Tempus AI CEO Eric Lefkofsky discusses Q3 results, Personalis acquisition, and data moat
Summary
Tempus AI CEO Eric Lefkofsky reports Q3 revenue of $411.5 million, up 23% year-over-year, with adjusted EBITDA expanding significantly. He outlines a dual revenue model combining clinical sequencing billing and pharmaceutical data licensing, citing a dataset exceeding 600 petabytes. Lefkofsky highlights the strategic acquisition of Personalis for $1.5 billion to expand into Minimal Residual Disease (MRD) testing. He argues that investors misprice the data business, suggesting it warrants a higher multiple if spun out. The company aims to reinvest two-thirds of gross profit growth while targeting future EBITDA expansion.
Episode sections
- 0:00 – 1:26
Intro and Financial Results
Host Austin Hankwitz introduces CEO Eric Lefkofsky of Tempus AI ($TEM). Lefkofsky outlines the company's dual revenue model: clinical sequencing and data licensing. The host notes Q3 revenue of $411.5 million (+23% YoY), adjusted EBITDA growth from $1.5 million to $25.5 million, and a raised full-year revenue outlook to ~$1.6 billion.
- 1:26 – 9:18
Business Model and Data Moat
Lefkofsky explains the two revenue streams: billing insurance for sequencing patients and licensing de-identified molecular/clinical data to pharma. He describes the 'moat' as the difficulty of establishing connections with 8,000 hospitals, comparing it to 'mowing 8,000 lawns.' The dataset exceeds 600 petabytes and grew by 50 petabytes in one quarter.
- 9:18 – 15:49
AI in Healthcare and Physician Tools
Lefkofsky argues AI has made patients 'smarter' but hasn't yet made doctors 'superhuman.' He highlights Tempus's 'Hub' platform, which uses AI to generate case summaries, prior authorizations, and clinical trial matches for oncologists. He notes the goal is to expand these tools into neurology, cardiology, and immunology.
- 15:49 – 19:24
Personalis Acquisition and MRD Testing
The discussion covers the ~$1.5 billion acquisition of Personalis. Lefkofsky explains the strategic importance of Minimal Residual Disease (MRD) testing, which detects tumor cells in blood before they appear on scans. He notes Natera is the leader in this space with ~400,000 tests/quarter and predicts MRD volume could eventually exceed therapy selection volume for Tempus.
- 19:24 – 28:00
Pharma Data Value and EBITDA Economics
Lefkofsky states pharma companies are shifting R&D budgets toward AI/data to reduce high failure rates in Phase 2 oncology trials (60-70%). Tempus has over $1 billion in total contract value remaining. He outlines a capital allocation strategy of reinvesting two-thirds of gross profit growth and letting one-third drop to EBITDA, expecting a future flip.
- 28:00 – 30:32
Investor Mispricing and Future Catalysts
Lefkofsky claims investors discount the data business to zero, trading Tempus at a diagnostic multiple despite having a $500M+ data business growing at 30%. He suggests the data business alone could warrant a 20-40x revenue multiple if spun out. He identifies potential catalysts in large frontier model licensing deals and reimbursement for algorithmic diagnostics.
Claims worth verifying
Q3 revenue was $411.5 million, an increase of 23% year over year.
Unidentified speaker (UNKNOWN) ·0:39
How to verify: Check Tempus AI's Q3 2026 Earnings Press Release or 10-Q filing on SEC EDGAR. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Primary sourceAdjusted EBITDA grew from $1.5 million to $25.5 million in Q3.
Unidentified speaker (UNKNOWN) ·0:39
How to verify: Check Tempus AI's Q3 2026 Earnings Press Release or 10-Q filing on SEC EDGAR. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Primary sourceFull-year revenue outlook raised to about $1.6 billion.
Unidentified speaker (UNKNOWN) ·0:39
How to verify: Check Tempus AI's Q3 2026 Earnings Press Release or 10-Q filing on SEC EDGAR. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Primary sourceThe dataset is more than 600 petabytes and grew by 50 petabytes in one quarter.
Unidentified speaker (speaker_1) ·7:28
How to verify: Verify against Tempus AI's investor presentations or Q3 2026 earnings materials. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Data is pulled from over 8,000 hospitals in the US.
Unidentified speaker (speaker_1) ·7:46
How to verify: Verify against Tempus AI's investor presentations or Q3 2026 earnings materials. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Natera is running close to 400,000 MRD tests per quarter.
Unidentified speaker (UNKNOWN) ·18:31
How to verify: Check Natera's quarterly earnings reports or investor presentations. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Primary sourcePhase two clinical study failure rate in oncology is 60-70%.
Unidentified speaker (speaker_1) ·20:44
How to verify: Verify against industry reports such as Nature Reviews Drug Discovery or biotech analysis firms (e.g., EvaluatePharma). For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Tempus finished last year with over $1 billion in total contract value still to be delivered.
Unidentified speaker (speaker_1) ·21:41
How to verify: Check Tempus AI's annual or quarterly financial filings for deferred revenue or contract liability disclosures. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Primary sourceData business is $500 million growing at 30%.
Unidentified speaker (UNKNOWN) ·29:24
How to verify: Check Tempus AI's Q3 2026 earnings materials for segment revenue breakdown. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Acquisition of Personalis valued at about $1.5 billion.
Unidentified speaker (speaker_0) ·15:49
How to verify: Check the definitive proxy statement or merger agreement filed by Tempus AI on SEC EDGAR. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Primary sourceDo your own research
- Verify Q3 10-Q filing for Tempus AI ($TEM) to confirm revenue of $411.5 million and EBITDA figures.
- Review SEC S-4 or proxy statement for the Personalis acquisition details, including purchase price and integration risks.
- Check Natera ($NTRA) quarterly reports for MRD testing volume claims of ~400,000 tests per quarter.
- Analyze Tempus's 10-K for the composition of the $1.6 billion full-year revenue outlook and segment breakdowns.
- Search for recent FDA clearance or reimbursement codes for algorithmic diagnostics mentioned in the 'Hub' platform.
- Review pharmaceutical R&D failure rate statistics for Phase 2 oncology trials to validate the 60-70% claim.
- Assess the valuation of comparable data licensing companies to evaluate the claimed 20-40x revenue multiple.
- Check for any conflicts of interest or insider trading disclosures by CEO Eric Lefkofsky around the time of the interview.
Full transcriptexpand
0:00Unidentified speaker (speaker_0): Hey everyone, and welcome back to After Earnings. My name is Austin Hankwitz, co founder of Wits Ventures and co host of the Rich Habits Podcast. Today we're joined by Eric Lipkowski, founder and CEO of Tempest AI. Tempest sits at the intersection of health, diagnostics, data, and artificial intelligence. The company uses diagnostic testing to generate massive amounts of clinical and molecular data and then uses that data to build tools for physicians and pharmaceutical companies. So
0:31Unidentified speaker (speaker_1): I think for us to get a real benefit from AI, we need to make good doctors great doctors and great doctors superhuman doctors.
0:39Unidentified speaker (UNKNOWN): Now
0:39Unidentified speaker (speaker_0): Tempest just reported $411 plus million of third-quarter revenue, an increase of 23% year over year. Adjusted EBITDA grew by 25.5 million. from just $1.5 million a year ago. And the company raised their full year revenue outlook to about $1.6 billion.
1:00Unidentified speaker (speaker_1): It's like mowing 8,000 lawns. You know, you can't cheat the system, you have to mow them one at a time.
1:06Unidentified speaker (speaker_0): So today we're gonna unpack how the Tempest business This model actually works, what's driving the growth, how AI could change the way. cancer and other diseases are diagnosed and treated, and what the next several years could look like for this company. So, Eric, welcome to After Earnings. Wow. Eric Tempest is one of those businesses, in my humble opinion, that sounds super interesting, but can be difficult to explain how it all, you know, the flywheel kind of comes together here in one sentence. So starting really simply here. how does Tempest make money?
1:41Unidentified speaker (speaker_1): Well, we make money uh two ways. We have two main businesses. So, unlike some companies that could explain their business model like one way, we have two ways. It's like, for example, if you were talking to Amazon, they might say, I run, I have an e-commerce site and I run this thing called AWS. We make money two ways. We are one of the largest sequencers of patients, particularly cancer patients. in the world. And in that business, we sequence patients when they're sick clinically and we bill insurance and get paid. So that's our uh a biggest business. But then we also, by virtue of sequencing patients, we marry that data with clinical data for those patients, including how those patients are responding over time to therapies. And we generate this very large, rich kind of multiple. multimodal data set that is licensed most of the big people that make drugs for cancer patients. And so we have a very big data and AI modeling business. And so it's broken down into two ways. So sequencing patients, building insurance, licensing de-identified data and building models for people, and then we charge them for those.
2:44Unidentified speaker (speaker_0): Great breakdown. So let's maybe dissect that a little bit further. How does a cancer test that's performed on one side of your business today ultimately become valuable to maybe a pharmaceutical company developing drugs that are not going to come to market for maybe several years into the future?
3:01Unidentified speaker (speaker_1): Yeah. So I mean, the first is, you know, we sequenced the human genome, let's say 25 years ago. And it was very expensive If to sequence patients back then. Over the decade that followed, sequencing costs came down dramatically. And for a whole variety of reasons, one being that cancer is a very complex disease. It's very heterogeneous. It involves a bunch of like molecular things that go awry. We put a bunch of money into cancer research that started in the Nixon administration. For a bunch of reasons, we kind of pointed this amazing capability of sequencing the human genome. at patients with cancer. And so we built up a set of capabilities to sequence those patients and match them to a bunch of targeted drugs that would perform better if you knew what mutation was driving someone's cancer. So when TEMPA showed up, we were already, let's say, a decade into that work. And our big aha moment was hey, we're starting to sequence lots of patients. We're generating lots of data. It's really hard for physicians to contextualize this and figure out like, what do I do with all this data? And so we need to marry it with clinical data to allow them to actually come up with those insights. Like, what's the best clinical trial for my patient, or what drug should I actually put them on?
4:19Unidentified speaker (UNKNOWN): So
4:20Unidentified speaker (speaker_1): we built technology to do that. As we became a much bigger sequencer of patients and started amassing all this data.
4:26Unidentified speaker (UNKNOWN): the
4:27Unidentified speaker (speaker_1): data became so large that it was super valuable to biotechs and big pharmaceuticompanies because they're trying to mine that data to really figure out okay, like, are there new targets I should go after that a patient might respond to? If I'm designing a clinical trial, like, do I need to understand the molecular composition of my patients to make sure that they don't fail? the trial for some reason I didn't understand? Like, how do I watch this like the real-world standard of care as it's evolving rapidly and make sure that my drug doesn't perform worse over time for things I didn't know? So there's a bunch of reasons why they want that data. And as a result, we built a very large data licensing business over the past, you know, five, six years.
5:12Unidentified speaker (speaker_0): Yes, definitely a large data licensing business. You guys just reported 411.5 million of revenue. That's an increase of 23% year over year, with your oncology testing volume actually up 32%, and that data and applications revenue that we were just talking about outpaced all of that at 33%. So when you look underneath those numbers, you know what's driving the business most right now is there maybe one part of Tempest that you think could be, you know, reaching an inflection point that's really just going to start driving a lot of this growth into the future?
5:46Unidentified speaker (speaker_1): Because the business has these two big, two, two kind of big components, they they both have to be doing well to generate the kind of growth we're generating. And we're fortunate that both are doing well. We have just kind of tons of natural tailwind in both those two main businesses. So as you mentioned a second ago, oncology testing, our units were up 32%, which is, you know, a lot of growth given our size and scale. I mean, we'll run something like a million tests uh over the next year. So we sequence a lot of patients. So to have something growing at that rate is pretty is pretty dramatic. And and then our data licensing business is growing even faster because what's happened is that with the last three plus years of the benefit of generative AI and these large language models, the capabilities that now exist inside our biotech companies and big pharma is pretty extraordinary. And so now they're hungry for more and more data to build those models, to generate insights, to
6:47figure out what to do next. And we have the kind of large largest and most valuable repository of that data at the present moment, not just in oncology, but now in other disease areas as well.
6:58Unidentified speaker (speaker_0): I kind of want to talk more about y'all's data set here. At what point do you think the scale of this data set becomes something that's really difficult for maybe another company to replicate? Or, you know, we just talked about AI and sort of how these different you know, models are being used by these researchers. Like, do you think there's a a world in the future where it it's your data set is like I'm sure it is now like the gold standard, but like it really becomes so big, so useful that people just can't even compete with their own. It's just you guys.
7:27Unidentified speaker (UNKNOWN): Yeah.
7:28Unidentified speaker (speaker_1): I mean, the data set's already pretty extraordinary. I mean, if we just measure it by size, it's more than six 600 hundred petabytes uh matched multimodal data.
7:38Unidentified speaker (UNKNOWN): It grew
7:39Unidentified speaker (speaker_1): by 50 petabytes last a quarter alone. To put that number in perspective,
7:45Unidentified speaker (UNKNOWN): you know,
7:46Unidentified speaker (speaker_1): 50 petabytes would be more data than, you know, like some of the largest healthcare systems in the country would have in their entire history. So it's a very large A data set. And the what makes the data set so unique is that we're pulling data from over 8,000. hospitals in the US. And to develop these connections, you have to essentially, these are bi-directional connections where we pull data out of these electronic healthcare record systems and then we put data back in, insights back in. So you have to go to the hospitals and say, we want to build these data pipes. You have to get through regulatory compliance, legal, you have to get through IT, you have to actually do the work. And I, you know, we used to say to people when we were going going public, it's like mowing 8,000 lawns. You know, you can't cheat the system. You have to mow them one at a time. And it's just that's the way it is. And so I think the fact that we've been able to over the last decade establish that many connections and the data sets growing as rapidly as it is, it just becomes harder and harder. for other people to build comparable data sets, which is why our business is growing so much more quickly than other people in the space.
8:59Unidentified speaker (speaker_0): Mowing 8,000 lawns. I mean, sometimes doing the hard thing quite literally can just be the moat of the business because no one else has the capacity or bandwidth to do that same hard thing. Now, I wonder from your perspective, like, do you think the real moat ultimately, is it the diagnostic testing? Is it the data set? Is it the models, or maybe a combination of how all three of them feed into one another?
9:20Unidentified speaker (speaker_1): Yeah, it's it's it's it's all the above. I mean, each of these things becomes a component of like proprietary differentiation. So you need proprietary access to data. And we built these 8,000 connections, but that's only for the clinical data. In our case, the data set's only valuable when it's matched with molecular data. And so we've built proprietary sequencing capabilities as one of the largest sequencers in the country. So you need to have both. You need to be able to generate enormous amounts of molecular data at scale and low cost and high quality that doctors order, because you have to be part of the clinical flow. You need to have it matched with clinical data. And then you need to,
9:59Unidentified speaker (UNKNOWN): if
9:59Unidentified speaker (speaker_1): you could get past those two, you then need to build a uh tool. tools that make that data useful. Otherwise, you just have massive amounts of data. And we've spent the last decade, we've had something like six or seven hundred engineers a year for the last decade working on this problem. So it's an enormous technical challenge. And then on top of that, we've spent the last several years building large compute infrastructure. We have two clusters,
10:24Unidentified speaker (UNKNOWN): one's
10:25Unidentified speaker (speaker_1): about a 1,000 thousand each H200s, two hundreds, one's about a comparable size G V200s that basically are running large pre-training model building and post-training exercises regularly to kind of build proprietary models off our data. So each of these components, you know, you have to kind of you'd have to replicate if you wanted to truly compete with Tempest, have access to vast amounts of clinical data. I think we have 50 million patients that. in the US, have access to massive amounts of molecular data, build technology and tools to make it useful, and then build uh AI models on top of it. So that's a lot.
11:00Unidentified speaker (speaker_0): So from your perspective, it's certainly it it's definitely a lot. And you, you know, kudos to you all for building such a durable business that I feel like is at the cutting edge of just Yes, everything technology and AI related. Where are you seeing, because I think this is a big discussion topic in the country, I think in 2026 and will continue to be in 2027. We hear the word AI or the acronym, right? AI in SI or whatever you want to call it. And we think about, okay, it's impacting work this way or it's helping people do this more efficiently. And we're promised this idea of like AI is going to revolutionize healthcare. From your perspective, where are you seeing AI actually changing payments? patient care, maybe making existing workflows different? Or like are is there a clear no AI is positively impacting this specific thing? And because of maybe data you guys have or access to customers or things like that, like you're seeing that take place in real time.
11:54Unidentified speaker (speaker_1): Yeah. I think look, I get asked this question all the time. As the, you know, we're the literally, I think, at the epicenter of AI in healthcare. So, you know, you can imagine. I get asset a bunch. I mean I think there's been a there's been a clear significant um ad um advantage to patients at scale loading some of their clinical data into these large frontier models andor loading um you know questions about their clinical data into these models and getting insights. We've essentially if you think of it, we've kind of made the average patient smarter. Okay, there's no question. The challenge is, and that works very well, I think, for a lot of the lower level applications, like, you know, how often can I take Advil and Tylenol, or what should I do here? You know, whatever. Okay, things like that. Where it tends to break down
12:50Unidentified speaker (UNKNOWN): is
12:51Unidentified speaker (speaker_1): where the US healthcare system struggles the most, which is people are actually really sick and they're being treated in hospitals and there's a lot going on. These models weren't designed to interrogate that level of complexity. They also typically don't have the clinical data for those patients at enough depth to even be able to do that. So, and that's where the breakdown occurs. And so we've made, if you think of it, we've kind of made patients, we've made um unsophisticated patients more sophisticated for sure. But what we haven't done is done the same thing on the physician side. And so I think for us to get a real benefit from AI, we need to make good doctors great doctors and great doctors superhuman doctors. And we need to basically bring some of the benefits of. these large language models into the clinical practice so that we don't make physicians don't make mistakes. They always have their patients on the optimal therapeutic path. They kind of have vast amounts of information at their fingertips. And those things really are very hard today given the current tools they have.
13:57Unidentified speaker (speaker_2): I want wanna to talk about the physician component there for a little bit too. Obviously, you know, your product is supporting physicians in such an important way. What maybe can a physician do
14:08Unidentified speaker (speaker_0): with Tempest today that maybe just wasn't even possible three, five plus years ago?
14:14Unidentified speaker (speaker_1): I mean, in cancer, they can do a lot. So where we're where our where our solution is most mature, which is oncology in the US, they have access to really incredible uh tools that are all
14:27Unidentified speaker (UNKNOWN): uh
14:27Unidentified speaker (speaker_1): enhanced by AI. So when they order sequencing from us, we run a series of algorithms, AI-based algorithms, that are able to predict all kinds of things. Like if you didn't know, for example, where the primary site of your tumor was, we can predict it. If you might take an immunotherapy and you're not sure if the immunotherapy drug drug's is going to work for you, we can predict that. So we can make a, we have a bunch of AI predictions. running in the background that make our tests super smart. But we also have technology now where when a doctor logs in, our main platform is called Hub. That's where what doctors would log into, or Hub's actually available in their EHR. But when they log in, they can build, for example, agenc summaries of the patient case they show up on the when you open up the system. So we the tools are reading. everything about that patient and summarizing it for a doctor. You can use those same tools, for example, the physician to generate an automat automatic prior authorization. So let's say I want to put my patient on a drug and I have to like write a letter to the
15:27insurance company, you can use our tools. If you wanna, if I wanna figure out if there's a clinical trial my patient's eligible for, you can use our agentic agenetic tools to figure that out. So we've kind of embedded them throughout multiple steps of the clinical workflow in oncology. And our goal is to really now do that
15:44Unidentified speaker (UNKNOWN): in
15:45Unidentified speaker (speaker_1): in neurology and cardiology and immunology and all kinds of other diseases.
15:49Unidentified speaker (speaker_0): That is that is really powerful, Eric. Let's switch gears here. You're in the process of acquiring personalis in a deal valued at about one and a half billion. Why was MRD? See, I know what MRD is, but are listening. our listeners might not. So maybe talk about sort of why that what that is specifically as it relates to oncology and cancer. But why was MRD important enough strategically that you decided to buy personalis in instead of maybe like building out your you know capabilities more internally? And then what does Tempest look like now with personalis that maybe it doesn't look like today? So after that acquisition and all that. stuff gets sort of the synergies as a lot of people like to talk about, what does it look like maybe two, three, five years into the future from your perspective?
16:35Unidentified speaker (speaker_1): So there basically there are in in cancer testing, there's a compendium of tests you you could order as an individual. So the first is: are you at risk of getting cancer? Do you have some kind of hereditary risk that predoped predisposes you to get cancer? And so we're the largest sequencer in that space because we acquired a company called Ambri a few years ago. And they they were a leader in that space and we acquired them. And we felt it was really important to be able to serve our patients even before they get cancer. Then there's, okay, I'm diagnosed with cancer and I have to be treated. And there's two main forms of tests that are run. One's called solid. tumor profiling and one's called liquid biopsy. And what's unique about Tempest is we are a we are a leader, meaning number one or number two, in both of those. So we're equally strong in solid tumor profiling as we are in liquid biopsies, which is unique. Most companies are strong at one or the other, not both. And then, and now there's a whole new category, which is amazing, called MRD or minimal. visual disease detection or monitoring. And what happens is we have this new capability of essentially detecting tumor cells in the blood long before they would cluster into a mass big enough to be
17:54identified by a scan, like a CT scan. And we can detect that these cancer cells in the blood six months, nine months, a year. before the scan would show up. And so this has become incredibly powerful because really, when you're treating a cancer patient, especially somebody who's got late stage or metastatic disease, you can't wait six months or a year. By the time you've determined the cancers come back, it's often too late and impossible possible to treat. And so you got to get ahead of it as much as you can. And so this MRD testing space. which is relatively new, has become quite large. In fact,
18:31Unidentified speaker (UNKNOWN): the
18:31Unidentified speaker (speaker_1): leader in that space is a company called Netera, and they're running something like close to 400,000 tests a quarter. It's an enormous number of tests. And so the the the the kind of second company in in Salud in It what's called tumor tumor-informed informed MRD, which is a the space Natera is largely in, is a company called Personalis. We, we, we began distributing their test a few years ago. And we just suspect MRD is going to be a growing and important part of our uh business. And, you know, it would not surprise me if several years down the road the volume we generate from MRD is greater than the volume we generate from our therapy selection work. And it just has a it's just a big category. So it's a it'll be a big driver of growth for us, I think, for quite some time.
19:18Unidentified speaker (speaker_0): Wow. That was actually my my follow up, which was how big do you think this opportunity is? And just stole it from me here, Eric. But
19:24Unidentified speaker (UNKNOWN): wow,
19:24Unidentified speaker (speaker_0): that is a massive opportunity. Especially, you know, talking about sort of that that leader right now, right? The 400,000 tests that take place on a quarterly basis. Like Whew, that's a lot. Um you guys have more than 550 million dollars year to date of data licensing bookings, and you're significantly increasing large agreements with pharmaceutical companies. Why is Tempest's data so valuable to the Bristol, Myers, Squibbs, or the Modernas of the world?
19:53Unidentified speaker (speaker_1): Yeah. I mean, I think they they
19:56Unidentified speaker (UNKNOWN): as
19:56Unidentified speaker (speaker_1): I mentioned before, like these companies all have, I think, figured out over the past few years that this kind of data
20:04Unidentified speaker (UNKNOWN): can
20:04Unidentified speaker (speaker_1): be transformational to the efficiency of their RD budgets. So the they they right, they're going to spend, these big farmers are going to spend $1 billion, $10 billion, $15 billion, 20, whatever it is. And their entire job is to allocate those significant investments from the earliest stages of discovery all the way through development to get drugs approved and in market. And when they run these trials, then they invest a billion dollars or more and they don't pan out, it's a loss for everybody. And so, and this happens, you know, regularly. I mean, the failure rate of uh phase two clinical studies in oncology is still sixty 60 or seventy 70%, percent, meaning m
20:46Unidentified speaker (UNKNOWN): far
20:47Unidentified speaker (speaker_1): more fail than succeed. And it's an even later stage. they have a far from a perfect track record. So our as these companies come to realize they can use our data and our AI modeling capabilities to help them predict
21:01Unidentified speaker (UNKNOWN): the
21:01Unidentified speaker (speaker_1): likelihood of success, help them design a more intelligent study, uh, help them build um better control groups, help them figure out if there's adverse events that might cause issues, all those things. They just get more efficient. And the more efficient they get, the more they can allocate to technology. And it's basically going to be a content a compendium. Like you, you, you're spending a hundred cents on the dollar in biology and chemistry and clinical development. And you were spending 0% in technology and AI and data. Let's say five years ago, whatever it is. And over time, it's just going to move. We're obviously we're in that movement now. which is why our revenues are growing so rapidly and our data bookings are so high. We have, we, we uh finished last year with over a billion dollars in total contract value still to be delivered to our clients. And that number's done nothing but go up since then. And we just have a it's it's all because of this migration, as more and more of these companies realize they have to allocate a certain percentage of their RD
22:01budgets to technology and AI and. data. And we're just kind of the go-to source.
22:05Unidentified speaker (speaker_0): Now tell me more. I mean, you were just talking about, of course, the growth. Tell me about the growth of your adjusted EBITDA, right? Grew from about one and a half million to 25 and a half million this most recent quarter. What's maybe structurally changing
22:20Unidentified speaker (speaker_2): in
22:20Unidentified speaker (speaker_0): the economics
22:21Unidentified speaker (speaker_2): of your business
22:22Unidentified speaker (speaker_0): that's allowing that to happen?
22:24Unidentified speaker (speaker_1): The the the business basically is just in a really healthy spot in that you have this business that, you know, if you look at most analysts' projections, it's supposed to generate a couple billion dollars of revenue next year. It's it operates at a pretty high gross margin. Let's call it
22:45Unidentified speaker (UNKNOWN): You
22:45Unidentified speaker (speaker_1): know, high 60s, low 70s, like very high gross margin. And the fixed and very variable cost needed to run the business is low. So if I if I if my revenues grow, let's say 400 million next year, using just a number, and let's say that generates 300 million a gross profit.
23:04Unidentified speaker (UNKNOWN): I
23:04Unidentified speaker (speaker_1): don't need 300 million dollars to run the business. So I'm maybe maybe I need 50 million, right? So you have 250 million that you can spend in a in in a in a in in any manner you want. You could choose to invest that back in the business, or you could choose to drop it all to Ebita. And one of the reasons that our our our EBITDA is rising so quickly is, you know, even though we're investing aggressively, we will be we're fortunate given our growth that we'll be able to invest aggressively and generate meaningful improvement to EBITDA. even. So we just have a business that's, you know, when you have a business producing lots of gross profit dollars. you're you're kind of fortunate to be able to either reinvest them or retain them.
23:43Unidentified speaker (UNKNOWN): And
23:43Unidentified speaker (speaker_1): that's the stage right now.
23:45Unidentified speaker (speaker_0): Is there a point where the incremental economics now of the data business can cause profitability to accelerate faster than revenue? Kind of like what you were alluding to here. And then like, you know, dropping it down to the, you know, Justy Justed Diva EBITDA or reinvest in the business. Like I'm just curious, I want to understand. as the CEO of this business, right, are how are you thinking about this growth and balancing growing at 20 plus percent, you know, year over year, as well as growing and sustaining EBITDA?
24:18Unidentified speaker (speaker_1): We we think about we've and we've told the street Treat this we this is not a perfect science, but we think about some kind of ratio like two-thirds of gross profit dollar growth reinvested back in the business, and one-third of gross profit dollar growth
24:35Unidentified speaker (UNKNOWN): um
24:35Unidentified speaker (speaker_1): dropping to EBITDA. And that eventually that flips to one third invested back in the business and two-thirds dropping to EBITDA. EPITA. So, you know, you can,
24:48Unidentified speaker (UNKNOWN): and
24:48Unidentified speaker (speaker_1): we haven't given long long-range long range guidance yet, but like you can imagine there's going to be this inflection point, assuming we keep growing at the rate we have been, where the business is gonna is this going to generate lots of
24:58Unidentified speaker (UNKNOWN): uh
24:59Unidentified speaker (speaker_1): incremental EBITDA.
25:00Unidentified speaker (speaker_0): Amazing. Um yeah, not asking you to give guidance, uh of course, but what are maybe the biggest things that could over the next maybe twelve, eighteen, twenty four months, you know, we talked about this recent acquisition of personalis, uh, the growth that's part of that category. Um obviously we're at the early innings of what AI slash SI could mean for healthcare, but what are the biggest things that could make Tempest materially different over the next 12, 18, 24 months as a business when you're kind of looking at it today as we film November of 2026?
25:29Unidentified speaker (speaker_1): Yeah, I think we're fortunate that we built like a little machine here and it's just cranking. So I think, you know, when you're growing at 25, 30 percent, you know, it's you you want to maintain that. for a long horizon of time. I've long said I'd rather have a business grow at 25% for a decade than grow at 40% for a year and then down to 15. So you really want to maintain that long-term kind of growth trajectory and at scale, these be it's a law of big numbers, right? I mean, we're, you know, we've got to grow to maintain that, you got to gotta grow go 400 million, 500 million. So big, these are big numbers. So I think we're most focused on that. To the extent the business surprises people to the upside in a dramatic way, because I think it could surprise people in, you know, moderate ways. But the dramatic w would be, you know, some of the big frontier modeling folks, you know, who are testing our data to the extent they wanted to license large amounts of data. They just have such large um budgets in terms of what they can spend on data and modeling. that these could be like, you know, just a phase shift higher than our normal deals, which are already $100 million, $200 million plus deals. So that is certainly out there. And then the second is our algos business. We we didn't talk about it. These two main businesses, sequencing and data. We have a third business called algos, where we basically take the data we've generated and
26:48develop purely algorithmic diagnostics. that are basically just zeros and ones, like they're just models that are diagnostically relevant. So, for example, from a digital pathology slide, can I um can I like uh run an algorithm to predict whether a breast cancer patient is estrogen positive or HER2 negative? Um, can I uh if somebody gets an electrocardiogram, an ECG, and they're told they're fine, but we know some percentage of the time these people have a heart attack or stroke within a year. Can I predict that? So these algorithms that run on top of healthcare data we collect at some point will be reimbursed by the U.S. healthcare system. And when they are, that could also be quite large, right? We run lots of these tests where we don't get paid today. And if we do get paid, it becomes a big number, right? When you're running when you're running, you know, 10 million tests, 50 million tests, even a small amount of money is a big difference. So I think those are
27:51things that are out there, but we're fortunate that the main business is just is just is just helping.
27:55Unidentified speaker (speaker_2): I I appreciate that answer. And maybe you're about to answer my final question here for you
28:00Unidentified speaker (speaker_0): with a sim in a similar vein. What do you think is the most important thing about Tempest that public market direct investors still don't fully understand?
28:11Unidentified speaker (speaker_1): A significant number of investors historically have not appreciated the strength of our data business. They've actually they've discounted it to like zero. So if you look at it, if you look at it, we operate uh at the at the at the you know kind of top end of our diagnostic peer group. So if you look at like companies who seem to be executing at the top end, you you in terms of like unit growth and all that stuff, you'd say like, oh, it's it's Tempest, it's Gardent, it's Netera. Just assume those are three names you'd come up with. You might have others, but like let's assume those are three. So we currently trade at a discount to both those other. companies, just purely on a diagnostic basis. And yet, and yet our data business, you know, for as a if you were a tech investor, you'd be like, oh my God, that data business is worth more. Like those kind of businesses can trade at 20 times revenue or 30 times revenue. So I think what's happened is because we've been, you know, a relatively small company, like you know, we're not Amazon, and we've
29:13had an investor base that is. mostly diagnostic oriented. They have kind of discounted our data business to kind of zero. They've been like, I don't understand it. I can't follow it. So,
29:24Unidentified speaker (UNKNOWN): and
29:24Unidentified speaker (speaker_1): so we traded a discount to like, you know, some of our peers, and we certainly traded a discount to our technology peers. I mean, if you look at our data business, when you've got a data business, 500 million growing at 30%, it's a rule of 40 company, you know, at our size and scale, given the AI capabilities we have, you know, if that thing went public, it would probably trade at, you know, 20 or 30 or 40 times revenue. So, which is created the entire market cap of our company. So I think that to me is the thing that people, I think, haven't fully appreciated and one day uh I probably will, and then it will be an inflection point for our stock.
30:02Unidentified speaker (speaker_0): What an incredible. answer. I've asked that question to about a dozen other people, and that's been my favorite answer so far. That was a great
30:08Unidentified speaker (UNKNOWN): answer,
30:08Unidentified speaker (speaker_0): Eric.
30:09Unidentified speaker (speaker_1): Thank you. Thank you.
30:10Unidentified speaker (speaker_0): That's it for today's episode. I'm Austin Hankwitz. Thank you for tuning in to After Earnings, the show that brings you up close and personal with the executives behind the world's most interesting publicly traded companies. If you learned something today, do us a favor, share this episode with a friend, and don't forget to like and subscribe wherever you're tuning in from. We'll see you next time.