. Etched was a dorm room project; I was on campus and Zoom calling every other day with Gavin, who was finishing up a gap year before we took things a little more seriously over winter break. After working together in person in the spring, we decided to drop out.

Was leaving Harvard a hard decision to make?

It was a little tough. Harvard’s a nice place, and I really enjoyed getting to know the people there and taking classes, but the choice boiled down to realizing that there’s no way I could double dip, going to school and working on Etched. I figured this would be one of those things I would regret so hard if I didn’t try it–whether or not it succeeds is another question!–but I felt that being able to work on an AI startup when AI is booming was such a valuable opportunity that I wanted to commit to it.

Things have moved quickly: The WSJ reported your seed round valued Etched at $34 million. Was it hard to make the move from academia to enterprise?

We’re designing hardware, and if you can make this computer chip for AI models, the product-market fit is pretty obvious. One of the great things is that, on the business side, we didn’t have to struggle because we had well-defined goals for what needed to be done and how customers would interact with the product. It was actually a natural progression from academic study to a more fleshed-out real-world startup. Near the end of last year, we wrote a paper and realized that it was feasible to offer 140 times the performance per dollar of traditional graphics processing units used in running generative AI models. We decided to do something bigger with that insight. 

Why did you decide to move from Massachusetts to the Bay Area?

Long story short, it’s because AI is happening here–if you look anywhere around the world for research and development and new companies working around AI, they’re all in the Bay Area. Being in this community and staying in the know about what’s going on is really valuable. We’re also a chip company, and there are good semiconductor engineers here, so the choice was pretty natural.

It’s crunch time for many AI companies as the cost of computing is skyrocketing. What is the roadmap for Etched to bring its product to market in Q3 2024, and what are your long-term milestones?

Machines to fabricate chips cost several million dollars, so most companies in our space create the design and send them to a company like TSMC in Taiwan. We’re currently talking with those fabs and establishing relationships, but still very early in this stage, as it’ll probably be more of a driving issue six months down the line. Long-term, while we’re just making one chip that can run AI language models that generate text right now–which we’ve nicknamed Sohu–the entire company shouldn’t be based on a single product. Many new technologies exist, including image and video generation and protein-folding simulations. Our five to ten-year vision is to make other chips for different kinds of AI models.

What are you most excited about within your field?

These AI models are getting bigger–over the past couple of years, we’ve seen that the sizes of these models have been increasing by 10x, and the capabilities of these models are directly correlated. GPT-4 is much better at various tasks than its predecessor, GPT-3, and people are experimenting with their own models. We believe that the end goal is improving people’s quality of life as AI can replace many mundane tasks and accelerate different workflows. Let’s say you’re a programmer and have a question about how a particular piece of code works. Previously, you had to spend 15 minutes Google searching to find discussions in online forums. Now, you can just paste in your code and ask an AI language model a question, and it’ll respond within seconds. That reduction in turnaround time allows people to work so much faster and be more productive. We can leave all the grunt work to the AI, and humans will do the creative thinking. Being able to have that partnership is what makes these tools so valuable.

You arrived in the Bay Area around two months ago. How is your Private Office membership at CANOPY Menlo Park supporting you in developing your company? 

It’s been really great! We love the peaceful environment and soundproofing, the beautiful views. We looked into many options, including empty office spaces, and determined that setting up an office or hiring someone to do it for us wasn’t the best use of our time and resources. A co-working space was our best bet, and CANOPY clinched it because the building is well-finished, we have a huge office with eight desks–currently, there’s just four of us–and you can walk to lots of great restaurants, which is really convenient. 

Will every member of Etched’s staff work out of CANOPY? 

We’re building an in-person company, and we want everyone to be in the office daily. Seeing people in person and the small day-to-day interactions are crucial to establishing a strong company culture. To that end, we’d like to hire people from around the Bay Area so they can commute to the office. As we scale up, one option could be to get another room here or take a larger space at CANOPY Menlo Park. It’s great to have that flexibility.

What does your nine-to-five look like? 

I wish it were nine to five! Kidding. Typically we come in in the morning and stay in the office until seven, eight, or even nine p.m. People often ask me why I’m working so late; as cheesy as it is to say, I’m enjoying what I’m doing. After a college schedule, where you are working until 1 a.m. with your friends on a Friday night, this is a more regimented schedule. Regarding the actual day-to-day, every person at Etched wears many different hats. We might get on a hiring call in the morning, then manage some accounting before doing some technical work and taking a call with a vendor or supplier.

Do you have a favorite podcast? 

One of my favorite podcasts is by Lex Fridman, a well-known CS professor and researcher at MIT. He hosts a wide breadth of interviews with many different people: Elon Musk, Zuckerberg, famous YouTube folks, and influencers.

Do you travel much?

One destination I never get tired of traveling to is Franconia Notch in the White Mountains in New Hampshire. It’s one of my favorite spots for hiking–I’m a huge fan of hiking and backpacking. I’ve visited at least a couple dozen times in my life, and it never gets old.

What’s your favorite thing to do so far in the Bay Area?

Along the lines of hiking and backpacking, I’ve heard there are lots of great trails, and I’m looking forward to exploring them–I did a quick hiking trip last weekend. I’m also a bit of a foodie, and there are lots of great restaurants out here that I’ve heard about, especially in SF, so I’m looking forward to trying them.

Is there something you’ve discovered or learned in the past month of being here that you’d like to share?

How to run a company is one thing we’re continuing to learn about. A lot gets done by taking calls with people, which is not something I internalized before–when I was a student taking classes, I used to joke about people having fully booked out calendars. Now we’re here, and scheduling and taking meetings is such a big thing.