200. That's One Cell of a Machine: Jason Kelly of Gingko Automates Labs to Accelerate Discovery

It would be like saying the only use of a computer is, I don't know, computational drug discovery or something, right? Like, we obviously understand that it's a machine that could do a million other things too. A cell is a machine that could do a million other things.

Yeah, finally made it. I'm so excited.

The mission of Ginkgo from the beginning is to make biology easier to engineer. And the root of that is just a fundamental truth about biology which is it's the other sort of like coded substrate we have other than computers right so a computer runs on digital code forms of zeros and ones that's what's going on down at the bottom you have all these layers of abstraction on top and we get the whole world of software and all the variability the magic of like installing new code and making that machine do something new and we only really have one other substrate like that which is a cell. And the code in this case is not zeros and ones. It's A's, T's, C's and G's of DNA. But it is coded. And if you install new code, which is what genetic engineering and biotechnology fundamentally is starting in 1978 when we moved the first gene in this case for human insulin into E. coli to start Genentech. Since then we've been able to install new code into cells. And the big question is are we able to do it well? Like can we make the cell do what we want? Can we make biology do things that are useful for humans? And that's like my life's work. That's Ginkgo's whole mission is what's the tooling that makes that easier year-over-year. And there's different ways to attack that technology problem. We can talk about them. Ginkgo has tried a lot. You know, we spent $2 billion over the last 10 years, uh, you know, since doing Y Combinator back in 2014. We've done a lot on laboratory robotics, which we can talk about. We've done a lot on design tools for genetics for designing genomes and pathways. Uh a lot of lessons learned across those two things. But fundamentally that's what we've been trying to do. And the thing I try to explain to people because we spend a lot of time in pharma now. The majority of our customers are these days in the pharmaceutical industry is that that isn't the only use of a cell.

It would be like saying the only use of a computer is I don't know computational drug discovery or something, right? Like we obviously understand that it's a machine that could do a million other things too. A cell is a machine that could do a million other things. And so we have pushed that at Ginkgo a lot over the years. We've done projects with Bayer Crop Science for engineering microbes to produce fertilizer. We did a lot of work in sort of the heyday of animal free meat with a lot of those startup companies there. Cosmetics, nutritional ingredients. We tried a lot of these things. A lot of them have not worked that well to be honest commercially which we could talk about too. And so a lot of our customers in those sectors don't exist anymore. Or if they're the startup companies or the large companies, they have programs, but nothing like what you see in pharmaceuticals. So, I think we're in this interesting spot where it is still the case that the main app for biotech is drugs. But the underlying technology very obviously can do many more things than make drugs. And that is the thing we have to unlock. And the thing that unlocks that is better tooling. It's as simple as that.

You're so close to biology that you don't think of it like a tech. But like just imagine that uh Tim Cook gets on stage and he's like, "Oh, Apple has a new product. We've invented the seed." Right? You put this thing into dirt, you cover it up, you add air, water, and sunlight. And here's what it's going to do. It's going to start pulling carbon atoms out of the air. It's going to self assemble solar panels to collect additional energy to power more carbon capture. It will continue to then do molecular assembly at a level of atomic resolution like Intel and check it out. And we would be like this is the most substantial technology breakthrough in the history of mankind, right? Like very obviously, right? Like what's weird about biology is naturally existing biology is the existence proof that this substrate is capable of absolute magic. Self-repair, self assembly, creating its own energy, general intelligence, hello. Like what the hell are we dealing with here, right? It is magic. But unlike every other engineering field, everything, mechanical, chemical, computer science, electrical engineering, humans did not invent biology. So we have this totally bizarre situation that looks nothing like any other engineering field where you've got absolute magic and capabilities, but a substrate that you do not understand. And God bless, we do have the ability to like read and write code from this thing, but you're programming a computer you did not design. And so that's it. And it's awesome. Like I think if you're a young person today, if you like tech, like biology is the place to go. It is where there's magic, right? Like we still have magic there. And it's in the physical world, which by the way, as you guys talk about on this podcast a lot, these are the big markets. Like information tech is weird. It kind of collects a lot of of the world inside of it. But if you really think about what impacts people, it's food, it's medicine, it's building materials, it's fuels, it's like all the atoms. And the the cell is the computer of atoms. It is.

Yeah. Yeah. 100%. If you're a longevity bro, this is also the place you need to be hanging out. Yes, like [laughter] let me tell you like the AI model is not the thing that's going to make you live forever. It's something in the world of biotech, right? We are made of biology. Like it is figuring this out. That's it. There's no other way.

2008, right? Back 2008. Yeah. I think the thing that and I'll get back to your question about the you know the history there but we have not gone through a tool revolution like say computer science or robotics has over the last 20 years like if you go in the computer science department grad students at MIT and you look at their tooling compared to when I was in grad school the power of the computers the infrastructure completely different and like certain areas of biology are better our sequencing is cheaper you know synthesis is a little cheaper but most of it's the same. And and that's a problem, right? Like my conclusion now 20 years in is that's the root problem. The root problem is we have not gone hard enough on our tools in biotech. But what we started Ginkgo very much with a tools mindset from the beginning like we were kind of tool nerds coming out of MIT. We were not like therapeutics people at all, right? We had been like synthetic biology led by Tom Knight and Drew Endy and we started the company with Tom on this idea that maybe you could standardize genetic components and use simple organisms to make biology more predictable. Okay, that's hard by the way that did not work as well as I would have liked to be honest and also that you could potentially standardize the lab work. That has worked great. So those are kind of the two things Ginkgo worked on since founding the company since particularly the last 10 to 12 years where we had capital was in those two lanes. Could you simplify organisms to make them more predictable? And Drew's lab just actually put out this like spud cell, which I think is a great idea. Like that's like the extreme form of simplification. So there was like that approach to making the organism more simple. And then the other thing we worked on was could you make the lab work cheaper with scale? Because if you think of it like a computer scientist, when we compile and debug DNA code, that's a physical process. We actually build the DNA with chemical synthesis. We install it in organisms with cloning. We grow the organisms up in a lab and then we test them in high-end analytical equipment. It's atoms. And so if you want that debug cycle, that compiling and debugging of code to be rapid, you got to automate all that stuff. And so that's been the journey of Ginkgo for the last 10 years on the lab side has been that. And then on the other side, we tried a lot on sort of standardizing and that was tough. We couldn't simplify the cells enough to make them easy to engineer. That's where we I've kind of landed on that. And there's a new generation over there with like virtual cells and things like that. If we generated enough data, maybe you don't need to simplify the cell. You can actually manage it in all its complexity. And so that's sort of the new lane for the people that are trying to simulate the cell. And in that world, I'm happy to have other people do that. We just want to make the data for them with the robotics because it is still, I'll be honest, it's still hard to start companies like this right out of grad school. It was hard for us in 2008. Financial crisis fine, but we weren't making a drug, which means you make no sense to biotech VCs. And we had a lab, so we made no sense to tech VCs. Oh, and we were like kids, which in tech people like, and in biotech they don't.

I'm now uh what, 45? I'm still like a young biotech CEO.

18 years. Like like what in what world, right? Like it's so different than tech, right? That that said, I'd say that the two big pieces of advice for like grad students or um deep tech folks tuning in is number one like don't be afraid of grants. So actually the first like seed money into Ginkgo, the first three million or four, five million bucks we brought in over five years was grants, RPA, NSF, SBIR, DARPA. Like that was great. It was non-dilutive. You could, you know, how to apply for it as a grad student. Everyone's going to tell you that's lame and slow. Just do it. Okay. And so that's one thing. And then two, once you raise venture capital, if you are anywhere that is not San Francisco, go to San Francisco to raise money and then fly home. Fly home. All my labs are in Boston. It's great here. It's like the labor market. There's a lot of great things about being in Boston. The only thing that's bad about being Boston is fundraising. And And so just money travels very easily in airplanes. [laughter] Okay. Right. Like go raise your money in California. Like it's just no-brainer. And so we did we were the first biotech to do Y Combinator in 2014 for that reason. We were sort of like biotech VC rejects, right? Like like we made no sense on all axes to our local venture ecosystem. And that's true in Boston. It's definitely true if you're in Wisconsin or you're in Ohio or you're in Indianapolis. There's people starting companies all over the place. They should just go to San Francisco to raise startup capital. Don't go to your local VC ecosystem. Those are my two pieces of advice. That that's kind of like what I would say for the first five years of Ginkgo. That that's the lesson. Other than that, like don't have a high burn rate, personal burn rate.

For VC, it's fine, but it's not nearly as good as San Francisco. Because it's like, think of it this way. When you take your company public, you do a road show. All right. Which means you meet a whole bunch of investors in a very short period of time. Do you know where 90% of them are?

New York City. Because the industry of New York City is public finance. Like on the equity side, that's their industry. And so that's just where they all are. So you get on a plane and you go to New York City and there's maybe a few other places that you also meet, but the majority are in New York. The industry of San Francisco is seed capital. Okay, there of course there are venture capitalists, but the industry of San Francisco on the finance side is seed and venture. Just like when you go to your IPO road show, you go to New York. When you do your seed road show, go to San Francisco. That complicated, guys. And by the way, venture, they do not care where you put your company. They're like in Tahoe half the time anyway. They don't care. They don't care that it's in San Francisco. If you're doing software, I think there can be reasons. But if you're doing biotech or other deep tech things, there's no reason to be there. That's a misunderstanding for a lot of entrepreneurs.

Yeah, this is a really good question. And so I'd like to say it's just purely my entrepreneur intuition, but some of it was beaten into me by like the the business realities of the markets we were in. So what happened to us basically was we spent 10 years growing on a business model where we were basically doing outsourced R&D for customers and then taking like a royalty. Okay. They actually worked pretty good like in the sense that like we had 250 deals like that over 10 years. Yeah. A lot actually. A bunch were with startups like I was mentioning the animal free meat and things like that but a bunch were with like giant you know and the chemical industry work with Sumitomo flavors and fragrances with Givaudan and IFF like we worked with Bayer and like it was like all all like big a big chemical and then right before the IPO we started to just stack pharma companies. We did Biogen then we did Novo Nordisk then we did Pfizer for RNA. It was like bang bang bang bang bang and and and so the issue was uh 22 comes along and interest rates go up all of the weird biotech companies so like all your the stuff you guys like flavors fragrance food all those things they got wiped right like SynBio that year was sad Carl if you remember and so it was tough because like the way we do it in the US anyway is we make our bets on longer-dated stuff when interest rates are low like we're not China we don't have like central planning, but we do centrally plan our interest rates relative to inflation. And so when we were in this ZERP era, the 0% interest, that's when SpaceX got started and Tesla got started. These like crazy long-dated things could play because money wasn't getting more expensive. Like it didn't matter what year you got your revenue, you're getting 0% interest on money anyway. And so people were more patient. All right. When that went away, all the patience for like non-pharma biotech kind of went away. So I instantaneously lost a bunch of customers that like were these startup companies that just were not going to get another round. All right. And then on the pharma side there was just like a post-COVID depression. But like when when that dropped a lot of the pharma companies kind of put their arms around their own R&D orgs and and were like listen we're going to cut back on external partnering and sort of protect our people because we're all belt tightening for the next couple years. And so I lived on external R&D partnerships. So that also kind of evaporated. So I had a hard time bringing in additional R&D fees and then but the real kicker is biotech products take a long time to go to market and so I was kind of pushing an app store model which I was hoping would work outside of pharma because things could go faster right like pharma is brutal right pharma is a seven-year journey right like we did a project with like Cronos that Canadian cannabis company that went to market they scaled up and started making stuff putting it in their gummies for cannabinoids and I got a something like a $50 million check right like that is awesome. If every deal had looked like that, then the business model would have worked well. It didn't work well enough when all those fast things went away and then I was just left with slow pharma. And so the R&D, I think, loop, it's real tough. Very few companies have been able to pull off that business model. Like the only one that's even closed is Atom. And they mostly did it by like staying low spend. And I wasn't low spend. And so I was like, "All right, we're going to have to adjust." This was like in 22 23. So we did a big layoff. We had half the company, cut costs massively. And I basically shifted my business model to we still have those types of research partnerships. I still work with Novo. I still work with Bayer, a few others, but increasingly the business model is basically Thermo Fisher. So life science tools, we sell our robotic hardware and we sell what you called lab data as a service, Carl, but like our CRO services. Like we will just generate data for you. No royalty, no nothing. Pay by the data point. We actually call it data points is our service. And so that's how it works. And so that's been great actually. So like from a business model standpoint, I think that's a good fit for Ginkgo today. I think it might have been harder 10 years ago when we were a smaller company to do that business model, but fine. Then we could talk about the technology side, which is what I'm most excited about because as you mentioned, we went to these kind of bet on lights out total autonomous labs and that has been awesome. But the business like if you want to talk the business side of things if you think of Ginkgo today like if you talked to us five years ago some company listening it would have been about royalties and research partnerships and now it's like think of us like Thermo if you like what we're doing in AI like a project with OpenAI or or what we're doing in robotics you can have that installed in your lab just like Thermo would install it for you and that that's the business model today.

Yeah, [laughter] this is a very good question. Okay. Um, do you want me to give you like the super nerd out on on autonomous labs?

Okay. So, this is this is I think getting very hot right now because in Silicon Valley general purpose robotics is getting hot. So, like robotic hands and

I think it's 20,000 bucks for the first one. It's like not crazy. Is it

Yeah, remember the Jetsons? Yeah, 100%. Yeah, that's coming. Definitely. So, okay, I'll give you my analogy for like what is when I say an autonomous lab, what am I talking about? I'll draw an analogy to the transportation industry. I think it's important because it's not like we haven't had robots and labs for a long time. We actually have. And so, I need to put it in context. All right. So transportation industry imagine the y-axis is the amount of automation in the transportation technology and the x-axis is like the level of variability a customer can ask from the technology. All right. So, low variability, high automation in transportation. That's the subway. You sit down. We have had automated transportation for a hundred years, Carl. We have. You sit there, you do nothing, and you are transported away on the subway. Okay? Right? But you better not expect it to go to your house. Okay? It only goes to these five stops in front of you. Not variable. Low amount of automation, high amount of variability. We have that too. It's called the car. And we put our hands on the wheel and our foot on the pedals and we drive it right into our driveway or to the grocery store or the restaurant or wherever we want to go. It's totally variable, but it's no longer automated. There's a human in the loop handling all the variability. Does that make sense? That and that's the transportation system for the last hundred years. Have you been in a Waymo?

Okay. So, so it's amazing experience. You should definitely try a Waymo. It is the automation of a subway. In other words, you sit in the back seat like you were sitting in a subway car, but it goes right to your house. Okay? So, you're getting the variability of the car, but the automation of the subway. And that's so surprising. You'll notice we don't call it an automated car because that's like an automated door or something. You're like, it must just go do the same thing over and over. Automation just repeats. Automation doesn't do variability, right? And so the we call it an autonomous car. We gave it a new name. And by the way, it's actually kind of a misnomer, right? Like because you think autonomous, it's thinking for you. It by the way, you are telling it where to go. It's not like the Waymo decides where you're going to go today. You tell it where to go. All it's we've done is we have managed the variability of driving so that you no longer need a driver for this transportation technology. Does that make sense?

Or not though. Like you could control that too, but like sure it could, right? Okay. Now, here here's the other thing though. Miles traveled,