Tempus AI, Inc. ($TEM)

Hello, everyone, and welcome to this next session of the 25th Annual Needham Virtual Healthcare Conference. I'm Ryan MacDonald, and I lead Needham's Health Tech research efforts. And in this session, I'm pleased to be joined by Tempus AI CFO, Jim Rogers. Jim, thanks for joining me today.

Yes. Thanks for having me, Ryan.

So thanks for those for everybody who is joining us today. For those listening in, we've got about 40 minutes to go through a list of questions on a fireside chat here. But if you do have questions for Jim, feel free to put them into the chat box, and we'll make sure to get those asked and answered over the last 5 minutes or so of the fireside chat. But with that, let's jump right in. So Jim, for those who are less familiar with Tempus, how about a brief overview of the business?

Yes, of course. So we -- at Tempus, we've been spending the last 10 years kind of focused on building a platform that allows us to bring the power of data and AI to health care to positively impact patients, starting in oncology. And so we've focused on kind of 3 different categories to date. One is kind of an AI-enabled diagnostics for providers. And so we run a very large sequencing laboratory or several of them across the United States, where we process samples on behalf of ordering physicians for their oncology patients and kind of contextualize the results even further through some optical integrations that allow us to access clinical data. We also partner with life sciences companies to improve their drug discovery efforts through the licensing of the identified data. And we also developed kind of a suite of AI application tools that can do things such as match patients to clinical trials, close care gaps or some of which even become kind of algorithmic diagnostics of their own, such as our cardiology efforts, which we can get into later on.

Awesome. Yes. And you touched on it a little bit there. But in my view, the tech backbone of Tempus is a real differentiator for the business. You talk a bit more about how Tempus' combination of data, AI and diagnostics is really unmatched in the industry? And how difficult would it be for another company to replicate what Tempus has today? And how does AI play into the company's moat?

Yes. So we've been a data -- we very much view ourselves as a technology company. And so we've been focused on data from day 1. And the original idea was how do we aggregate all of the siloed data to say something insightful to a physician or to a life sciences company. And so at the same time that we are kind of building out our wet lab capabilities to generate genomic information, we went to hospitals and had conversations with them about kind of building these EMR integrations that would give us access to clinical information. So molecular information is obviously incredibly powerful in its own right. However, when you pair it with clinical information, that allows you to do 2 things. One, as I mentioned before, you can contextualize the results for the physician that is ordering the test. Give them additional insight that you can't lean from just the molecular information. That also -- that data is identified and kind of leveraged in drug discovery allow you to kind of design more effective trials and really take a different approach than what life sciences companies have historically done because they didn't have access to this multimodal kind of longitudinally updating information. And so that was core to what we built from the beginning. We're now connected to over 5,500 institutions. And so we have a pretty broad coverage in terms of integration, and that's allowed us to amass a very large data set that really is kind of the driver of the flywheel that we have is the more data that we generate and the more data that we collect allows us to generate and garner even more insights that we can embed back into the test or with our life sciences partners. And the challenge is that a lot of people have access to data. There's other sequencers that have obviously a tremendous amount of molecular data. You have some EMR companies that have access to clinical information, but it's really the marrying of those 2 that kind of sets Tempus apart and why it's important to not only get access to data, but build a suite of tools that allows you to kind of structure, harmonize that data and also tools that our partners can leverage to kind of interrogate the data itself. So we have a physician-facing portal called Lens -- or sorry, a research-based portal called [ Lens, ] where our life sciences partners can go in, view the data, build models right within the environment. And that's really incredibly important for them to glean the insights that we provide.

That's really helpful context there. So as we think about the sort of 2 revenue segments of the business, you've got the diagnostics, which is sort of testing business and then obviously, the data business. Maybe just starting with diagnostics, given sort of how large it is as a percentage of revenue. For those getting up to speed, could you just cover sort of the tests within your genomics or diagnostics segment and maybe how you think about the growth prospects across those tests?

Yes. So within diagnostics, we kind of have 2 offerings. We are kind of 2 call points. We sell to oncologists through our kind of therapy selection and MRD offering, and then we sell to genetic counselors through our hereditary offering. And the hereditary offering, largely the legacy Am business that we acquired back in 2025. Within oncology, we have a very robust portfolio where we do DNA, RNA, liquid, solid, germline. We have a tumor-informed offering and MRD through our partnership with Personalis and then a tumor-naive test that we've developed internally. And so we can really serve as kind of a one-stop shop for oncologists to meet all of their genetic testing kind of needs. In the hereditary space, the majority of that business is in oncology today, although there is a small rare diagnose disorder business that is growing. And so different call points. The growth trajectories for the businesses are different. Obviously, in oncology and therapy selection, the industry continues to grow. All of our peers are experiencing strong growth as well. Our oncology growth rates in Q4 were 29%, so very strong kind of volume growth in oncology. On the hereditary side, in 2025, we benefited from some competitor disruption. And so we had outsized growth for 2025. That began to moderate in Q4 and will continue to moderate in Q1 as we've highlighted for folks. the hereditary space is more mature in oncology. Rare is still relatively untapped. And so that represents an opportunity for that growth rate to kind of tick up. But that's kind of a quick overview of kind of the breakdown of our assays and kind of anticipated growth for each.

Very helpful. And before we kind of drive into the growth drivers within diagnostics specifically, I wanted to ask you about some recent news. Late last week, the Diagnostics group of stocks and Temps included were pressured by some updates around the crush RFI and announced efforts just more broadly from CMS to sort of eliminate wasteful spend. Can you provide a bit of color on the crush RFI and what potential impacts it could have on the diagnostics business over time?

Yes. So the crush RFI, as you just described, was it an effort aimed at kind of minimize wasteful spending within CMS. And we -- for us, the reimbursement landscape has constantly evolved since we've started sequencing patients, and we've always taken approach of having kind of multiple mitigating strategies that hedge against any reimbursement risk. We have multiple labs in different kind of MAC jurisdictions. We've been bringing assays to the FDA to kind of see coverage under the national coverage determinations, and those efforts will continue. So we believe we're well positioned to kind of adapt to any potential changes that could come in the future. And that's always been our strategy from it.

Got it. Got it. All right. And so as we think about sort of the Diagnostics business, obviously, it's P times Q. So volume, we'll talk about volume and pricing from a growth perspective. But if we start with volumes, can you talk about the drivers of growth -- volume growth for Tempus across oncology and hereditary. How much of the success in oncology is sort of market share gains versus sort of a simpler, broader market growth and sort of more patients actually just getting their cancer sequenced?

Yes. So certainly, there's -- as I mentioned, we're all benefiting from the fact that the market as a whole continues to grow. We used to say that we thought about 1/3 of cancer patients were receiving this type of a sequencing. We believe that number is probably about 55% today. So there's still room for growth in the overall market, but everyone is benefiting from the fact that these are -- this type of testing is being ordered more frequently. For us, the advantage has always been one of the insights that we can provide. And so we certainly have experienced over the last several years market share gains as a result of kind of our differentiated offering. As I mentioned, we grew 29% in Q4. That has accelerated over the course of 2025 largely due to some sales force realignments we did in the previous year that had caused some disruption. By the time we got to Q2 of last year, we had certainly kind of see efficiency kind of take off from the sales force, and that continued over the balance of the year. On the hereditary side, as I mentioned, the hereditary cancer market more mature however, we still don't sequence -- if you look at the number of people that should receive hereditary cancer screening, it's still a relatively untapped market, but you have to be able to kind of see that pull through to kind of drive higher growth rates there. [ RARE ] is a little bit different just because that's relatively new. Reimbursement only kicked in over the last couple of years. So our competitors in that space are also growing quite nicely. And so we would anticipate that growing faster than the oncology.

Makes sense. And as you think about sort of the oncology segment, you mentioned 55% probably over the market is -- or patients are getting sequenced today. Within the existing portfolio, how much of that sort of existing addressable market can you serve? And then how does that expand as you get more tests approved across sort of solid tumor liquid and MRD the categories you talked about?

Yes. I mean I think expanding the 50%, 55% largely comes from Chi guidelines, right? And doctors adoption of this type of testing. You've seen some of that over the last couple of months where there has been expansions in things like pancreatic cancer to cover earlier stages and things. So as you look at that I'd say less so tied to new assays coming to market, more so that, again, these doctors are now leveraging these types of tests in term circumstances where they may have previously not done so.

All right. So then as we shift to sort of ASPs within the oncology segment there, you've talked about, there is a potential here to expand ASPs by about over $500 over the next several years here. If we look at sort of the categories of that, you can talk about an incremental $200 ASP growth that could come from just the XT to XT CDx transition. Can you remind investors sort of what percent of tests have been migrated to CDX thus far? Sort of what's inhibited the transition at all? And then when you think this migration is going to be complete?

Yes. So there's kind of 3 initiatives that we have on the ASP front. The first being that we got XTD was FDA approved, got ADLT status, and we began migrating volume in 2025. we ended the year between 30% and 40% of kind of volume that had been migrated. We're confident that we'll end the year in 2016 with the vast majority of the remaining volume kind of migrated over to the ADLT version of the assay -- there's a variety of factors that lead to kind of the timing of that. But again, we're confident that we'll be able to exit the year with the majority of that being on the ALT version. The second initiative relates to our liquid biopsy portfolio XF, which we submitted to the FDA earlier this year. won't have a 26 impact on rates. But as we get into 2017 and get approval, that will also lead to additional upside in reimbursement for XF. And then the last initiative relates to kind of commercial payers. -- we're largely still out of network lab and kind of payments for these -- for our assays varies greatly amongst kind of commercial payers. We do think that over time, we will continue to kind of make incremental gains with those payers obviously, as I just mentioned, like changes in guidelines that recommend the testing obviously need to coverage in certain instances when we get at the approval of assays, that also helps with some of those conversations with commercial payers. We'll continue to crack at the number of times that we received very little or no payments from commercial payers, and we think that adds $100-plus over the next couple of years.

Yes. And how much -- as you get sort of assays approved across solid liquid and then also at MRD, like how big of a factor is having sort of test in each of those categories into, let's call it, the broader commercial reimbursement where you could start to see that final $150 million, I think you talked about that factors in ASP growth.

Yes, I mean, I think certainly, again, it's payer by payer. They all have their own policies and kind of what they cover or don't. So may only cover solid tumor, some may only cover FDA-approved assays. So it's very fragmented, it's kind of how you chip away at it. we're doing all of the things that all of our peers are in terms of trying to drive up that ASP with commercial payers, which has lagged kind of where Medicare and metadata mine.

That makes sense. Okay. Then as we shift to the hereditary segment, you mentioned obviously nice market share gains due to some disruptions at a competitor last year. Starting to see that market sort of moderate in terms of the growth rate there. How should we think about sort of the blend of volume growth versus ASP expansion within that segment and as we think about 2020?

Yes. So ASP is obviously more mature in the hereditary screening space than they are in kind of therapy selection. And so on the kind of anticipated expansion we would see was as the rare business grows and becomes a larger piece of the pie. Obviously, the reimbursement for that testing is higher than what we had for hereditary oncology, -- and so that would lead to some increase over time. But we're not anticipating kind of significant changes in reimbursement for the hereditary space for -- until that cures.

Yes. And then I'm not directly comparable. Like do you see any knock-on effects from sort of the news out of Grail with some of the concerns around their test? Or does it change or shift sort of consumer perception of how sort of hereditary screening or free cancer screening is viewed by consumers in the marketplace in your view?

We don't believe so. I mean that is obviously kind of an early detection test the testing that we're doing, which is identifying whether you may be a carrier of a mutation that [ Luverne ] has the same mutation they developed cancer. So it's a different type of test and one that plays an important role in people's care. So we don't think that there will be much impact there.

Yes, absolutely. Okay. So if we shift over to the data business, we saw total contract value climb to north of $1.1 billion from $940 million net revenue retention and what drives. What are some of the key drivers that drive like future strength in the data segment?

Yes. So we have always said that over time, it kind of TCV growth in NRR should largely align with your revenue growth. There's obviously -- for us, when we sign kind of some large chunkier deals that has a larger impact on TCV in the short term. What we did say at year-end was greater than $1.1 billion of TCV, $350 million of that relates to 2026. By way of context, we did about $317 million of data and apps revenue in 2025. And so that backlog provides us a level of visibility as being through 2026 that is really strong. And so we do think that the way that these relationships are kind of established are that they are set up to kind of grow over time. meaning that they typically start small or someone is licensing kind of one cohort of data. They see value from that, and so they expanded to multiple indications or assets that they're working on internally. Ultimately, they arise to kind of some of these strategic collaborations of which we've disclosed a handful of those over the years. So the -- we work with 19 of the top 20 big pharma companies, a couple of hundred biotechs. So very good or strong customer footprint today. The good news is that many of them are still kind of in that small or mid-tier in terms of the number of cohorts that they're licensing. And so there's a tremendous opportunity for us to march all them up to kind of the strategic collaborations or the majority of up to those strategic collaborations that will kind of fuel the growth over the next couple of years.

Yes. And you've had a strong start to the year, frankly, within the data business and it's proven through sort of the announced deal with Merck, which I think is a new strategic collaboration partner, you just announced an expansion with Gilead. Can you just talk about sort of the environment for the data business right now? What kind of got Merck over the hump and sort of some context around what's involved in that deal?

Yes. So we mentioned that at kind of year-end on the earnings call was that we've had really strong engagement from our biopharma partners. Obviously, that's continued into the quarter with the announcement of several our deals, Merck being kind of the headline deal. The thing, as I just mentioned, these relationships kind of evolve over a multiyear period. And so these are folks that we have been working with previously on a smaller scale -- they've obviously seen the value that our data can bring to their drug discovery efforts and have decided to expand kind of those relationships more broadly across kind of multiple indications. So this is not a type of business where things can just kind of come up on a nowhere their relationships that we're cultivating over kind of multiple years, making sure that they see value in the data that they're licensing before they kind of expand. So it's really exciting to see Merck kind of join that group. Great to see the lead expansion kind of as well. But 1 that we're not surprised by, given obviously the engagement that we see from biopharma over the last 6, 9 months.

Yes. And maybe just as a clarifying point, because I think it's something that's interesting that as we've done meetings together, that I think it's really underappreciated within the investment community is sort of the opportunity for expansion within -- even within these large strategic collaborations with customers. Can you just talk about sort of what sort of in a typical deal like one of these collaborations, like what percent of the data set they're actually using and how there's opportunities to expand into additional cohorts over time. these aren't sort of just deals for the entire data platform from my understanding. Is it...

Yes. So I mean, I think that the reason why people kind of end up committing to multiyear strategic deals is really one of scale and discount, right? You can license on file from us or 10 files from us or you're just paying a higher rate than if you're committing to spend. And so what typically happens is that once organizations determine that they want to embed the data across their entire portfolio, they commit to those higher levels of spend to secure a higher discount. That said, it doesn't mean that you can download all of the records for these strategic collaborations. There's obviously limits in terms of what people can access. And it's a relatively small percentage of the overall database that they can download. So a lot of times, let's get in the door, let's expand this across the entire portfolio. That's a cap on what someone could spend in a given year. They may still come back and have 3 other projects that they need to work on. but it gives us the ability to kind of leverage the amount of spend that they have to secure larger discount.

Makes sense. Okay. One of the most notable deals you signed last year was sort of the foundation model deal with AstraZeneca. Can you just provide sort of an update on sort of where we are in sort of the development process of that deal? And how is that factoring into the pipeline of incremental opportunities for similar type of foundation model deals?

Yes. So we don't say kind of AZ pathos foundation model deal last spring as we talked about kind of on the earnings call, the development and training of that model has gone well. We've delivered the first version of that to Aztrazenica, and so we feel like that is going as well as we've expected. I think that in terms of pipeline, that's obviously a very large deal, and so it's difficult to predict when another one of those we've certainly seen an uptick from other folks that want to build smaller models, maybe indication specific. And so there's always ongoing discussions related to those types of deals. I don't think long term, because we get to retain a copy of the model and leverage it to make our data products and our diagnostics more intelligence, like that's not going to kick in, in the first version of the model, but you start to identify some of these insights that we can either turn into enhancements or insights in our diagnostics or leveraging the model to improve the effectiveness of our data products. That work is ongoing. So we're super excited about obviously, the work that we've done over the last 12 months to kind of get a first version of the model, it will be ongoing to continue to kind of tweak and further train the model. But certainly, we'll start talking more and more about the insights and things that we're seeing.

And as we've done sort of checks in the marketplace around sort of the competitive differentiation of Tempus data relative to peers in the group, it really sort of seems to boil down to sort of uniqueness of the proprietary data sets that you own and then sort of data visualization or ease of use in surfacing insights. So when you think about sort of where things stand today and the investments you're making on the data side of the business, what are the prioritization in terms of remaining competitive and differentiated there? Is it filling sort of gaps within the data set that you're looking at? Or is it working to sort of make the surfacing insights a lot easier or making the platform easy you are to use for your life sciences customers?

Yes. I mean I think we're always evaluating new data sets and how they can play a role in kind of the data offering. That said, what we have today allows us to kind of fuel the needs of our customers today. And so there's not kind of big gaps in the data -- the interesting thing about the data set is it's constantly updating, right? So every time there's therapy or a new guideline like you start to see that kind of flow through the data and our customers need access to kind of the latest data. So it's kind of maintaining that continuous flow of data is obviously important. And we certainly have made tremendous investments in the tool that I mentioned earlier called Lens, which is what researchers are kind of leveraging to access a lot of the data to build their models. And we've invented things like Tempus on there, where before you'd have to go in and kind of manually identify kind of to build your own cohort, you had to understand the data model. Now you can just talk to it and say, build me a cohort with these things and it can know automatically. So that ease of use is incredibly powerful, especially when you're dealing with the amount of data that we're leveraging with our life sciences partners that they're not spending all this time trying to just understand how to get to what they want. Have that be readily available for them so that they could spend their time modeling versus trying to build some of these couple of points.

Yes. Makes sense. When you acquired [ Ambry, ] I know it was noted that they didn't really have much of a data business at all. But there is an opportunity as you integrate it to sort of build products with and leverage that data sort of in the future. What's the product road map look like in this regard? And how near term could you start to leverage some of the data from [ Ambry ] in that data business overall?

Yes. So we've owned [ AMRI ] for about a year now, and they've operated largely independently -- what we think as Temps' role and kind of the ecosystem is how do you positively impact patients that comes in the form of intelligent diagnostics to providers also leveraging the identified data sets for purposes of drug discovery. And so we're in the process of kind of evaluating the nature of that data and how it gets incorporated into the data offering. And based on that, that will kind of dictate how quickly or long it takes us to kind of incorporate it into our life sciences partnership. So that is ongoing. That will continue to be ongoing in '26 and into 27, but certainly an area that we think the platform can be very applicable to -- given what we've built in oncology.

Yes, it makes sense. And then with the current sort of forecast, as you think about sort of the growth in the data business, it's not really dependent on integration of [ Ambry ] at all, I would imagine, with into that -- into the data center, any sort of starting of monetization there. Is there?

Yes. The vast majority of our data revenues today are within oncology in of late-stage oncology and so that it is not something that we're, again, having an impact in '26, '27.

Want to shift on the -- and touch at least on the AI applications business as well because I think this is where there's some really interesting sort of long-term future opportunity for Tempus. You continue to validate and gain reimbursement for a number of applications because obviously CG AFL go out there that's gotten some reimbursement on there. I think for you the sort of the framework developing for additional approvals in this portfolio? And what do we need to see to sort of hit an inflection point where the AI application business can sort of be a driver of growth and profitability over time.

Yes. And so when you talk about kind of that flywheel that I mentioned earlier, one of the exciting things is as you started masking all this data, you can start identifying very interesting insights that kind of get turned into algorithms of their own or [indiscernible] We have a suite of those in cardiology that you mentioned 2 of which have FDA approval. Have limited in reimbursements. We have some FDA-approved offerings in radiology. We had [ DigiPath ] through the Page acquisition. So we have a number of these efforts that are ongoing that have been deployed and are reaching some scale. But to your point, there isn't kind of a strong reimbursement framework. Unfortunately, there's not a great answer on kind of timing of when that changes. But our viewpoint is that it would be very odd for us that the health care system wouldn't want to pay for things that can have a positive ROI on kind of overall spend. So if you can identify a patient that's at a higher risk for something early on and prevent catastrophic events, the math probably works in your failure to reimburse those. Our approach has been to deploy these things, so to develop home deploy them and start kind of having them run in the clinic and kind of start building, obviously, the evidence of kind of their impact, while simultaneously having discussions with folks and the government about why these types of things should be reimbursed. For us, this is not something that we're building in. We've said it's going to be a multiyear effort. So it's not in our '26th guide or we don't anticipate a significant impact in '27. But the really exciting thing is that they can scale incredibly quickly because we're leveraging a lot of the integrations that we've already built with hospitals. So we can deploy these things, they're relatively inexpensive for us to run today. But when reimbursement ultimately does come, which we hope and believe that it will, then they can scale very quickly. And so because we sit on top of this very unique data set, that affords us the ability to kind of identify these insights, turn them into aortic diagnostics and deploy them while simultaneously kind of trying to push the -- or move the football on the field in terms of reimbursement has been our approach today.

And obviously, a lot of work to do on the reimbursement side and sort of getting FDA approvals as you continue to build the portfolio of algorithms. But where is sort of private payer thinking on this right now? Is it on their radar? How much education are you having to spend and do right now with the prior private payers?

Yes. I mean I think the private payers are similar to some of the government conversations of -- there is a recognition that these things can have or add enormous value to the health care system. But how do we prove that out before we're willing to reimburse. So I think those are the conversations you're having of, okay, show me, how exactly this plays out in reality such that we see the value that we think that we're going to see.

And sort of short of AI applications sort of obviously being a material revenue driver or margin driver in the business, obviously, there's a lot of sort of provider education right now and sort of as the opportunity to sort of attach the applications onto existing tests. Where does that stand today in terms of what do attach rates look like? And how does that maybe even just create a retention or differentiator relative to the other diagnostics tests on the market?

Yes. So we have a handful of algorithms within oncology today that are basically add-ons to some of the wet lab procedures that someone may order. Today, it's about 40% of our orders come in with one of the algorithms selected which gives back to kind of my earlier point of where we win within oncology therapies like one of ease of use and additional insight that we can provide. And I think demonstrated by the fact that we've got a large percentage of our orders coming in with these algorithms, just highlights that doctors obviously see tremendous value of them.

Yes, makes sense. One of the questions we got in from the audience is around sort of where you see the value and incremental investment within the Diagnostics business in terms of sort of investing in sort of new assays or refinement of assays versus sort of spending more of your investment on sort of the data side? How do you think about sort of the differentiation relative to the more traditional diagnostics vendors here?

Yes. I mean if you think -- you look at the investments that we've made historically, we've taken a slightly different approach than some of the other diagnostic players that are kind of always bringing the first assay to market in kind of first mover, where we've taken a slightly different approach we've been focused on being a very comprehensive sequencer, but then having these additional kind of insights. And so that's not to say that we don't invest obviously in new assays. We're obviously building out our tumor-naive MRD portfolio. We continue to iterate on versions of our core assays as well to add additional features but we just don't make all of our investments in that space. We also make investments on kind of the data and platform side and on these integrations that allow us to kind of provide that additional kind of insight. And so there's not a right or wrong answer as to kind of where people choose to make those investments. But it just gets back to kind of at our core, we view ourselves as kind of a data technology company unless so kind of a traditional diagnostic company that's always worried about kind of being the former because there's only kind of one transaction, which is what you get paid for the test and when it get cost, right? We have kind of multiple shots on goal through kind of the various offerings that we have.

That makes sense. Okay. Another question we got from the audience is around sort of the -- some of these regulatory proposals around attention for nationalizing [ MolDx ] and what sort of the implications or change could be going from moving from a regional system to more national. I guess how concerning is that? Or what sort of process would that sort of results for Tempus to do? What kind of disruption could it create?

Yes. So this is -- it's similar to the crush RFI discussion kind of earlier. Our view has always been -- the reimbursement landscape continues to evolve. It's evolved over the last 10 years that we've been operating in the lab. We always have kind of mitigating strategies to minimize any impact. And that's why we've been bringing things to the FDA to get approval and get under the NCD to the event that there's a change in any of the MAC kind of dynamics. But again, I think we're well positioned to minimize any impact to make.

Awesome. Good questions coming in. And so of those who have incremental ones, please let me know. We can get those asked and answered. Jim, I want to shift to obviously -- we'll give you a few CFO questions. Why not? Talk about the -- it's obviously a very large growth opportunity ahead of Tempus, both in the diagnostics business, also in the data business and the AI applications, plenty of ways to sort of allocate capital. Can you just sort of provide some color on sort of the framework of how you're thinking about growth versus profitability? And sort of how you're thinking about sort of reinvestment of gross profit dollars over the next couple of years and then sort of beyond that?

Yes. So we've told people that if you assume a 25% growth over the next couple of years, our plan is to reinvest 2/3 of the incremental gross profit dollars back in the business, 1/3 drop down to adjusted EBITDA. We think it's important to demonstrate that there's obviously continued -- we've demonstrated leverage but that there's continued leverage in the business. And so there's obviously -- we're operating at a scale now where that's a fair amount of discretionary spend that we're choosing where to invest. And so as we kind of look across the various offerings, you have diagnostics, which is growing at a healthy rate. We have MRD still to come, so that it's exciting. And so that's an area that's definitely worth doubling down on -- and on the data side, as we mentioned, we've been able to establish a business that operates at scale, but it's still relatively early days with most of our customers there. And so we want to continue to invest, so they continue to see that value that leads to kind of that expansion of those relationships. And then the last category is kind of the AI application is obviously isn't a near-term revenue driver. But we do think it's a worthwhile investment because that will afford us the ability to continue kind of growth longer. After about 3 years, we think we can reverse that trend of reinvesting 2/3 and that can drop down to 1/3 because at that point, you're generating a lot of gross profit dollars such that your investments you think that you can sufficiently invest while letting more drop down to the bottom line. So while we're mindful of profitability and kind of continued leverage in the business, we do think it's important given where we're at in our life cycle to continue to invest in these initiatives.

Makes sense. And then as you think about -- you talked about the 25% growth or trying to sort of work towards multiple years of growing at that consistent rate. How do you think about sort of where to invest to meter that growth in times Wednesday, last year, you had hereditary that was growing much faster than what you initially expected. This year, data is off to a really hot start. How do you think about sort of balancing and sort of shifting those investments as you're seeing sort of the changes in the end market activity on a year-over-year basis.

Yes. I mean I think there's always going to be periods -- you're going to have periods of time where you're going to have a reimbursement uplift, right? And that's going to have really downsized growth. And our view is, okay, let's see what is working and where can we double down on those investments? We're very thoughtful on that when we enter a new disease area, for example, we do so very small to be in with. We don't kind of get ahead of our skis. We want to prove it out before we kind of dove down. But then when you're running kind of ahead of plan in a certain area, then you want to identify those areas that are going well and kind of double down on those efforts. And so it's something that we look at on a quarterly basis to identify what are those areas that we should be looking at and then making sure, obviously, that they end up seeing that ROI.

Makes sense. Makes sense. And then -- and how are you thinking about sort of M&A as a capital allocation strategy moving forward? Obviously, historically been acquisitive over recent years, makes sense with Ambry to sort of build out the portfolio page is a nice interesting asset to add to sort of give more context when maybe sequencing doesn't -- isn't affected. But how are you thinking about sort of incremental M&A to supplement organic initiatives?

Yes. So we think that the -- on the diagnostics side, we have a pretty well-rounded portfolio at this point. Obviously, [ AMRI ] was an outsized acquisition relative to the ones that we typically do. Many of the companies that we look at are kind of smaller tuck-in acquisitions where somebody has built something very interesting from a technology or integration standpoint or in Pages example, a digital pathology kind of offering we think can be complementary to the current business but also has some future growth potential as well. And so we evaluate kind of all of these the same way of how does it fit into the offering -- is it something that ultimately will improve patient care? And if it checks and doesn't divert us from our kind of continued leverage in the business. then we look at kind of taking some of those folks in.

Awesome. We're not seeing any other questions from the audience. So we are going to leave it there. Jim, thank you so much for taking the time today, and thanks for everyone who joined us and for all the interactive questions from the audience. Have a good day everyone.

Thanks, Ryan, really appreciate it.