Ginkgo Bioworks Holdings, Inc. (NYSE:DNA) Q4 2025 Earnings Call Transcript

Ginkgo Bioworks Holdings, Inc. (NYSE:DNA) Q4 2025 Earnings Call Transcript February 27, 2026

Daniel Waid Marshall: Good evening. I’m Daniel Marshall, Senior Manager of Communications and Ownership. I’m joined by Jason Kelly, our Co-Founder and CEO; and Steve Coen, our CFO. Thanks, as always, for joining us. We’re looking forward to updating you on our progress. As a reminder, during the presentation today, we will be making forward-looking statements, which involve risks and uncertainties. Please refer to our filings with the SEC to learn more about these risks and uncertainties, including our most recent 10-K. Today, in addition to updating you on the quarter results, we’re going to provide insight into the autonomous lab, how we believe it will transform biotechnology and how we plan to commercialize autonomous labs going forward.

As usual, we’ll end with a Q&A session, and I’ll take questions from analysts, investors and the public. You can submit those questions to us in advance via X, #ginkgoresults or through e-mail, investors@ginkgobioworks.com. All right. Over to you, Jason.

Jason Kelly: All right. Thanks, Daniel. So Q4 was really a breakout quarter for us in sort of defining and really leading in the category of autonomous labs. And so you’re going to hear a lot from me about that in the future. But I want to start by saying our mission remains to make biology easier to engineer. But in 2026, the technological focus for the company and really the business focus is going to drill down on investing to win in this category of autonomous labs. And this is really a part of what I see as an emerging movement around robotics and AI and autonomy that’s coming to a lot of sectors in the economy. And I think we happen to be in a sweet spot in bringing that into really a high-value area around laboratory research that there’s an increasing amount of excitement about, and I intend to win that.

All right. So how are we going to do it in ’26? So first, we want to focus our investment in our platform into that area primarily. And I’ll talk in a minute, but we mentioned in our recent earnings announcement just now that we’ll be divesting our Biosecurity business. That allows me to focus Ginkgo’s investment and our dollars really into autonomous labs and bring in other new investors to invest alongside us into Biosecurity. So that’s focused in investment. Second, internal to the company, we want to demonstrate the capabilities of our large autonomous lab here in Boston. And the way we’re going to do that is we’re going to start to systematically decommission our lab benches, our walk-up automation, our work cells, the way that we’ve traditionally done our R&D services over the last 10 years and move more and more of that work on to a single large autonomous lab that’s software controlled here in Boston.

And the reason I want to do that is that serves as a demonstration to the Mercks and the Takedas and the Pfizers and all the folks who have huge investments in traditional manual laboratories that it is possible to take open-ended research and run it through a large autonomous laboratory system. And so I think that’s really fundamentally the most important work we’re doing this year. And then finally, I want to book sales of autonomous labs. So you’re going to hear in one of our big announcements from last quarter that we did a $47 million deal with Pacific Northwest National Labs. So I want to sell autonomous labs to National labs like that DOE deal, but I also want to sell them to biopharma. I want to sell them to research universities. And so that sort of bookings and landing new deals is the other thing we want to do in 2026 in this direction.

Q&A Session

I do want to take a minute and talk about that Biosecurity divestiture. So you might remember, over the last 5 years, we’ve invested a ton of energy into this space. This really came about starting during COVID, because we honestly just saw a need. COVID was sort of a global scale biological disaster, and we felt we should lean in and help where we could. The niche that we found in that moment was doing really monitoring for — so not diagnostic testing, but rather monitoring testing in order to reopen congregate areas and in particular, reopening schools here in the United States. So I’m really proud of this is a decent sized business for us. But really importantly, we helped open 5,000 schools nationwide. And this is one of these like really political topics.

And I think what’s neat about technologies, you can sometimes find a third way between and one end at the time, which was hey, we really should be closing the school, [ standard ] for teachers. We care about spreading disease. And then on the other side, hey, this is hurting kids, and we need to open the schools and everyone should just go back and whatever comes comes. And there was a third way, which was why don’t we open the schools and have persistent monitoring so that if an outbreak starts to happen in a school, you can send 2 or 3 kids home and stop it. And that’s exactly what we want to build at a nationwide level and what’s continued after COVID in our monitoring at airports that we do in partnership with the CDC, looking for viruses in the wastewater of planes and other inputs, both here and internationally in places like Doha and Qatar, at the airport there.

And so that sort of identify it, put it out, put that fire out before it spreads is something that’s needed nationally and globally for the U.S. to be secure. The other thing that’s been happened in that period of time, you might have noticed, companies like Anduril, Palantir, Shyam Sankar, our Board Chair, is the CTO of Palantir. This sort of defense tech sector really exploded over the last 5 years. And so there’s been increasing interest from pure-play investors in the defense space who want to see next-generation sort of biodefense primes. So again, these are companies that would be focused on serving the government and others on biodefense needs directly. That’s very exciting because it means there’s lots of new capital interested in that.

But to my point earlier, where I want Ginkgo to focus very clearly in 2026 is on autonomous labs. And so one of the great things that happened was we got a lot of inbound from these types of investors, and we saw an opportunity to say, all right, why don’t we share in the upside of biosecurity by taking that business unit in the company, spinning it off, taking it private, bringing in investment from some of these great investors. Ginkgo will still hold a minority position in that. So we get a piece of the upside of what we’ve built, but the investment needed to build that biosecurity prime doesn’t need to come from the $430 million, as I’ll mention in a second that we had on our books at the end of the year. We can focus that into autonomous Labs.

So I think this is a win-win all around. I also think bringing in these types of great folks that we have coming into the private entity are really going to open doors with the defense sector and so on and having it be a sole branded biodefense company, it’s the right time. So I’m super excited about this. I think it’s — I want to give again credit to the biosecurity team at Ginkgo, who did absolutely amazing work through Ginkgo — sorry, through COVID and now has a real opportunity here, I think, to build a generational business coming up in the defense sector. Okay. Last point I want to make before I hand it to Steve. So again, I think tremendous work over the last 2 years. We sorted 2 things at the same time. We dramatically cut back spending as we saw sort of a downturn in the biotech sector and a lot of our customers pulled back on outsourced large R&D projects, which was really our bread and butter here at the company over the years.

Because of that, we drew down on our spending and pretty substantially. So in fiscal year ’24, we were at $383 million and just last year, $171 million. So a 55% reduction in our annual cash burn. That sets us up very nicely. You’re going to hear from Steve on our target for cash burn for this year, even with the investment, our focused investment in autonomous labs and moving that investment in biosecurity into a separate private entity, we’re actually able to do better than what we spent in ’25. But that, I think for investors is important to understand where we’re at from a cash position and how we’ve done a really nice job getting cash spending under control as we continue to make investments and get in the right place at the right time with autonomous labs.

All right. So I’m going to hand it to Steve to dive in a little more on the financials.

Steven Coen: Thanks, Jason. I’ll start with the Cell Engineering business. Cell Engineering revenue was $26 million in the fourth quarter of 2025, down 26% compared to the fourth quarter of 2024. In the fourth quarter of 2025, we supported a total of 109 revenue-generating programs. This represents a 4% decrease year-over-year, primarily attributed to ongoing program rationalization as part of our restructuring activities. Turning to the next slide. On a full year basis, Cell Engineering revenue was $133 million in 2025 as compared to $174 million in 2024. As previously disclosed, revenue in the first quarter of 2025 includes $7.5 million of noncash revenue from a release of deferred revenue relating to the mutual termination of the BiomEdit agreement.

In the third quarter of 2024, Cell Engineering revenue included $45 million of noncash revenue from a release of deferred revenue relating to the mutual termination of the Motif FoodWorks agreement. Excluding these impacts, Cell Engineering revenue was $125 million in 2025 and $129 million in 2024. This decrease was primarily driven by customer program rationalization related to the restructuring, as all discussed previously. The Biosecurity business generated $7 million of revenue in the fourth quarter of 2025 and $37 million of revenue in the full year 2025. It is important to note that our net loss includes a number of noncash and other nonrecurring items as detailed more fully in our financial statements. Because of these noncash and other nonrecurring items, we believe adjusted EBITDA is a more indicative measure of our profitability.

A full reconciliation between segment operating loss, adjusted EBITDA and GAAP net loss can be found in the appendix. Cell Engineering R&D expense decreased 44% from $50 million in the fourth quarter of 2024 to $28 million in the fourth quarter of 2025. For the full year 2025, Cell Engineering R&D expense decreased 42% from $272 million in 2024 to $159 million in 2025. As reported last quarter, the full year 2025 period, R&D expenses included a $21 million shortfall obligation related to our multiyear strategic cloud and AI partnership with Google Cloud. In October 2025, we amended and reset the annual commitments for future years and settled the shortfall obligation for $14 million. Resetting the commitment reduced our future minimum commitments by more than $100 million, compared to the original terms and extended the commitment term from 3 to 6 years.

Cell Engineering G&A expense decreased 40% from $20 million in the fourth quarter of 2024 to $12 million in the fourth quarter of 2025. For the full year, Cell Engineering R&D — I’m sorry, G&A expense decreased 51% from $115 million in 2024 to $56 million in 2025. These decreases were all driven by our restructuring efforts. Cell Engineering segment operating loss was $17 million in the fourth quarter of 2025 compared to a loss of $38 million in the 2024 period. For the full year 2025, Cell Engineering segment operating loss was $96 million compared to a loss of $219 million in 2024. The lower loss was directly related to our restructuring efforts, while partially impacted by the matters previously mentioned. The Biosecurity segment operating loss improved 60% in the fourth quarter of 2025 compared to the 2024 period.

And the Biosecurity segment operating loss improved 38% in the full year 2025 compared to 2024. Moving further down the page, you’ll note that total adjusted EBITDA in the fourth quarter of 2025 was negative $36 million, which was down from negative $57 million in the fourth quarter of 2024. Total adjusted EBITDA for the full year 2025 was negative $167 million, which was down from negative $293 million in 2024. Again, the period-over-period declines can be attributed to our restructuring efforts, while partially impacted by the matters previously mentioned. Turning to the next slide. We show adjusted EBITDA at the segment level to show the relative profitability of our segments. The principal differences between segment operating loss and total adjusted EBITDA relates to the carrying cost of excess lease space, which was $54 million in 2025, and this carrying cost was $15 million in Q4.

The cost represents the base rent and other charges related to leased space, which we are not occupying net of sublease income. This is a cash operating cost that is not related to driving revenue right now and can be potentially mitigated through subleasing. And finally, turning to cash burn. Cash burn in the fourth quarter of 2025 was $47 million, down from $55 million in the fourth quarter of 2024, a 15% decrease. Cash burn for the full year 2025 was $171 million, down from $383 million in 2024, a 55% decrease. Cash burn does not include the proceeds from the ATM issuances or certain cash restrictions. The significant decrease in cash burn was a direct result of the restructuring. Turning to guidance. In terms of the outlook for 2026, as Jason has mentioned and we will go into further, 2026 is about continuing to be cost efficient while investing in our AI robotics and software to bring autonomous labs to our bioscience customers, including the build-out of our frontier autonomous lab in Boston.

We have turned the page on our pure focus on restructuring actions for the last 2 years to focus this year not only on cost efficiency, but on investing in what we see as our opportunities while continuing to provide our customers the advanced services they have come to expect. We will also close our transaction for the Biosecurity business as announced and disclosed. For these reasons, in 2026, we will not be providing revenue guidance as we believe cash burn best reflects our continuing services and tools and further investments in autonomous labs. For 2026, our overall expected cash burn guidance is to be in the range of $125 million to $150 million. This range reflects a firm balance amongst cost efficiency, continuing services and tools and the further investments we are making.

In conclusion, we are pleased with the continued improvements in cash burn and cost reductions in 2025 and are excited for what will come in 2026. And with that, I’ll hand it back over to you, Jason.

Jason Kelly: Thanks, Steve. Okay. So before I jump into my section, I want to spend a minute talking a little bit more about what Steve was talking about there at the end in terms of our guidance for the year, how we’re going to be guiding on cash burn rather than on revenues and sort of why we’re doing that. And so this is in line with my theme for this earnings call, which is Ginkgo’s focus. So one thing is we want to be focusing on investing in the right things. And so we’re — I believe, again, it’s important for our investors in Ginkgo to understand what we’re doing with our cash supply, how fast that’s being spent down, what we’re spending it on. And again, the highlight here is we are spending it very deliberately on autonomous Labs, and we’re doing it in a controlled way.

We’re hopefully spending substantially less than we spent in the last year and our relative position there to our cash file, it looks pretty good. So from my standpoint, we have a solid margin of safety as we’re investing to lead in this area of autonomous labs going forward. But the second thing we need to keep focus is our attention within the company. And so the majority of our revenue today does come from our R&D services. We love serving those customers. I’m hopeful we grow those services. But as I mentioned earlier, the focus of the team in 2026 is not on hitting a short-term revenue target around a service run on top of our autonomous lab to make sure we hit a target or trying to predict exactly what that revenue is going to be over the next 12 months.

What I want their focus to be on is decommissioning all of the different labs here at Ginkgo and moving that work onto our autonomous lab so that we can show all of our customers that this works, that autonomous labs can be a true replacement for the humongous spending they have across their manual laboratories in both biotech and academic science. That’s the main event. And I felt that, again, continuing a focus on revenue targets and things like that was going to take people’s eye off the ball. And I also think it sort of takes away from a long-term orientation, which I think is going to be critical for Ginkgo. So that’s why we made that decision. Happy to talk more about that in questions or otherwise, but just so you know where it’s coming from.

All right. Okay. So as I said, our mission is to make biology easier to engineer. We had 3 really amazing things happened last quarter. So I was going to run through them. So first, we had an announcement of a project we’ve been working on for the last 6 months with OpenAI. This is their blog post about it, where we talked about connecting GPT-5 as sort of an AI scientist, so doing the work of a scientist designing experiments, except they would submit those experiments to our autonomous lab here in Boston. The lab would conduct the work, send that data back to GPT-5, and then over the course of 6 rounds of doing that, we were able to beat state-of-the-art on a pretty complicated sort of scientific — experimental scientific challenge in cell-free protein synthesis by 40%.

What I think is cool about this is, number one, the sort of views on this X post where they announced Codex on the same day, about equal to what they saw with Codex, right? So I think there is really a lot of excitement right now in how reasoning models can enter the physical world, all right? I’m going to talk in a minute about that in the area of transportation where like Waymo’s have brought them into the physical world. But I think we really stand to be the ones to bring AI into the physical world of the lab. We are absolutely in the pole position on that. So I’ll talk more about that in a second. Second, we were — I had the absolute privilege to do a press conference with the Department of Energy Secretary, Wright up at Pacific Northwest National Labs in Washington, where we announced in December that the first 18 robots that we were installing for PNNL as part of the Genesis project.

This is a new project out of the White House to bring AI into science and AI into the national labs in particular. But alongside that ribbon cutting and the Secretary got to sign one of our system, you can see him signing it there. We also announced a new $47 million contract with the Department of Energy to build a 97 robot, 97 RAC autonomous lab at that same site in PNNL in the future. So really exciting, and I think this showcases that autonomous labs are of interest to the federal government, which is the other big pool of research spending. So a place like the NIH is spending $40 billion a year on lab work, frankly. And that’s pretty close to what you’re seeing in the pharma companies as well. So those are your sort of big pools of spending.

And so I think it’s important to see it coming from the federal government as well as from pharma companies. Last but not least, we had SLAS. This is the Society for Laboratory Automation and Screening Conference. I was just at the conference center. It’s about 5 minutes away from here, very fortunately for us at Ginkgo. So we hosted tours of Nebula, our now more than 50 rack RAC lab set up here in Boston. We had 590 people come through, and it was very eye-opening to see what a difference it made for people to see a lab like this actually doing real science during the day, right, like people coming in and just seeing what our scientists were doing with it. It was eye-opening for them. And so I think this makes it more and more clear to me that we’re making the right choice with this focus in ’26 on really driving the further expansion of the system.

We’re going to go from 50 RACs to 100 RACs by H1. That’s the sort of stuff I want you to be following. How quickly are we able to expand that? How quickly are we able to add more of our work onto that system because that’s exactly what our pharma and national lab and university research leaders are going to be looking at to see if they want to buy a system like this. Okay. All right. So now I’m going to do a deep dive into autonomous labs because, again, I think this is really our focus, certainly in ’26, and I think the technological foundation for the company over the next decade. So I’m going to talk about [indiscernible], what is an autonomous lab? Why is it going to transform biotechnology? Secondly, what does it look like very specifically?

Like what do you need to have the lab be able to do in order to deliver biotech R&D? And then finally, how are we going to bring it to market? And the 2 ways we’re doing that is, one, we’ll build one for you like we did at PNNL. Two, the beautiful lab that you just saw pictures of, we’re able to run that sort of in a cloud service model through our R&D services and new services we’re adding coming up. All right. Okay. So here’s the analogy I like to give. I talked — give a talk at SLAS, and I [ did talked to ] this with a lot of people. I think it’s a good one, all right? So I’m going to start in the transportation industry to help explain what I consider autonomy to be. So if you look at this chart, on the y-axis, you have the amount of automation, all right?

And then on the x-axis, you have the flexibility of the request from a user to that automation that it’s willing to tolerate. So in transportation, if you have a low amount of request flexibility and a high amount of automation, that’s a subway, right? You sit down in the back of a subway and it just takes you away, right? You’re not having to do anything. It is fully automated. But you better want to go to one of the stops on that subway line because it’s not going to take you to your house or the grocery store or just wherever you want to go, it’s on rails, all right? So it’s very inflexible. Now low automation, high amount of request flexibility. That’s a car, right? You put your hands on the wheel, your feet on the pedals and you can drive it straight to your front door or to that grocery store, right?

And those 2 poles is basically what the transportation industry has looked like for the last 100 years. Forward slide, unless you’ve been in San Francisco over the last 4 or 5 years and you see these driving around. So this is a Waymo. You sit in the back seat, just like sitting on a subway seat, you do absolutely nothing. It takes you away, except unlike the subway, it will take you right to your house, right to that grocery store. So it has the flexibility of a car, but the automation of a subway. And that’s such a surprising thing that we’re giving it a new word. We’re calling it autonomy, all right? And I think you will see this replicate. When you’re seeing all this interest in like humanoid robotics and all this like, there’s a huge amount of investment going into it right now.

What we’re trying — what’s happening on a broad investor level is the industrial revolution was essentially the application of automation and systematization to all of the tasks that were like low flexibility required, like back to the loom, right? Everything that wasn’t that required a lot of flexibility, we kept manual. And what’s happening now is the AI models are getting good enough, the software is getting good enough to allow automation to be applied to flexible things, and we’re going to see how far we can push that. And the more you can push into flexibility, the bigger the opportunity there is for robotics. And so we’re going to drive that change in labs. Now last point, this is the kicker. If you were to look at the split between miles traveled on trains and subways versus cars and trucks in the United States anyway, it’s 99% cars and trucks.

because you need that flexibility to go places, right? It’s a requirement. It’s not like we didn’t know about rails. They just did not tolerate the flexibility needed. All right. So let’s look in the lab. Low amount of flexibility, high amount of automation, so up where that subway was on the last slide. We do have that actually. We call it automation work cell. And you can buy this from companies like [ Hirose Bio ] and Biosero and Thermo Fisher. And basically, you tell them what protocol you want and they build you a work cell that will run that protocol for you. And it’s great. It’s totally end-to-end. There’s not a person in the middle. It’s fully automated. It’s at top of that chart, but you better be asking for the same protocol that you asked it to do yesterday because it cannot handle variety in the request from the scientists that are using it.

Drop down that automation line and go over on request flexibility. So not automated, but very, very flexible. That is the lab bench. And we’ve had it for 100-plus years. It lets you do whatever experiment you want and the scientists, the human scientist in the middle is what’s providing the flexibility, all right? And that’s what the system has looked like. We’ve had work cell automation for 40-plus years now that we’ve been kind of those 2 poles for the last 40 years. And much like research — or sorry, much like transportation, I had a couple of heads of R&D and 2 pharma companies over my house during SLAS for dinner. And I asked the question, what’s your spend between work cells and lab benches? And they said, actually, 99% on the lab benches.

But let’s call it more than 95% of the research budget is going to the lab bench. And it’s for the same reason that 99% goes to the cars and trucks. You need the flexibility to do science. And you can tell this if you walk around at Merck or Pfizer, Takeda and you walk the hallways, you will not see robots. You will see lab bench after lab bench with benchtop equipment on top of it and scientists basically being the human glue that connects all that different equipment and manages to do liquid handling by hand with pipettes and all the things that they do. That is the overwhelming majority of research spending and pharma is doing, again, $40 billion to $60 billion of not development, but research spending every year through that platform. All right?

What are we trying to build? We’re trying to build that Waymo. What Ginkgo believes we have when it comes to our RAC hardware and then very importantly, the software that runs it is an autonomous lab. It is the flexibility of the lab bench, but the automation of the work cell. And that is, we believe, fundamentally different. It’s a much bigger market than the work cell market. The work cell market, again, just like the subway is very limited in terms of the amount of research dollars flowing to it. And so we want to go right at that autonomous lab market. The key technical question, the next slide is how do you get both high automation and high flexibility without having those human hands in the lab, right? And so that’s the next thing I want to talk about.

What do we actually have to pull off technically to make this a reality? What are people so impressed with when they come visit our lab here in Boston and see what we’ve built. All right. Also, I don’t know if you’ve noticed, if you follow me on LinkedIn, you’ve seen I’ve become a bit of an influencer lately. So this is what it looks like if you’re standing at a lab bench, doing your work by hand. And the real major activities is, number one, you’re serving as a manual liquid handler. In other words, you are moving small volumes very precisely between different liquid containers to set up an experiment with the right materials in it. Second, you’re moving samples, in other words, that liquid you just set up in a plastic tube or whatever it might be in to different devices across the lab.

So you are moving samples as the protocol demands across maybe 3, maybe 10 different devices depending on the complexity of the protocol you’re doing. And then finally, every time that sample ends up on a device, all those devices, these are all like complicated long-tail scientific devices. They have some settings that you need to set in order to have it do the thing you want it to do. And so you, as a scientist are the one putting those settings in, and you’re either doing that with a touchscreen or with sort of third-party software. Okay. All right. So to replace traditional labs, an autonomous lab has to do those same things I just said, that’s 1, 2 and 3, reliable liquid handling, material transport and parameterized control of the device.

But very importantly, if you think about what one of those labs at Takeda or Merck looks like, in one floor with a bunch of benches and maybe 20 or 30 scientists using it, you’re going to have more than 50 devices around that lab that those scientists are making use of, different ones, different days, different ones as part of different protocols. So you’ve got to be able to put at least 50 devices into one big setup. The other thing that those scientists are doing, when they — the first scientist gets in the lab in the morning, they do not close the door behind them, lock it and put up a sign that says “lab in use”. No one else can come in, right? It’s busy. Lab is busy. But on a work cell, like one of those subway system automations that we have in the lab, that’s exactly how it works.

Once it’s in use, you cannot interject yourself into that process and submit a new job. But in the manual lab, absolutely 10, 20, 30 scientists are all walking around that lab, basically sharing the equipment and avoiding each other’s usage of the equipment. So if I’m using something in the morning, you’ll use it in the afternoon. But other than that constraint, they have access to all that equipment and they can use it in parallel. And then finally, it’s very easy to use the lab bench. You don’t have to write software programs and things like that. I won’t have as much time to talk about it today. But in the coming earnings call, I’ll do a little bit of a deeper dive on our software. But one of the things we’re really benefiting from is all this investment in coding agents, things like Codex, from OpenAI and Claude code are now allowing human language to turn into pretty complicated software.

We want to turn scientific intent into work that runs on automation without scientists needing to code. I think that is going to be very doable, thankfully. And that’s number six. It needs to feel like when I go in the lab every day to do my work, I don’t have to sit down and write code. You shouldn’t have to do that for the autonomous lab. This is really a difficult set of challenges. Work cells today, do those first 3. They deliver liquids. We have liquid handler automation, companies like Hamilton and Tecan have been around for 25 years or great — more. They’re great. Second, reliable material transport can be done with arms. And third, parameterized control is doable. 4, 5 and 6 are not delivered well by traditional lab automation today, but we do have it working at Ginkgo.

All right. The first thing to understand about how to deliver 4, 5 and 6 is that a work cell, in other words, that subway is designed around a protocol. So the first thing one of those companies will ask if you’re going to build — they’re going to build an automation system for you is, what’s your protocol? Are you doing high-throughput screening? That’s one of the most common ones, antibody developability, protein production? What is it you’re doing, right? And you say, “Oh, I’m doing this, these are the steps. This is the equipment I need, and this is the throughput”. And then they’ll design a subway that delivers you to that stop. Autonomous labs are not designed around your workflow, but they’re rather designed around the equipment because this is exactly what happens when you’re setting up a new manual lab at Takeda or Merck.

If you’re the person in charge of that lab, you’re that kind of group leader, you ask your scientists, what equipment will they need to do their work over the next 5 years in that lab. They don’t know for sure what protocols they’re going to do, but depending on the type of work they’re doing, mammalian work, bacterial work, cancers, whatever, they’re going to use different types of equipment. So we oriented the design of our robotics hardware, not around a protocol, but around the device. And so this — here, you can see our RAC automation carts. Inside each cart is a device. In this case, that’s a centerfuge, a 6-axis industrial robotic arm and a piece of MagneMotion track. And that track allows you to deliver a sample between connected RACs. So each one of those RACs, their little tracks connect to each other and you can send samples around and deliver them.

If you go to the next slide, we can show a video here of samples moving through our autonomous lab here in Boston. And this is actually — interestingly, this is one of the protocols from OpenAI, right? And so what you can see as this runs is, we have the sample getting put on to the track. That’s a 384 well plate in each well of that plate is a set of conditions that were designed by GPT-5. The plates travel on that MagneMotion track. And in this case, they’re delivered to that centrifuge, right? And so the centrifuge is going to spin down that sample. So it just happens to be the first step in this protocol. Now it’s going to one of those liquid handling devices. So this is what’s called an acoustic liquid handler. It moves liquids with sounds.

So one of the things that’s great about this is it actually can handle smaller volumes at a greater precision than a scientist could do by hand. So we can move nanoliters of liquids around. As a scientist using a pipette by hand, you’re kind of limited to microliters in terms of your ability to be accurate. Now we’re going to be adding, in this case, DNA to each one of those wells. So the project we did with OpenAI, a piece of DNA was being added to what’s called a cell-free mix of reagents. And the idea is that cell-free mix turns that piece of DNA code into a protein. And the protein level is what we’re trying to optimize with OpenAI. We’re trying to see, could you change the conditions such that you got higher protein production than any scientist had shown before in the literature.

So once that DNA got added, we now shake it up, make sure it’s well mixed. And then it’s going to end up onto an analytical device in order to basically run the reaction and then measure the levels of protein that are coming out of that particular — of each well in that 384 well plate. And so to give you a sense for the OpenAI project, each time we did a round with the model, we ran 100 of these 384 well plates, collected all that data, gave it back to the model and then the model was able to design the next round of experiments, okay? So that’s what it looks like for a sample to move through the system. At the beginning of that video, you would have seen a quick picture of sort of the data coming in, like the particular designs from OpenAI and then the scheduler.

That schedule on that one was just running the one protocol. This is what the scheduler looks like when our people — scientists at Ginkgo have submitted 30 protocols to the system. And so what you’re looking at is each row in this is a different device on the system. And then the X-axis is time. So that orange bar in this case is like now, all right? And what is great is we can basically predict the future, right? We know the system knows exactly what piece of equipment is going to be used for what protocol and each protocol is a different color on this chart. What piece of equipment is going to be used for each protocol in the future. And what the scheduler does is if you showed up at Ginkgo as a scientist and you submitted a new job into this into the — our autonomous lab, you would say, okay, I’m using the centrifuge for 5 minutes.

And then I can wait any — up to 2 hours before I need to end up on the echo and [indiscernible] you would specify with time windows your protocol. The scheduler will check, could you fit in? And this is very analogous to one of those scientists walking around the manual lab asking their benchmates, Hey, when are you going to be done with the PCR machine? How long — is it okay if I run the HPLC overnight, here something, do you need to get on it, right? Like having conversations about the availability of equipment, except in this case, it’s all computer-controlled and computer scheduled, so we can essentially schedule it perfectly. And so as you add more protocols in, there’s a complicated algorithm to handle all this. We are the only people in the world as far as I know that are doing anything close to this scale of variable protocols on a single automated system.

And that was pretty well confirmed by wide open eyes during the SLAS tour when I was able to show this off to people. Okay. So we go to the next slide. This is just a different color. So each of those protocols, you can see being submitted by a different user at Ginkgo as well. So that’s, I think, actually really interesting where we have not just like a large number of protocols, but also a large number of unique users submitting those protocols. That’s also very unique in the world. When you have those work cells, there’s an automation engineer or 2 who are sort of in charge of it and everything funnels through them. In the case of our autonomous lab here in Boston, we have tens of scientists submitting protocols every day, different protocols from yesterday that are all scheduled simultaneously.

Okay. So hopefully, that next slide, hope that gives you a picture of how we’ve checked off the sort of 4, 5 and 6 on that list in terms of many pieces of equipment, all in one setup being run simultaneously in parallel, easy enough to use by scientists who aren’t automation engineers. Just note that system that’s now 50-plus RACs and Ginkgo started off as, I think, 7 or 8. It’s very expandable. In fact, on the next slide, after we finished up at SLAS, we were able to bring over the RAC carts that we had at the conference, it was, I think, 7 or 8 and install them all in a day on the system. So the ability to really grow this system is, I think, again, unique when compared to traditional sort of subway style automation. All right. So what’s the value prop to customers?

There’s 3 things, I think, that the — like a large biopharma or a national lab would get excited about. First, save overhead costs by closing your traditional labs. This is one of the things I’m most excited to do this year with our CRO or kind of research services that we run on — across all our labs at Ginkgo. As I move more and more of that work onto the autonomous lab, I can shrink the footprint of my labs, which saves me in EHS costs, saves me in rent, saves me in all these different things that you have to carry when you’re running these labs. Second, it increases the research productivity of your researchers. So right now, a lot of their ideas are ultimately bottlenecked by the amount of time they have to spend in the lab. We want to really open that up and get much more data per research dollar out of your scientists.

And then finally, like we did with OpenAI, you can have AI scientists run what are kind of in the industry called lab-in-a-loop experiments, where the AI model is designing experiments, they’re running on the autonomous lab and data is going back. And so we’re seeing a lot more interest in that from pharma companies as well. All right. Okay. So the last section I want to talk about is how are we going to sell these autonomous labs, and there’s sort of 2 ways we’re going to do it. One, we will place a system like we did Pacific Northwest National Labs. We will place it at a customer site. We’ll sell CapEx. We’ll sell basically service fees, both for the software and for our maintenance of the equipment. And in the future, I could see us even selling things like reagents and consumables and things like that to the users of our system that are sort of automation specific.

Additionally, we have this big autonomous lab in Boston that we can offer services on top of. All right. And so what’s like the overall market potential? This is back to that 1% on the subways, 99% in the cars. The overwhelming majority of research spending, that $40 billion to $60 billion in pharma, the $40 billion plus in — from the government and so on, that’s all funneling through ventures today. And that’s before we also — the other big industry we haven’t talked about at all is sort of diagnostics, and I also see opportunities there as well. So all of that bench labs spending, I think, ultimately has the opportunity to funnel through our platform if our autonomous lab is able to replace the bench. The way we’re going to get there is we’re going to start by commercializing in 2 ways: First, build those autonomous labs for customers; second, run the cloud lab.

All right. So cloud lab services. Two of them are ones you’ve already heard about. So our Solution services. This is where Ginkgo scientists use our autonomous Lab to deliver a research outcome to a customer. So our deals with Novo Nordisk and Bayer Crop Science and Merck and Pfizer and all these people over the years where Ginkgo scientists use Ginkgo’s labs to deliver a research outcome. We get a royalty, we get milestones. We can structure these in different ways. We do a lot of work with the government in R&D grants and things like that through our Solutions business. Second, in data points, customer scientists use our autonomous lab. The design what they want to run on it. This is run like a traditional CRO. There’s no royalties. There’s no milestones.

We send them a huge amount of data back usually to their ML team, and they use that to train bio AI models for protein design or RNA design or whatever they might be doing. The third, and I’ll have more about this in future earnings calls, but we’ll be announcing this soon, is our cloud lab offering. And so what we’re going to have here is customer scientists outsourcing small amounts of lab work directly to our autonomous lab. So think like a $50 order or a $200 order where the actual experiment will get run on the cloud lab and data will go back to the scientist. And I think this is a great way for scientists who are curious about what it’s like to engage with autonomous labs to sort of try it before they buy. And there’s lots of things to try there, different ways to bring it to market.

You’ll hear from us coming up on that. I’m really excited about it. Just to say, we’re not new to the Solutions business. We’ve done 250 partnerships in the last 10 years. We’re continuing to sign these every quarter. We’re doing a lot of business with the government and large pharma are really the 2 areas where we see the most of this, but also agriculture. Industrial biotech has been a lot harder since 2022, basically. But ag, pharma and government still will sign up for Solutions deals. The other area that’s grown really well for us over the last year, and I want to give a shout out to the Datapoints team at Ginkgo is we’ve been growing this business where we run our robotics to generate big data sets against the designs of customers. And this is a business that got started 1.5 years ago.

We’ve now worked with 10 of the top 20, I think, or 30 top pharma customers just in the first year we launched this thing. So people are really excited about it. It’s a good fit. We’ve actually released a bunch of public data sets. If you go to the Datapoints website, you can download some of the largest data sets for drug seek and for antibody developability and things like that. Next slide. I think that we’ve also done a really nice job being a community builder here. We’ve launched a developability competition. We have a virtual cell pharmacology initiative where we do free data generation as part of building up a big public data set. So really, I think if you’re interested in this area, if you’re doing bio AI, definitely check out data points, come to some of our events.

The last point I’ll mention about us running the CRO labs is that our scientists using our big autonomous lab in Boston is a little bit like the Waymo engineer 5 years ago going through Palo Alto, sitting in the driver seat with their fingers like this, right next to the steering wheel, like ready to grab it if the car turns into a mailbox or something. They are the first ones to push lab autonomy to the frontier, right? When you saw those 30 protocols running on a system there, we were the first people to do that, right? And so things break. And that allows us to very quickly speed our development cycle on the autonomous lab compared to companies that really just focus on robotics or on software or things like that because we are doing wet lab research on our own infrastructure, we are learning really fast about what works and doesn’t and very importantly, about how to onboard scientists into autonomous labs.

Like that is a cultural change, right? And so it involves technical tools to make that easier and faster so that they can still get their very important work done, but they can run it overnight. One of the things that our scientists have really enjoyed doing. If you watch like the ramp of protocols getting on to Nebula, our big autonomous lab here, it spikes in the afternoon and then people whose experiments run overnight and they come in the morning to data, which it’s been a while since I’ve been at the lab, but that’s sort of the dream is to show up in the morning with a coffee to a fresh data set. So I do think scientists get really excited about this as we bring the barriers down. But again, Ginkgo’s team gets to be the guinea pigs so that our customers of the autonomous lab end up being able to see what’s possible and also have a lot of that debugged in advance.

I’ll talk a minute about our OpenAI project. If you’re sitting in the back of a Waymo, an autonomous car, you tell it where to go. Once you close the door and get out or get out and close the door, Waymo’s AI takes over and it tells the autonomous car where to go. So what the autonomous car is solving is the problem of replacing the manual driving, not the directing. Same idea here. When we’re solving the autonomous lab problem at Ginkgo, what we’re solving is the manual lab work. not the directing of what lab work to do. And that could be done by scientists as is done every day at Ginkgo, as you saw all those protocols on the scheduler, but you can also try to have those experiments run by AI scientists. And so our project with GPT-5, like I was mentioning, we were doing 100 of these 384 well plates.

We’re giving that data back to the model. It was interpreting the data and then sending in new designs. We have a great archive paper on this, if you Google, if you look at the OpenAI blog post, you can find it. We learned a number of things about this. I think we did some really smart stuff with OpenAI here. I’ll just give you one quick vignette. So the model is designing the parameters of the experiment, but we don’t let it just run anything. We had what’s called a pydantic model, which was basically a software-defined set of rules, and we have open source this, you can download it, where GPT-5 submitted designs into that, and it had to pass a series of tests for us to be willing to run it. And if it didn’t, we would tell, it failed and it would redesign until it met the test.

So simple things, 384 well plate submit 384 wells. The volume of the well is this much liquid, do not exceed that amount of liquid or else it’s going to spill everywhere, right? But more complicated things, do your experiment and quadruplicate because we want to publish a paper about this and scientists are going to want to see replication. Include a set of standard controls, experiment to experiment, so we can fairly compare how you’re doing over time. So we put those rules in. But then within the experimental wells, like the rest of the plate, as long as it put the right amount of volume, it could do whatever it wanted. And so across 500 plates in the experiment, we had only 2 that we thought were just total nonsensical designs. And one of them was a problem with our pydantic model where GPT-5 designed negative volumes of certain reagents to try to squeeze more reagents in, under the volume limit.

Obviously, you can’t do negative volumes. So we added that to the model and it learned not to do that. So really, I think this is the first demonstration of really more open-ended experimental work, beating state-of-the-art. There’s definitely really great ways to take this work in the future that we’re going to continue following up on. OpenAI basically used us as a cloud lab, right? They paid us to do the data generation and their model was able to send and receive commands and data back from our autonomous lab in Boston. All right. I’m going to sort of end on this next one or nearly. Ginkgo is the right company to bring autonomous labs to market at scale. I deeply believe this. This is now apparent to me, in particular, over the last quarter.

We have our cash burn under control. That’s why we wanted to guide to that and keep the team and have investors understand how much we plan to invest in this. We have extensive practical experience automating lab work. This is what we have been doing the last decade plus at Ginkgo. We know what is hard. We know what it takes to move bench work on to liquid handlers to what are the little tips and tricks associated with each benchtop device when you run it at high throughput and high capacity. All of that information is getting embedded into models and our software to make this really just work magically for scientists as they move to the autonomous lab. I think we’re the only ones to do it, and it is a dead mission fit for making biology easier to engineer.

I’m convinced that the #1 problem in that space right now is the lab work. We are just not able to try enough genetic designs to get good at genetic engineering. That’s not Ginkgo. That’s the whole industry. Next slide. do you want to mention because I’m sure some folks don’t tuning in are scientists or potential customers and so on. A lot of times, we also hear from scientists, hey, should I be worried about this? Obviously, part of my job is working at the lab and generating this data. I really like this old advertisement from 1951, IBM, and it talks about how the mechanical calculator or electronic calculator, I should say, is going to have to do the work of 150 extra engineers at your company. And what I love about this is, if you’re not familiar with that device, [ you’re ] the younger folks or whatever on this call, and that is a slide rule.

So this is back when computation was done manually. And this device, this predates general purpose computers. This was literally just a device that like added and subtracted and divided the basic arithmetic was going to do the work of 150 engineers, and you might say that device will replace 150 engineers. Now of course, you fast forward 70 years, and there are 100x more engineers than there were back in 1951. And that’s because the return on investment on what is in the head of people who understand engineering increased dramatically on the other side of the automation of the manual work of computation. And if you go to the next slide, I very much believe that will be the case for the manual work of laboratories. What is — it is insanity that we take people who are PhD caliber, understand all the biology, like all the ins and outs of these ridiculously complicated biological systems.

They have to understand human biology, 18 other things. And while they’re at it, they have to be extremely careful laboratory technicians in order to move liquids and do this work with great fidelity to even be able to try and test and hypothesize their experiments. We need to divide those 2 things just like computation did back in the 1950s. And if you do that, I assure you, you will get many, many more genetic engineers, many, many more scientists than we have today when our ROI is limited by the manual work at the bench. And we just got to do it, and Ginkgo is going to do it. So please put down your pipettes and join us if you’re interested. All right. So next slide. That’s my e-mail up there. As always, feel free to e-mail if you’re excited about this stuff.

I appreciate the time today and happy to take questions.

Daniel Waid Marshall: Thanks, Jason. As usual, I’ll start with a question from the public and remind the analysts on the line that if you’d like to ask a question, please raise their hands on Zoom, and I’ll call on you and open up your line. Thanks, everyone. Thanks, everybody. All right. Let’s get started. So the first question that we have is from BBAGGUE, and this is on X. With the planned expansion of RACs capacity from roughly 50 to 100 units at the Boston facility, could you help us understand how this increased capacity is expected to translate into 2026 revenue? Specifically, what portion of that contribution do you anticipate will be recurring in nature, for example, software, operations, consumables versus project-based services?

Jason Kelly: Yes, I can take that. So for starters, the RAC expansion in part is to, again, be able to move — we have extensive labs here where there’s a variety of different levels of automation. There’s sort of walk-up automation where a person is going up to like a liquid handler, the liquid handler does some automated work, then you take that sample somewhere else in the lab. Work cell automation, which is that subway I was talking about and then benches for some of our lab work, we’re still at the bench, right? And so we want to take all that work and basically shift it on to Nebula onto that big 100 RAC system in the coming year. So that’s sort of the point of it. Now on top of that system, we’ll do those services, right?

So we’ll have our data point service, our upcoming cloud lab service and then our Solution service. So I can speak to sort of the repeatability of those services. The way Solutions deals work, those are usually multiyear R&D deals. So there’s some reproducibility like across a deal, like we do a big program for ARPA-H, our long-standing partnership with Bayer Crop Science that’s been going on for 5 or 6 years. So you have some repeatability inside a contract. But each time we do a new contract, that is hunting for a new project. Datapoints is a bit different. We are starting to see now, as I mentioned in the call, we’re in 10 of the top — again, I think it’s 20 or 30 pharma companies where we have — that is becoming more repeat business as we’re able to basically build trust with those partners that we can serve as that outsourced data generation for their teams.

The cloud lab one, we’ll see. I mean that’s a new experiment where I want to go after that smaller batch work from scientists at the bench. I think if you look at more narrowband CROs, the folks that like build DNA, express proteins, these sorts of things, I think you do see a lot of repeat business once someone has confidence with a vendor. We’re obviously trying to do more flexible work, and that will be a new experience. We’ll see how it goes. But I’m hopeful that would also look like repeat business on the platform. So of the 3, I think Solutions is the one where you’re really off hunting each time to add new research partnerships, but the other 2 are a bit more repeat business. And then just to mention it, that’s all on the side of using our system in Boston, which what you asked about.

But of course, we’re also selling our system. So like we sell that system to a PNML or a pharma company or whomever. When our robotics go in, there’s an initial spend on CapEx, but then we have a service and software license that’s ongoing over time that is also repeat, more like a SaaS business. And then if you were to — if we had like sort of specialized reagents, some high throughput things, as the system was used, that would also be repeat business as well. But the initial CapEx will be onetime.

Daniel Waid Marshall: Thanks, Jason. Our next question is from Brendan at TD. He actually has 2 questions. And so I’ll start with the first one. The first one is, how should we think about U.S. onshoring of manufacturing as a potential tailwind to RACs revenue growth? And what do you think Ginkgo will need to do to maximize share of this trend over the next couple of years?

Jason Kelly: Yes. So on the manufacturing side, we have been seeing interest for — on the sort of manufacturing QC, right? So again, what our systems are doing [ in a ] typical manufacturing environment, you’re going to be doing production in larger tanks. The RACs are really about integrating laboratory benchtop equipment. Now you do have a bunch of laboratory benchtop equipment in manufacturing plants, and it’s used to do quality control across batches, sometimes associated with even kind of like semi diagnostics work associated with following up on patients over time in a lot of these drugs. So there’s actually a decent amount of lab work tied to post clinical sort of like once a drug is on the market, that does keep going in a repeated way.

I think our sweet spot there would be being able to handle many different QC protocols on one big system, right? So again, the strength is complicated protocols, multistep, but today, you may be doing at the lab bench. You often have folks at those research centers that are part of the kind of manufacturing team and so on, not sort of open-ended research PhD scientists. And so being able to do sort of like the latest type of assay or something that — being able to deploy that out as a QC step, I think the RACs open up the door for that. So we are talking to some customers about putting our automation into manufacturing sites. So I think there may be a little bit of tailwind there. We’ll see.

Daniel Waid Marshall: Cool. Next question is, how is Ginkgo’s Datapoints offering being received among customers? Are there any material tailwinds that you expect for this part of the business over the next 12 months?

Jason Kelly: Yes, really well. I think we’re kind of finding the sweet spot in providing — so like what’s going — I guess I’ll just make this quick point on the AI side, right? So there’s sort of 2 halves to how AI is impacting the biotech industry. One is what I just spent a lot of time on the call talking about, which is reasoning models and coding models, the same sort of models that everybody is using in basically any information technology space are going to impact the ability for scientists to use autonomous labs and robotics in the lab. That’s all being made easier by reasoning models. Separate from that, there are bio AI models. And the most famous one of these is AlphaFold that Google came out with, which was a model trained on not human language, not human reasoning, but rather biological language and in particular, in that case, like amino acid sequences from proteins and the structure of those proteins.

So there’s a lot of work going on in that area. And in order to build those bio AI models, you need to generate very large data sets with sort of a variety in that case, say, of proteins and their structures, but there’s many other things in functional genomics, antibody developability, other areas where pharma companies are asking us to make those big data sets for their ML teams. That side of the house has gotten tailwinds. There was a lot of — if you’re at the JPMorgan conference this year, folks like [ Chai Bio ] announced their partnership with Lilly and [indiscernible]. There’s a whole bunch of companies in the start-up side who are starting to partner with the large pharmas because they have great bio AI models. And the reason they have great bio AI models is they have proprietary data.

It’s not just how smart they are at the modeling, it’s that they’ve been generating these large data sets. So that’s sort of opening people’s eyes up and I think creating a little bit of a wave there. And we’re definitely the right, I’d say, the leader in providing data sets to again to the large pharma, large biotechs, ML teams that are all ready to go for training and everything else. We can do the robot half and the data cleanup half and they can focus on the biological modeling. So it’s been good, I would say. Yes, I’m excited about it for this year.

Daniel Waid Marshall: All right. Next question is from at [indiscernible]. This is on X. The question is about basically not just the utilization of racks, but the manufacturing and deployment of racks. And so the question is, similar to how Tesla used the giga press to dramatically improve manufacturing efficiency. Is Ginkgo exploring any specific strategies, technologies or methods to significantly enhance the production efficiency and scalability of RACs?

Jason Kelly: Yes, it is something we are starting to think about. I mean, mainly because those things take time to put in place. We did make some good decisions. So over the last 4 years or so, I guess, since we acquired Zymergen where this technology started. We actually did do a like a generational upgrade to the hardware of the RACs and made them, and that design change was not really about, in fact, our old RACs that we had back in the day Zymergen needs are — the new ones are compatible, but the system has many less components. And that was all done for design for manufacturability for exactly this reason. We make them today in San Jose. We do final assembly in our — through a partner, and then we do final assembly and integration with the third-party equipment at our site in Emeryville, California.

As we were to scale up and selling more and more of these, I think you will see us invest in like basically larger partners to repeat that manufacturing process. But even as we have it now, we can actually scale pretty decently on this. So it’s something I think we want to plan ahead for, but it’s not the immediate problem we have.

Daniel Waid Marshall: All right. I think that that’s all that we have for tonight. So of course, if anyone has any questions in general, they can always e-mail us at investors@ginkgobioworks.com. Jason also put his personal e-mail up there earlier, so you can message him. And yes, thank you, and have a great night, and I hope everyone has a great quarter.

Jason Kelly: Thanks, everybody.