Q4 2025 Earnings Call Transcript

Good day, and thank you for standing by. My name is Tina, and I will be your conference operator today. At this time, I would like to welcome everyone to the Tempus AI Fourth Quarter 2025 Fiscal Results Conference Call. [Operator Instructions] It is now my pleasure to turn today's call over to Liz Krutoholow, Vice President of Investor Relations. Please go ahead.

Thank you, Tina. Good afternoon, and welcome to Tempus' Fourth Quarter 2025 and Full Year 2025 Conference Call. This afternoon, Tempus released results for the quarter and year ended December 31, 2025. The press release and overview of the quarter and our latest presentation are available on our IR website. Joining me today from Tempus are Eric Lefkofsky, Founder and CEO of Tempus; and Jim Rogers, CFO.

Before we begin, I would like to remind you that during this call, management may make forward-looking statements that are subject to risks and uncertainties that could cause actual results to differ materially. For a discussion of these risks, please refer to our 10-K and other subsequent filings with the SEC. During the call, we will discuss non-GAAP financial measures, which are not prepared in accordance with generally accepted accounting principles. Definitions of these non-GAAP financial measures, along with reconciliations to the most directly comparable GAAP financial measures are included in our earnings release, which is available on our IR page.

I would now like to turn the call over to Eric.

Thanks, Liz. 2025 was an exceptional year for Tempus with both of our businesses growing rapidly and performing well above expectation. Total revenue of our core business was up over 33% when you factor in the acquisition of Ambry, obviously much higher. If you look at our 2 main businesses, I'll start with Diagnostics. In Oncology, we had unit growth of 29%, which was and has been accelerating throughout the year.

We called out that our MRD growth rate was actually 56% quarter-over-quarter, which is extraordinary. Hereditary held up well with 23% unit growth. So all in, our Diagnostic business is accelerating and performing above expectation. In terms of Data, that business is growing even faster.

It's made up of really 2 product lines, our licensing business and our applications. Our licensing business, or Insights, was up 69% in the quarter when you factor in the onetime impact of the AstraZeneca warrant, and we're projecting roughly 40% growth this quarter. Total contract value was greater than $1.1 billion and most importantly, has been rising faster than revenue over the past several quarters. And net revenue retention was 126%, which is super strong, all things considered.

We guided to $1.59 billion, which is in line with our 25% long-term growth expectations and approximately $65 million of positive adjusted EBITDA. Our balance sheet is in great shape. Our products are resonating. Our AI advantages are continuing to take hold.

So all in, we're poised for a phenomenal 2026. With that, happy to take questions.

[Operator Instructions] And our first question comes from Kallum Titchmarsh with Morgan Stanley.

Eric, I wanted to zoom out from the financials. I'm sure that will be covered and go a bit broader. The markets are a bit anxious around AI and how value is getting distributed within that ecosystem. And we're now obviously seeing kind of traditional AI players push into the healthcare sphere.

So I'm curious how you feel your position is protected on the data side? And I guess, to that point, why you expect the large pharma companies you work with, particularly within Insights to keep coming back for more? The Q1 guide is obviously strong, but I'm kind of looking on that. And then finally, just on that point, I'm interested whether the feedback you're receiving from these customers suggests that they're getting better at what they do because of what you're providing them.

Just any sense of success stories and what you're hearing on the ground would be appreciated.

Yes. So I mean, I think the most interesting business models, I believe, surrounding AI, in particular, large language or large multimodal models, really center around access to proprietary data to train models and proprietary distribution once you have a model that generates insight. So as we've talked about historically, Tempus is uniquely positioned in that. We have both of those at scale.

We have over 450 petabytes of connected multimodal data, which flows from our Diagnostic business, which has real-time insights, real-time connection to outcome and response, is able to track patients longitudinally, rich molecular data, rich imaging data. So we have this really unique proprietary data set that you can use to train AI, to train models. And then once we generate insights or some kind of contextualization that we want to put in the hands of a doctor, because we're connected to more than 5,500 hospitals, because we have more than 8,500 regularly ordering oncologists and thousands of other physicians and other areas, we can deliver these insights in real time as part of routine clinical care.

So that's what makes us unique that if somebody wanted to replicate our data business, they'd have to go reproduce all that real-time data, which is quite hard to do. You have to enter into contracts with providers. You have to get the data, you have to work through IT issues. You have to structure the data, you have to build technology to make the data useful.

It's an enormous lift that we've been on over the last 10 years. And I think because of that, because of the work we've done to build the data pipes, to harmonize and structure the data, to build technology that wraps around the data, we just have a unique offering and other people have been unable to replicate it. And so if you look at our -- the scale of our data business, I mean, a few years ago, people thought we couldn't hit $100 million of data revenue. And we're now 4x that and projecting to grow 40% even at this scale.

We have 126% net revenue retention, which means, on average, our clients are ordering significantly more year after year than they ordered the year before. And all kinds of proof points. Our largest clients continue to re-up. Those contracts get extended.

We've announced a bunch of them over the last several years. And it's because we're demonstrating real -- our data is demonstrating real value where these clients are able to use our data and our technology to be more intelligent about what assets to go forward with, refine their early-stage discovery projects, design more intelligent Phase IIs and Phase IIIs, recruit the right populations at the right time and get their drugs approved and in market faster. And that is why the data business is just kind of having a moment and the growth is actually accelerating.

Your next question comes from the line of Subbu Nambi with Guggenheim.

Earlier in 1Q, you launched Paige Predict. Given you previously discussed that you don't expect this to contribute meaningfully to revenue this year, can you discuss the strategic value of the added capability when samples are QNS? How often is this case with xT and xR? And a quick one for Jim.

Jim, ASPs are expected to reach over 2,200 in the coming years, but what are you expecting in '26 guidance?

Yes. So I'll start with Paige Predict. So we have -- and I think I'd encourage you to read the letter. We try to spell some of this out in the letter we published.

But we've long talked about how there's enormous benefits to the contextualization of these diagnostics and the technology we wrap around them. And it's not just about having the next version of an assay or running a study that produces some kind of wet lab improvement, like that only means so much. If you look at our growth and the fact that our growth is this strong at this scale, it's in large part driven by the fact that we just have a technology advantage that makes physicians want to order our products because they get greater insights from those products. And those insights, like if you look at just a few that we called out, one is Paige Predict and other is what we call our Immune Profile Score, and there are dozens of others.

But if you look at Paige Predict, here's an example of we've built technology that at scale, allows us to digitize pathology slides and generate insights from those pathology slides by which we can predict mutations that exist that will show up when we do the next-generation sequencing. The benefit of having a system where you're sequencing tons of patients and following mutations and you're digitizing pathology slides and tracking those is you can begin to correlate these things. All next-generation sequencing has some amount of error. It's just inherent to the process of using Illumina and other sequencing companies where you don't get 100% output when you sequence a patient.

So some percentage of the time -- it's a small percent, but some percentage of the time, you have results that can't be returned to a physician. Being able to digitize the pathology slide and render insights even when sequencing fails just makes our tests a little better than somebody else. The fact that we can also render those results in hours, makes us a little faster to deliver those insights. So the same thing if you look at our Immune Profile Score, we're able to look at lots of different multimodal data we generate, could be digitized pathology slides, could be transcriptomic data from RNA, could be DNA data and refine what have historically been traditional biomarkers like tumor mutational burden or others.

And each of these insights that we can generate, again, it's not that one of them alone is the reason that a physician would use us, but they just keep stacking up. And if you look at the foundation model efforts that we're engaged in, that's going to do nothing, we think, but accelerate that dramatically. And so to the extent today that we're X percent better than somebody else because we can generate Y percent more insights, you should expect that to grow quite a bit over the next several years.

And then in response to the ASP question, so ASPs in Q4 were around $1,640. That was up about $40 quarter-over-quarter. As you noted in our investor deck and kind of what was discussed at JPM is we think that there's about $500 or greater than $500 of upside to ASP based on the current mix driven by a couple of factors. One is the continued migration of xT CDx from the LDT version to the FDA-approved version. As we previously stated that we plan by the end of 2026 to be exiting with the vast majority of volume on that FDA-approved version.

So that's kind of the biggest impact from an ASP perspective in 2026. We also announced that we've submitted our xF, which is our liquid biopsy to the FDA. That will unlikely to have much of a '26 impact from ASP, but as we get into '27, we'll start to contribute. And then lastly, there's still some upside from commercial payers as we kind of chip away at those. So that's kind of the build -- those 3 initiatives kind of get you that upside of $500 with xT CDx being the biggest driver, and that will play out over the course of 2026.

Your next question is from the line of Ryan MacDonald with Needham & Company.

So my question for Eric, you just sort of talked about the foundation model and how that can exponentially help the stacking initiative and development of additional algorithms, different additional modules over time. Can you just give us an update on where things stand on the development of that foundation model? I think you mentioned last quarter, you're hoping to have the first version of the model ready in the first quarter of '26 here. And so just curious if that's launched yet, how it's performing relative to expectations?

And then for Jim, just curious in terms of -- you talked about ASPs just now, but just curious how you're thinking about sort of underlying assumptions for volume growth across Oncology in particular, but Hereditary as well in the '26 guide?

Yes. So I'll start. So the foundation model had a deliverable in Q1 where we had to hit certain benchmarks that AstraZeneca had established. We think we've hit all those benchmarks.

We've submitted all that to AZ. They're testing the model now, but we feel great about the model's performance. And so those efforts will go on. We've also -- that was a particular cluster we set up of, about a little over 1,000 H200s dedicated to that Oncology foundation model.

We've also procured a second cluster of more than 500 GB200. So in terms of actual compute power, it's greater than the first cluster. And we're running additional models internal, not just in oncology, but across all of our data because we have enormous amounts of radiology data and pathology data and cardiology data, neuropsych and so on and so forth. So we're incredibly long on the value that these models are going to deliver, and we're doubling down on those efforts. We think they'll be catalytic both to our diagnostic business as we keep dropping insights into our tests that make our test smarter and better than others.

And so that should be an accelerant to growth. And then we also intend to propagate these insights through our data business so that our clients get more value.

Yes. And then in terms of the kind of underlying volume assumptions in the Diagnostics business, Oncology, as Eric mentioned, had 29% growth in Q4. The first quarter is off to a good start, and so we're not seeing kind of that pace slowdown. On the Hereditary side, volume growth was 23% in Q4. As we've talked about previously and what we highlighted in the letter is we do anticipate that continuing to moderate as we lap some of the share gains that they had. There likely will also be some lumpiness in the Hereditary growth rates in 2026. And so in the letter, we've called out kind of this high teens longer-term growth rate.

It might be a little bit lower in Q1 and then kind of pick up throughout the year. There will probably be some lumpiness, but we think that, that high teens is still achievable. So Oncology, again, we feel really strong about where the Oncology business is. And on the Hereditary side, again, it was -- we were anticipating some of the slowdown given the lapping of some of the share gains.

The next question comes from the line of Mark Massaro with BTIG.

Congrats on a strong year. So I think in your letter, you talked about MRD volumes came in around 4,700 tests in the quarter. And I think you made a reference that, that level could have been 20x higher if the entire sales force had been selling it. So am I understanding it correctly that approximately 5% of your sales force was selling MRD in Q4?

And is the right way to interpret this that if everyone sold it, it could be 20x higher in the Q4? Or is the 20x higher more of an aspirational longer-term outlook?

Yes. I mean -- so the -- obviously, it's just -- it's a hypothetical, so it's impossible to say what could have happened in Q1. We were simply highlighting the really unbelievable strength of our MRD offering on -- the main vector being we ran 4,700 tests. It's 56% quarter-over-quarter, not year-over-year, but quarter-over-quarter growth, and we are highly constraining this effort. And so yes, we have a very small percentage of our cumulative sales force that is currently selling MRD.

And if we were to let everyone sell it and completely unblock it, it could be 20x higher. So now whether it would be, obviously, you only know that when you unblock it. But our point is, the growth is just really amazingly strong with a highly constrained sales effort. Now we will eventually unblock that sales effort.

It's a function of reimbursement. It's a function of the appropriate timing, but we intend to, over time, ungate this and be in market fully. It took other companies that have really strong reimbursement like, for example, Natera, it took them quite some time to establish coverage in a broad enough way that this made financial sense. We're fortunate that we can kind of ungate this in an intelligent, appropriate manner and not kind of reap financial havoc.

But what we are calling out is when you look at our diagnostic platform, which is obviously broadly connected for hereditary profiling, broadly connected for therapy selection, both in solid tumor profiling and liquid biopsy. When you look at the number of EHR connections we have and the number of integrations we have and the number of feet we have on the street, deeply embedded in the workflows of such a large percentage of the U.S. oncology market, I would suspect when we ungate this, we will become a very large MRD supplier.

Your next question comes from the line of Kyle Mikson with Canaccord Genuity.

It's [ Darren Kirstein ] on for Kyle Mikson. So just taking a step back, in Oncology, I was wondering if you could provide a bit of more clarity on your tests. You briefly touched on ASP and volume dynamics, but could you just kind of walk through what each of the main growth drivers will be for xT, xR, xF, xH and xE in 2026 and beyond?

I don't even know how to answer that question. It's a fairly broad question. But I would say, in the letter, we call out, I think, our main platform advantage that provide -- that has been kind of fueling our growth. That platform advantage, the fact that we contextualize diagnostics, the fact that we're marrying clinical data with molecular data, the fact that we have such a comprehensive profile, the insights we can generate by virtue of that, that advantage exists in xT, which is our DNA profiling.

It exists in xR, which is our RNA profiling. It exists in xF, which are liquid biopsies. It exists in xH, which is our heme offering, which, by the way, we have a whole genome heme offering that goes live this year. And it exists in xE, which our whole exome offering.

So it's across our entire platform. So it's not as if we've got like one driver driving DNA and another driver driving RNA. Our core technology advantage drives the growth of all 5 of those assays.

The other thing I would add is obviously, the market itself is also growing. And so amongst our peers, everyone is experiencing kind of healthy growth rate. So clearly, sequencing is becoming more prevalent amongst our ordering physicians and ultimately patients. And as Eric noted, kind of that data advantage is what allows us to kind of capture additional market share.

Yes. With us, obviously, in solid tumor growing, it looks like at this point faster than others.

Our next question comes from the line of Casey Woodring with JPMorgan.

Can you walk us through what the guide embeds for data and services revenue in 2026? I know that you pointed to $350 million of current TCV being tied to '26. So can you just talk about the visibility into in-year bookings to get to that guide and the timing around that? And then maybe as just a follow-up, can you split out the guide of the 40% growth you're assuming in data in 1Q? Maybe just talk through how that will shake out across insights and trials and any contribution embedded from the current foundation model with Pathos and AstraZeneca?

So the -- as we called out, I think, during the JPM conference, the bookings have been so strong that we start the year with greater visibility into the 2026 revenue build than we've ever had by a long shot. So we called out, again, at JPM that it is normal for us to generate about $100 million of revenue within a given year from bookings in the year, meaning we -- somebody wanted data and we delivered it within the year. So the fact that we have such a high percentage of our revenue already committed for 2026 means that we expect to be doing our best to manage the growth of the data business because it has just such systemic growth drivers going into 2026. And that's directly a correlation between bookings and revenue.

We just have greater than $1.1 billion in the tank. A bunch of that applies to 2026. And so we just are starting the year super strong, and we just have got crazy amounts of demand for our data products. Feels like we are in a unique spot in that we're just pulling further and further away from the competition. So that business just is having a moment.

And as it relates to like how the rest of it stacks up, the vast majority of our data and apps is data licensing. It represents the biggest chunk of it. And so everything else is kind of relatively small, but it all adds up to the other piece, and that's our clinical trial matching business, TIME, our care gap product called Next and a few of the ancillary products that are connected to that.

Our next question comes from the line of Douglas Schenkel with Wolfe Research.

This is Colleen on for Doug. So you delivered high 20s clinical oncology volume growth exiting 2025. How should we think about the durability of that volume growth into this year and next? And how should we be thinking about liquid versus tissue CGP growth throughout this year? Any color you can share on repeat testing with xF?

And if we should continue to think about xF being about 25% of total clinical oncology volume going forward?

Yes. So I'll start and then Eric, you can chime in. So in terms of the 29% growth and how that stacks up, again, as I mentioned before, kind of we're off to a good start in the first quarter here. And so we don't see a massive slowdown in the Oncology growth rates.

Obviously, you're getting to larger and larger scale. So there -- it's tough to continue growing at the same rate, but we're off to a good start. In terms of the breakdown, outside of MRD growing dramatically faster than the rest of the portfolio, we see strong growth across both xT and xF. XF may be growing slightly faster than solid, but not -- there's not a dramatic variation there.

And we don't see kind of that trend -- that's a trend that we've seen for quite some time. So we don't see a lot of variation in kind of the product mix as we look at '26.

Yes. And I'll just add to that. Like so we -- our guide implies 25% growth year-over-year. So we've called out that our Hereditary business is growing slower.

It's kind of like for the year, we think kind of high teens, mid- to high teens in that range. And we have a few other businesses that we historically have called out are also not growing. Like, for example, we deemphasized -- we have a little relatively small, but maybe $20 million CRO business that we don't spend time on. Things of that nature.

So you have a little bit of revenue with Ambry -- a lot of decent amount of revenue that's growing kind of significantly less than 25%, which means that our data business and our core Oncology Diagnostic business are growing 30-plus percent, right? And just as we told the world, we expect it to grow at 30% or so last year in our core business and ended up growing at 33%, I suspect something similar this year, right? You can -- if you do the math, we're going to be growing in the roughly 30% range plus in those businesses. So they're super healthy.

Liquid is growing a little faster than solid. So that's been a long-term trend for us. We don't disclose breakdowns of each and so on and so forth, but liquid is growing a little faster. Both are super healthy.

And there's just no one driver of the growth. It's not as if like -- it's not as if repetitive testing or concurrent testing or this kind of testing or that kind of testing or having an outsized impact. We're just seeing really good solid growth. We're seeing really good liquid growth, slightly better, but that's been a long-term trend.

And more and more people want the benefits of tumor normal profiling, more and more people want the benefits of DNA and RNA, more and more people want our connected platform that's intelligent. So our unit growth in Oncology is really strong and showing no signs of slowing down.

Next question comes from the line of Andrew Brackmann with William Blair.

Eric, it's sort of been just over a year since you closed on Ambry. And if we sort of go back to when that acquisition was announced, sort of a big part of that thesis was really around the data that you would be able to sort of generate across the entire drug oncology testing spectrum. So can you maybe just sort of talk about the acquisition in that lens now that it's sort of been a part of the company for some time, just sort of what you're seeing in that regard and how that's led to growth in the data business as well?

Yes. So I think we called out multiple reasons to acquire Ambry. The first by far was to broaden the comprehensive nature of our testing compendium. So when if someone said to me, why did you buy Ambry?

I wouldn't say data. I would say the #1 reason we bought Ambry was they had a very strong hereditary profile. More and more of our clients want a comprehensive solution. I think we've said this historically, I very much believe that over the long term, when it comes to treating cancer patients, it's going to be like e-commerce shoppers going online.

I don't go to one e-commerce site for books and another for clothes and another for consumer electronics and so on and so forth. I go to Amazon because they have kind of everything I need in one place. And I believe that's going to be the case as it relates to sequencing. So more and more providers want somebody who can help them manage risk, help them treat patients once they've been diagnosed and help them monitor those patients post treatment.

And so we want to have a broad menu. That said, we're also seeing a trend of more and more of our provider partners and certainly, obviously, to the benefit of patients, wanting to contribute de-identified data to platforms like ours to be used to accelerate research and to accelerate drug discovery and development. This is a very big trend that I don't see stopping. We still have 600,000 people a year that die of cancer.

We're not making nearly enough progress as it relates to eradicating that. And so I think you're seeing a movement among institutions that are saying, we need to help stop the waste to make sure these patients get better drugs and get them in market faster. And so we see that as a constant movement and a benefit of the kind of data we collect and then on a de-identified basis, take to market.

Our next question comes from the line of Dan Brennan with TD Cohen.

Can't hear you.

Sorry about that, Eric. Just maybe one on MRD and on Insights. Just on MRD, any update on the first-gen CRC assay? I know it's sitting at MolDX.

Just any color there? And on the next-gen tumor-naive assay, have you guys discussed kind of what type of performance advantage you would expect to get out of that? Obviously, you're filing for 2 tumor types this year and then you have another 2 that you mentioned in the letter. Just wondering what kind of performance enhancements do you think that could offer versus the existing tumor-naive landscape?

Yes. So we're back and forth with MolDx now. We didn't call out when that gets resolved, I mean, because I don't control MolDx. I mean, so it's possible that we have reimbursement in a month, and it's possible that it takes longer.

I don't have any idea. We also didn't call it out because it's just not that relevant to our current MRD offering, which is, I don't know, like 95% tumor-informed. So because we're largely in market with a tumor-informed product in partnership with Personalis, it just represents the vast majority of our current market penetration. And this particular product, which I do expect will be reimbursed by MolDX, is just not going to be a needle mover because the goalposts keep moving.

And so what's happening is tumor-naive products have to keep getting better and better to really compete. In CRC, I think it's going to be quite some time before you have a significant amount of the volume moving away from tumor-informed to tumor-naive. So I think the naive market is either episodic or it's for those patients where, for whatever reason, they can't get tissue. But tissue is pretty pervasive in colorectal cancer.

And so I think tumor-informed wins the day in that subtype for a while. There are other subtypes like, for example, lung, where tissue is more sparse, where I think tumor-naive products can do quite well. We realized that we just weren't getting the performance off the first version of our assay. So instead of like -- we're running a ton of studies to collect samples, I mean, a ton.

And so we decided to kind of pivot and begin working on the second generation of assay instead of like burning those very precious study samples on an old version, we wanted to move to the new version. So we kind of pivoted and we were fortunate -- everything we do takes into consideration this notion of like a comprehensive portfolio. So we were fortunate that we had a tumor-naive product that was just kind of doing super well in the market, more than -- giving us more volume than we candidly want or need. And so we didn't have to kind of overly push on the accelerator for tumor naive.

But the second version is coming along well, and I suspect at some point, we'll have a really nice assay market.

Next question comes from the line of Mark Schappel with Loop Capital Markets.

Eric, the start of the year is typically when companies adjust their sales organizations and their go-to-market strategies. First, if you could just give us an idea whether you've implemented any meaningful changes to the sales org this year? And then as a follow-up, maybe you could just sketch out what you believe are the firm's kind of key investment priorities for the coming year.

In terms of the sales force, we've made no big moves to reorg the sales force. We did that, obviously, early '25, and we announced the impact of that. And so the good news is we're long lapping that. And in terms of our priorities, they remain the same, to bring the benefits of technology and AI broadly to diagnostics and make sure that every decision, whether that's a decision in clinic or a decision for research, is data-driven.

And I would expect us -- given that we're growing so rapidly in our business accelerating, I would expect us to stay the course.

Our final question comes from the line of Bradley Bowers with Mizuho.

Just wanted to get into some of the ASP outlook and maybe the gross margin implication. Obviously, a lot of upside here with the 2,200 test outlook. I just wanted to kind of hear about what the expectation should be for gross margin. It assumes a big lift.

I mean, if you assume that the costs hold and do some math, it kind of gets towards gross margin in the genomics side of 70% plus. What could you say about that progression? And is that 1:1 with ASP? Or are there some offsets we should be considering?

Yes. I think when -- obviously, any time ASP increases, that would lead to an increase in gross margin. I think our -- we've long kind of taken the approach that as ASPs kind of increase or as cost of sequencing come down to reassess kind of how much you kind of increase the size of panels to generate more data. Obviously, that's great for patients and great for doctors and all that.

And so that's something that we do on kind of an annual basis. We're not -- given the fact that we have kind of these 2 businesses, Diagnostics and Data, we're less reliant on maximizing gross profit within the Diagnostics kind of product line as some of our peers may be. That said, as ASPs increase, we would anticipate seeing some increases in gross profit, but always balancing to make sure that we're bringing to the market the broadest panels possible because it obviously has a bunch of downstream implications.

There no further questions in queue. I will now hand the call back over to Liz Krutoholow for closing remarks.

Thank you all for joining us today. We look forward to updating you again next quarter. Have a great day.

Thank you again for joining us today. This does conclude today's conference call. You may now disconnect.