Tempus AI, Inc. (TEM) September 15, 2026 Conference Transcript

Perfect. I think we can get started. Callum Tishmarsh here from the Life Sciences team at Morgan Stanley. Welcome to day two of the Global Healthcare Conference. Really pleased today to be joined by Eric Lefkowski, Tempest's founder and CEO. Thanks for being here, Eric.

Thanks for having me.

A lot of news to discuss, but maybe we can just hit on first the second quarter results. I think very strong across the board, broad-based performance. Anything that surprised you within that, maybe just go through the good. anything that surprised you to the bad, you know, maybe just level set up, you know, then we could dive into some specifics from there.

Yeah, I think both the quarterly performance and I think the year performance has really been anchored around two main themes. One is the core strength of our diagnostic business, in particular around comprehensive genomic profiling and how kind of durable those growth rates, those unit growth rates are, both in terms of solid tumor profiling and liquid biopsy. So you have this, like, you know, very strong therapy selection business in terms of, like, demand that also has rising ASPs, and so that's kind of one of the big bellwethers of our business. It's the biggest part of our business. So even though MRD has high growth rates, higher growth rates, it's the biggest part of our business. So our diagnostic unit has been buoyed by really strong demand that we don't see slowing down. And on the other side, the data business continues to perform really well. I think our data licensing business grew to about 36% or something in the last quarter. So really, you know, continuing strong growth, and we've had something like three quarters in a row of more than 100 million of TCV. I think last quarter was around 200 million of TCV ad, meaning our bookings for that quarter, people signing new data licenses, or, you know, 200 million or whatever it is. And so really strong demand that I think is starting to really be catalyzed by migration to AI solutions by biopharma. Every time you turn around, they're announcing a deal to bring in NVIDIA chips or cut a deal with Anthropic or OpenAI for inference and compute. and they just, you know, we're the fuel that makes a lot of that spend intelligent. And so we just have seen kind of a frenzy of demand.

Amazing. You've talked about expectations for at least 25% growth for the business over each of the next few years. Maybe just go through, you know, what has given you that confidence where you have the most visibility across the different business lines to that, you know, sustained growth rate.

Yeah, so we've long said that we would rather have a business growing at 25% for a decade than 35% for a few years. So we just kind of architect ourself around long-term growth. On the diagnostic side, that business is really going to be benefited by not just this robust demand in terms of therapy selection, but really significant rise in ASP. We have two big levers that are going to work in our favor, I think. One is our main solid tumor assay got FDA approved, not just tumor normal but also tumor only, which allows us to migrate that entire platform to ADLT pricing. That should add somewhere in the neighborhood of $80 to $100 million of kind of incremental ASP lift and profit and free cash flow and all the things that comes with it. if you didn't reinvest next year alone. And then our second largest product, our liquid biopsy, which is also quite large in terms of units, that is in front of the FDA now. We expect that will be approved at some point. That also will have EDL-2 pricing. And the kind of benefit to us is that that assay is kind of roughly similar in size to Garden's larger assay that just got approved, and they signaled that they're getting ADLT pricing or they expect ADLT pricing north of $8,000. So we had historically thought that assay would be priced at, let's say, $5,000 or $6,000, but now it looks like it'll be priced closer to $7,000 or $8,000. So that should add something like $250 or $300 million of lift on an annual basis once that's approved. So you just have these really big kind of revenue drivers that will accrue to the benefit of the diagnostic business in therapy selection. And then we have the same thing happening in MRD. We have even stronger unit growth rate, rising ASPs, and so that business for the next, you know, whatever, three to five years will likely just have really robust revenue growth. At the same time, as I just mentioned, the data business, you know, you have record demand and you have, this migration to people leveraging these large language models for their benefit is not going to slow down or stop. In fact, it's probably going to accelerate because these people are just now starting to leverage those technologies. And so I think the data business also has really long-term visibility. And so if anything, if you said to me today, are you going to under or over-deliver that 25% growth rate, you've got to bet it all. every day of the week and twice on Sunday we over-deliver.

Amazing and you didn't have to deal to acquire Personalis. You've obviously been involved with the company for some time before that commercially. Maybe let's just start with the rationale behind that acquisition and the path you see forward for MRD across the market.

Yeah, so I mean we have long thought that Personalis' assay was best in class. It's whole genome-based. It has incredible sensitivity and specificity. Its limits of detection are really extraordinary. So we've long thought that it was a best-in-class assay. And we had cut a deal with Personnel years ago to be their exclusive distributor in the U.S. or in the world.

I'm not sure if it's U.S. or world-based in the U.S.

And so, you know, we thought that that was a great way to enter the market, and the deal was structured where they were paying us basically a sales and marketing fee to distribute the assay, whether or not they got paid or not, and so for the last several years, this deal was largely in our favor, right, because we were getting paid about 400 bucks per test. They weren't getting paid because they didn't have reimbursement approved, and so it was just in our favor, and I had been pretty vocal that, like, why would you want to change that. You have to wait until the things flip. When they got lung and breast and recently IO approved by Moldex, it became apparent to us that we were getting close to that point, that their ASPs would rise pretty dramatically, and that there would come a time in the next few months, few quarters, whatever, where their ASPs would actually be higher than they're paying us. So we began having conversations. This is all kind of detailed in the S4 that was filed. We began having conversations. Those conversations were accelerated when another party showed up and made an offer to buy the company. So, you know, whereas we may have bought them toward the end of the year or maybe early next year, we accelerated it by a few months because there was all of a sudden some activity. But it makes a ton of sense in that it's a best-in-class assay. The unit growth is really strong. We're already distributing it, so it's already a part of our portfolio. And they will have ASPs that will rise. They'll get to $1,000. They'll get north of $1,000, similar to Natera. And it will become a really strong financial product. And so it makes a ton of sense inside our platform.

Interesting timing as well with the Merck, Moderna, V940 metanoma vaccine data. It seems like you had some involvement there in kind of excluding what Personalis was doing. So can you maybe just detail how involved Tempus is and then how that Personalis addition would kind of come into the picture?

Yeah, yeah. So basically, Personnelis was selected to be the sequencing partner for that clinical trial some time ago, and they'd been performing the sequencing for that trial. At some point, Moderna and Merck felt they needed a national partner for the rollout, and so they ran an RFP that I think most people participated in on scale. And they ran an RFP, and we won that RFP, which we put out on social the morning the news came out.

We won that RFP.

So once the product's FDA approved, Tempest, unrelated to personalis, will be handling a significant amount of the kind of revenue and volume associated with that approval. In terms of who does the sequencing, you know, those details are still being worked out in terms of what percentages. Personalis may do some sequencing, we may do some sequencing. If we're one company, it won't really matter. But in terms of, like, If you were to look at this like pizza slices, we were getting the majority of the pizza slices anyway by virtue of winning that RFP. It's a really important program because it's really the first time that I'm aware of where the sequencing is actually a component of the manufacturing process and product. So it's unlike kind of any other form of companion diagnostic where somebody could get a drug approved using foundation medicine, but Keras or Tempest could do the same sequencing. Or Garden could be part of the Serena 6 rollout, but I can sequence for ESR1 as well, and you can prescribe the drug. Here, if we don't sequence you, you can't get the drug, and so you can't use somebody else. The manufacturing process is approved by the sequencing we do and a whole bunch of other logistical things we do as part of the process. So Merck and Moderna needed a partner that could handle that kind of national scale, redundancy, can't go down, all that good stuff. So I think it now works out great that we're going to be one company, but that process kind of ran its course the end of last year. So for a long time, we've been a big beneficiary.

And perhaps this is more my job, But have you worked to size that opportunity up for what this could be? You obviously have melanoma today, and obviously other indications are perhaps coming through the line. So how should we be thinking about how big this could perhaps be?

Well, it's okay. I mean, obviously I don't know because I don't know how the other trials will read out, and I don't know how big this will ultimately be. My best guess is that it's quite big. both because the performance of this drug, its ability to essentially allow patients that don't respond to immunotherapy to respond is pretty extraordinary. I suspect it will work quite well in multiple subtypes and that it will be a very big drug. So there's revenue associated with the work we do as part of those clinical trials, of which there will be, you know, many, and that's great. And then there's the ancillary benefit that, you know, one of the reasons our unit growth rate is so strong at Tempus is we just have given our oncology partners, like, more and more reasons to work with us. We said this years ago. We said you people are all kind of chasing performance of assays if that's why you make decisions, but it's never why you make decisions. Oncologists make decisions because of a whole bunch of other reasons, including ease of use, logistics, administration, contextualization, like do I get the information I need in a timely manner in a way that's better than other people? It's the same reason we shop at Amazon and we don't shop at eBay. If you go back in time 10 or 15 years, eBay and Amazon were kind of neck and neck or 20 years or whatever. And today that's just not the case. We all go to Amazon and we don't go to eBay. And it's all those kind of logistical benefits. And so this program will be another reason to use the Tempest because why would you want to sequence with somebody else? And then if you want to get this drug, which is a part of your mainstay therapeutic options, you then have to kind of resequence with us. So I suspect that we'll have both real-term revenue benefits as these trials roll out, and then ancillary benefits as more and more people just choose our plan.

And then just reading through the S4, one of the things that caught our eye was just the differences in how Personalis and Tempus was underwriting the kind of projections for the two businesses, for the one business. So maybe just talk through where that difference was. I think Personalis came out with $758 million of revenue in 2030, Tempus $333 million. So why do you think there is a big difference there?

Yeah, it's almost entirely ASP. So our expectations of unit growth, I think, are both quite robust. We have a robust pro forma. They have a robust pro forma. We expect the Unicro to be really strong. You can make different assumptions about how ASP rises, what percent of the orders are for, like, IO response, how fast do they get, for example, CRC approved. These would radically change your projections. We have kind of tried to take the approach of being conservative with those particular estimates, and so we just don't feel any reason to kind of like be overly aggressive but the short answer is their forecast could be better than ours I don't know yeah makes sense and there were some unique features of the deal Purson is currently trading above the deal price how do you think about that? I mean I think it makes no sense you know the and I don't know what it's trading at and I didn't look at it yesterday but there was a while it was trading like $17 or $18, which made no sense. I mean, first of all, you can read in the S4 that it was a competitive process. There were all kinds of people contacted. They ran a full and robust process. And the spread between the bids was only like $0.75. I think somebody was at $17. We were at $16.25. So it wasn't even like this. There's kind of no logic to think there's some kind of magical price out there that's much bigger that I don't think exists. On top of that, you know, it's the idea that somehow you would be trading above the price when you have Tempest as the largest shareholder, obviously voting in favor. Merck is the second largest shareholder, also voting in favor. I think ARK is the third largest shareholder, and they're kind of a huge Tempest fan. It just doesn't make a ton of sense to me That in and of itself could be 35% or 40% of the vote, and these things never get 100% of voting. So I don't see any topping bids coming. There's no kind of logic there. We already have significant shareholders voting in favor of it, so there's no kind of rationale to be buying their stock at a significant price above where it's going to close. The only counter to that would be some kind of nefarious short-term trade. Like you somehow think because there's an exchange ratio that's calculated a few weeks before closing that you could short us or buy them or whatever. But those trades, in my opinion, are kind of riddled with risk that I wouldn't be taking.

Makes sense. Maybe we can transition onto the genomics business and focus specifically on oncology for a little bit. I think the investor day, you described three key trends that will drive therapy selection, you know, in the years ahead. Physician penetration, earlier stage testing, and more comprehensive testing as well. So when we look kind of three, four-plus years out, how do you think that market evolves? Obviously, it's a more established market in the oncology space, but we still obviously have plenty of room to run, I think.

Yeah, I mean, I think there are some fairly good studies that came out that the kind of penetration rates are in the roughly 50% range for kind of later stage cancers, you know, stage 3, stage 4, metastatic, high risk. And I think that's probably right. In addition to that, we're going to be sequencing patients earlier. where more and more biomarkers will show up. The evidence behind concurrent testing is extraordinary in terms of the benefits you get from doing solid tumor profiling and liquid biopsy. The benefits of RNA are extraordinary in terms of enhanced fusion detection. So I suspect, and then the benefits of MRD are obviously kind of also well-known. So I think we will be in a cadence of broadly sequencing newly diagnosed cancer patients, you know, probably all stage two plus even certain stage one categories like liver and pancreatic, and then broadly monitoring patients post-therapy. And I think over the next decade, I can't see anything that's going to slow that down. And then I think you'll start to see other disease areas that begin to catch up because they've seen the benefits in oncology. certainly rare and undiagnosed disorders, certain immunological conditions. So I think molecular profiling for therapy selection and monitoring post-therapy will be a growing space for a while.

I know there were some headlines earlier this year, Crush RFI that came out that freaked a few investors out. Incoming R-way has kind of cooled off since then, but any way you're thinking about the durability of reimbursement for therapy selection, you know, mRNA, sorry, RNA, You know, your kind of dual XTR, XR orders. How are you thinking about that, I guess, in the years ahead?

Yeah, I mean, so on the Medicare-Medicaid side, reimbursement seems to be quite stable. There are kind of took a long time to establish national pricing, which has been like kind of fairly stable for the past, let's say, I don't know, four or five years now at this point. So I think the space has pretty good pricing in terms of solid tumor profiling and liquid biopsy. We obviously benefit from national coverage policies that are in place by virtue of the fact that our main assay is FDA-approved, our liquid biopsy will be FDA-approved, so we're even out of some of that at that. We'll have ADLT pricing and be part of national coverage. But I think the space has quite durable reimbursement in terms of Medicare and Medicaid. It took a long time to establish it. I don't see any material pressure coming anytime soon. On the MRD side, right now those assays are basically being reimbursed by Moldex. I suspect the other MACs will start to pay for those tests as well, because that tends to be a... So I think over the next, let's say, three to five years, you're likely to have very good, very stable reimbursement from Medicare-Medicaid. I suspect over time, commercial payers will start to pay more for these tests. They're still radically underpaying. I think others have said the same thing, whether that's Gargant or Natera, so I think that's a pretty common supposition at this point, which means I think you're going to see margins in this space from the top providers that get extraordinarily high. One could argue too high. But long-term, I think that will start to normalize a bit. You know, if you fast-forward 25 years from now, you might see margins in the 60% range. But I would not be shocked if over the next decade you see margins in the 80% range.

Amazing. We have two of those highest-volume tests, XT and XF, going through pretty, you know, significant regulatory and reimbursement upgrades. I think perhaps it's underappreciated in the story, at least from my conversations. So maybe just remind investors how important that is, the kind of uplift we could see into next year.

Yeah, so as I mentioned a little while ago, so, you know, we have two main products in therapy selection. One is our solid tumor assay, and one is our liquid biopsy. The solid tumor assay was divided in really two parts, tumor normal and tumor only. Tumor normal represented a minority of the volume. We got original FDA approval for tumor normal, and that had ADLT pricing at $4,500. And so we weren't able to migrate our platform fully to the ADLT pricing. A few months ago, we got tumor-only FDA approved with similar ADLT pricing, identically ADLT pricing. So effective January 1, we'll be running all of our solid tumor assays under that pricing. So you have this kind of immediate step up for more than half of your solid tumor portfolio from like $29.23 at list price to $4,500. And then sometime toward the back half of $27, we expect both approval and pricing of our liquid biopsy to be in market, whether that's Q3 or Q4, it's unknown. But at some point in the back half of the year, you'll have that. That step up goes from, I think, 3,200, which is roughly our current liquid biopsy pricing, to somewhere in the mid to high 7,000 range, likely. So that's a very significant step up. So whereas the ADLT pricing for tumor-only adds, let's say, $80 million to $100 million of revenue and margin benefit liquid biopsies, like $250 million to $300 million.

I want to hit on the hereditary business quickly. I think expectations for growth have fluctuated a little since you acquired Ambry. Is mid-teens the right way to think about this business longer term? And maybe just touch on the underlying drivers you think of this business three, four years out.

Yeah. Yeah. So, you know, we have bounced around a bit like a yo-yo on their growth rate. I think, fortunately or unfortunately, you know, we've only owned Amory for like 18 months or something. So, you know, we've had to learn a lot about how their business performs in this forecast. We could see early on that the growth rates they were experiencing, let's say six quarters ago, felt extreme to us. So we tried to call that out. And we tried to call out that it felt one time to us, that it was not one time, but not repetitive, and that it was largely a function of Invitae going bankrupt and a shift of volume from Invitae, who was one of the largest suppliers over to Amory. So I think we tried to call some of that out, but certainly as you lap it, we now can kind of fully see the impact of that. So you have a business that should be growing in the 15% range that was growing at 30% or 40% for a while that then was growing in the low single digits or is now growing in low single digits. and we begin to lap that toward the end of Q3. So you'll start to see growth rates kind of look better because we're just lapping that period of excessive growth. So I think we get back to mid-teens toward the end of this year because it's kind of almost just math, and I suspect we'll get there. Long-term, I think the business sustainably grows in the 12% to 18% range, just call it mid-teens, low-mid, high-teens. I think it grows at that range just based on current dynamics, meaning understanding hereditary risk is important. We keep, just like in cancer profiling, everyone's publishing papers looking at genes that are correlated with risk. People want to understand risk. And so this is a mid-teen grower space. I think the best estimate for the space is growing around 12%. So we should be a little better than that. If you can ever unlock what is, to me, the most insane amount of latent demand, I think this becomes a really big business, which is at the present moment we run about 2 million of these tests a year, something like that. and yet there's current coverage policies in place for about 70 million tests and we are gated by the number of genetic counselors that can order these tests and genetic counselors are not revenue generating for hospitals so they don't make money off that so you have this massive amount of people that are in categories where there's reimbursement established they're black, they're Ashkenazi Jews They have known familial risk, and yet we don't test them. So I think that problem has to get solved. And when that problem gets solved, and we're thinking through lots of ways to solve it, this could become a very big business very quickly. So I think conservatively, over the next three to five years, those mid-teens, if we get any of that right, it should have growth rates that are equal to or greater than our cancer business.

Understood. And I want to make sure we cover the data business. I think a major theme in 26 is AI-driven drug discovery. It feels like you have the head start here in the market. So what part of the AI-enabled drug discovery thesis feels credible to you? What feels a little more speculative, and what role do you think Tembus can play in this evolution?

Yeah, I mean, none of it feels speculative. It all feels very well-established at this point. We've been licensing data in oncology for probably seven years, eight years, 2018. So this has been going on for a long time. people thought our data business would never get to $25 million, would never get to $50 million, would never get to $100 million. It's obviously now, you know, way, way larger. And we have had now multiple people enter into, you know, $100 million-plus long-term data licensing deals with us, whether that's AstraZeneca or, you know, BMS or GSK or Merck or BioNTech. It just goes on and on. And I say to people all the time, our pricing works identical to AWS. You can license one file from us for a few thousand bucks. So the only reason these people are entering into long-term contracts is they want access to data and they want discounts. And so I think that speaks to when you have this many people licensing this amount of data for this amount of time. It just speaks to the durability of that business. and AI is only a catalyst to that. Our data is invaluable for understanding synthetic controls, understanding how to design a phase two and which mechanisms of action are driving response. Do you have the right design for your phase three? How do you think about site selection? How do you think about commercializing that asset given that therapies are changing? There's so many reasons to buy our data. And spending $25 million a year on our data when you can make decisions that are going to save you $200 million or $500 million. It's just a no-brainer. So I think in oncology, right now we have a significant number of people that are these very large strategic clients. I don't know how it's not almost everybody over the next three to five years. And I think that extends into biotech and it extends into other disease areas.

I think you mentioned at the Investor Day, those top 20 pharma, biopharma companies had only really just dipped their toes in the water with respect to those data offerings. So how big do you think this could become? And I'm just trying to, as an investor, think about that timeline and the cadence when this could really, really expand meaningfully.

I think it's kind of similar. In my mind, it's similar to genetics. So only the base case is a much higher growth rate. So the business probably grows at 30% just in oncology, just under current trend line. So if things just keep going as they are and we're really mostly oncology-based, this thing could grow at 30% for the next, you know, five or ten years easily, like for a while. If other disease areas really take off in big ways and we've got some, you know, big projects in flight, or if the hyperscalers choose to get into this space, which I suspect they will at scale, then those growth rates are going to seem small. because the amount of money people like OpenAI or SpaceX or Anthropic or Google or whoever have to try to use our types of data to train their AI models is extraordinary relative to the amount of money Pharma has, which is crazy because Pharma has lots of money.

Yeah, and maybe just talk through that competitive mode as well. You obviously have the data from your tests and then the identified data from the hospitals as well. So why couldn't someone come in and replicate this model that you've established?

Well, I think, look, I mean, I've been asked that question. So we went public two years ago-ish, a little over two years ago, and we began that process. We were delayed for a couple years, and then we had a testing in the water. So for five years, people have been asked that question at scale, meaning every three months for five years, people have been saying the same thing. So I guess there's a question. At what point is it like it just starts with to build the data product we built, we had to connect to thousands of hospitals. We had to enter into legal agreements, have BAs in place, build pipes, ingest the data, harmonize structure that data so that it's usable in a longitudinal format, match it to molecular data at scale, match it to digital pathology slides, the radiology scans, then build tools around that data because otherwise it's just 500 petabytes of useless data. So we've just done all that. And if you look at it, everyone who's tried to launch a data business, which is most of our major competitors, if you look at the launch of their data business, it's five, six, seven years old, have just had no traction, basically. And we continue to have significant traction. So in terms of our immediate cohort, The people like us that have rich molecular data, I think we've just outpaced them dramatically. It doesn't mean there aren't people out there with competitive products. I mean, there are. Concert AI did a deal with Keras, and people have done deals with Flatiron and Foundation Medicine. So you can buy data from lots of people today, and in certain use cases people do. They'll license Flatiron data, they'll license Acuvia. So there's lots of competition in the market, but this competition isn't affecting on any level. our growth rate or the kind of proprietary fuel behind that growth rate.

And is it fair to assume that pharma is just going to demand more and more data from you guys as we look forward? It seems like that's the obvious play, and it seems like that should flow through nicely into the growth rates into those years ahead. Anything we should be keeping in mind there in terms of the quantity of data that pharma is demanding? Anything you've seen?

You know, I think if you look at R&D budgets, and not just from biopharma, but researchers, payers, life science companies, certainly the big hyperscalers, anybody who wants to build products that advance healthcare in any way, shape, or form are going to need vast amounts of de-identified multimodal data, and we just happen to be sitting on a very large lake. It's 50 million patients, 10 million in cancer, 40 million outside of cancer. So we just have an enormous repository of data that I think is going to power a lot of this AI development. Amazing.

Eric, thank you so much.

Thanks for having me.