템퍼스(TEM) 2026년 3분기 실적 발표회: 매출 23% 증가, 가이던스 상향

핵심 요약

• 템퍼스(TEM)는 진단 및 데이터·애플리케이션 부문 전반의 성장에 힘입어 2026년 3분기 매출이 전년 동기 대비 23% 증가한 4억 1,150만 달러를 기록했다.
• 진단 부문 매출은 3억 350만 달러로 20% 증가했다. 종양학 매출은 1억 7,260만 달러로 28% 늘어났으며, 종양학 검사 건수는 32% 증가했다.
• 데이터 및 애플리케이션 매출은 인사이트(Insights) 데이터 라이선싱 부문의 29% 성장에 힘입어 33% 증가한 1억 810만 달러를 기록했다.
• 조정 EBITDA는 2,560만 달러를 기록했다. 템퍼스의 분기 말 기준 현금 및 유가증권 보유액은 8억 5,440만 달러다.
• 경영진은 2026년 매출 가이던스를 16억~16억 500만 달러로 상향 조정했으며, 조정 EBITDA 가이던스는 약 6,500만 달러로 유지했다.
• 템퍼스는 BMS 및 파소스(Pathos)와의 계약을 통해 향후 약 5년간 1억 9,250만 달러의 추가 데이터 및 애플리케이션 매출이 발생할 것으로 예상한다.

주요 재무 데이터

사업 및 영업 실적

종양학 부문은 진단 사업의 주요 성장 동력 자리를 유지했다. 매출은 28% 증가했고 검사 건수는 32% 늘어났는데, 경영진은 이를 수년 만에 가장 강력한 성장률 중 하나라고 설명했다. 또한 회사는 조직 생검과 액체 생검의 동시 검사 및 연속 검사 활용 증가에 힘입어 액체 생검 수요가 더욱 강해졌다고 밝혔다.

경영진은 유전성 검사가 전년도의 이례적으로 강력했던 성장에 따른 기저효과를 지난 후 긍정적인 방향으로 진전되고 있다고 밝혔다. 9월과 10월의 검사 건수 증가율은 10%대 초반으로 회복했다. 회사는 현재 가이드라인과 보험 급여 대상이 약 7,000만 명의 미국인에 달하는 반면, 템퍼스의 연간 검사 건수는 약 200만 건에 불과하다는 점을 들어 이 분야의 침투율이 여전히 낮다고 판단하고 있다.

데이터 및 애플리케이션 매출은 33% 증가했다. 템퍼스는 렌즈(Lens) 플랫폼을 통해 이 사업을 일회성 데이터 거래에서 다년간의 구독 모델로 전환했다. 고객은 전용 GPU 환경 내에서 데이터를 분석하고 모델을 구축할 수 있어 더욱 통합되고 지속적인 거래 관계를 형성할 수 있다.

템퍼스는 이번 분기에 추가된 50페타바이트(PB) 이상의 데이터를 포함해 총 600페타바이트가 넘는 데이터를 축적했다고 밝혔다. 회사의 인프라는 약 8,000곳의 의료 제공자와 연결되어 있다. 템퍼스의 광범위한 데이터셋에는 약 5,000만 명의 환자 정보가 포함되어 있으며, 이 중 80% 이상이 종양학 이외의 분야다.

회사는 브리스톨 마이어스 스퀴브(BMS)가 5년간 최대 1억 3,750만 달러 규모로 종양학 데이터 및 모델 라이선스를 취득할 수 있는 전략적 협약을 체결했다. 이와 연계된 BMS, 파소스 및 템퍼스 간의 협업에서는 알츠하이머병 데이터셋과 모델을 개발할 예정이며, 파소스는 템퍼스에 5년간 5,500만 달러를 지급하게 된다. 경영진은 이 두 계약이 향후 약 5년간 1억 9,250만 달러의 추가 매출에 기여할 것으로 예상한다.

또한 템퍼스는 키트루다(KEYTRUDA)와 함께 평가 중인 개인 맞춤형 신생항원 치료제의 상용화 가능성을 지원하기 위해 모더나(Moderna) 및 머크(Merck)와 다년간의 협력 계약을 발표했다.

경영진 가이던스

템퍼스는 2026년 매출 전망치를 16억~16억 500만 달러로 상향 조정했다. 회사는 조정 EBITDA를 기존과 동일하게 약 6,500만 달러로 예상하고 있다.

2027년과 관련해 경영진은 미국 식품의약국(FDA) 승인을 받은 xT 검사의 종양 전용(tumor-only) 버전에 대한 개정된 보험 급여가 1월 1일부터 적용될 것으로 예상한다. 회사는 이를 통해 2027년 동안 약 8,000만 달러의 종양학 단가 상승 효과가 발생할 것으로 추정한다. xF의 추가적인 단가 혜택은 FDA 절차 완료 및 승인 여부에 달려 있다.

경영진은 종양학 검사 건수, 보험 급여 상향, 데이터 및 애플리케이션 성장에 힘입어 장기적으로 최소 25%의 총매출 성장이 가능할 것으로 믿고 있다. 이는 공식적인 연간 가이던스라기보다는 경영진의 목표를 의미한다.

템퍼스는 증분 매출총이익의 약 3분의 2를 지속적으로 재투자하고, 나머지 약 3분의 1을 조정 EBITDA로 반영할 계획이다. 경영진은 신규 사업의 규모가 커지고 더 많은 매출총이익을 창출함에 따라 수년 후 이러한 배분 비율이 역전될 수 있다고 밝혔다.

리스크 및 주요 점검 사항

• 진행 중인 퍼스널리스(Personalis) 인수 건은 규제당국의 심사를 통과했으며, 11월 30일 주주 총회 투표가 예정되어 있다. 인수 종결은 거래 절차 수속에 따라 달라질 수 있다.
• 퍼스널리스는 현재 미세잔류질환(MRD) 검사에서 손실을 보고 있다. 템퍼스는 대장암 관련 보험 적용 가능성을 포함해 보험 급여 확대로 단위당 매출총이익이 흑자로 전환될 것으로 기대하지만, 경영진은 그 시점이 2027년 2분기 또는 3분기가 될 수 있으며 여전히 불확실하다고 언급했다.
• 본격적인 MRD 상업화 출시는 인수 완료 여부보다 보험 급여 적용 및 검사 단위의 수익성에 더 크게 좌우된다.
• 템퍼스는 인수 완료 후 퍼스널리스의 손실을 흡수할 것으로 예상한다. 경영진은 조정 EBITDA 개선세를 지속적으로 기대하고 있지만, 성장 투자를 희생하면서까지 단기 수익성을 극대화할 계획은 없다고 밝혔다.
• xF의 추가적인 보험 급여 인상 가능성은 FDA 승인 여부에 달려 있다. 경영진은 구체적인 승인 날짜를 확정하지 않았다.
• 유전성 검사는 전년도의 이례적으로 강력했던 성장에 따른 기저효과가 작용하고 있어, 최근 검사 건수 추세 개선에도 불구하고 이전 기간과의 비교가 까다로운 상황이다.

애널리스트 Q&A 하이라이트

진단 단가: 경영진은 PAMA(임상검사수가개선법)가 유전성 암 검사 수가에 미치는 영향이 제한적일 것으로 예상한다. 고가의 희귀질환 검사 비중이 커짐에 따라 시간이 지나면서 유전성 검사의 평균판매단가가 소폭 상승할 수 있을 것으로 전망한다.

MRD 확장: 템퍼스는 2027년에 MRD 영업 활동을 확대할 계획이며, 검사의 단위당 매출총이익이 흑자로 돌아서면 더욱 공격적인 확장이 가능할 것으로 본다. 경영진은 보험 급여 적용이 예상대로 진행될 경우 2027년 하반기부터 2028년까지 더 강력한 성장이 나타날 것으로 기대하고 있다.

퍼스널리스의 입지: 경영진은 퍼스널리스의 종양 정보 기반 MRD 분석법이 매우 민감하다고 평가하며, 템퍼스의 병원 연계 인프라, 영업망 및 연결된 검사 포트폴리오가 이 분석법의 도입을 지원할 수 있다고 믿는다. 회사는 2023년부터 해당 분석법을 독점 유통해 왔다.

데이터 매출 가시성: 바이오 제약 고객들과의 다년 계약 체결로 데이터 사업의 일회성 성격이 줄어들었다. 경영진은 이러한 계약들이 2027년, 2028년, 2029년 매출에 대한 더 높은 가시성을 제공한다고 밝혔다.

AI 모델 개발: 템퍼스 고객들은 전용 GPU 클러스터를 사용해 렌즈(Lens) 환경 내에서 분석 및 모델 개발을 수행하는 경우가 점차 늘고 있다. 계약에는 32개 또는 64개 GPU에 대한 3년 약정이 포함될 수 있어 플랫폼에 대한 고객의 결합도가 높아진다.

종양학 이외 분야로의 확장: 알츠하이머병 관련 협업은 종양학 외 분야에서 템퍼스가 체결한 첫 대규모 데이터 계약이다. 이를 통해 알츠하이머병 및 말기 치매 환자 약 50만 명을 아우르는 데이터셋이 구축될 것으로 기대된다. 경영진은 이를 데이터 사업의 새로운 성장 동력으로 보고 있다.

실적 발표 콘퍼런스 콜 전문

전체 실적 발표 컨퍼런스 콜 녹취록

경영진 발표

Operator

Hello, everyone. Thank you for joining us, and welcome to Tempus Q3 2026 Financial Results Call. [Operator Instructions] I will now hand the conference over to Kendra Webster, VP of Investor Relations. Please go ahead.

Unknown Executive

Thank you, Matt. Good morning, and welcome to Tempus' Third Quarter 2026 Conference Call. This morning, Tempus released results for the quarter ended September 30, 2026. The -- the press release, overview of the quarter and our latest presentation are available on our IR website at investor.tempus.com. Joining me today are Eric Lipkowski, Founder and CEO of Tempes and Jim Rogers, CFO.

Before we begin, I would like to remind you that during this call, management will be making forward-looking statements that are subject to risks and uncertainties and 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 mostly to the most directly comparable GAAP financial measures are included in our earnings release, which is available on our IR page. I will now turn the call over to Eric.

Eric Lefkofsky

Q3 was an incredibly strong quarter for Tempus with broad-based growth across both our Diagnostics and data and applications businesses. Momentum from the first half of 2026 carried into the third quarter. With total revenue of $411.5 million, increasing 23% year-over-year, led by accelerating unit growth of 32% in our oncology business and increasing momentum of 33% revenue growth in data and apps.

Our Diagnostics business delivered $303.5 million of revenue, an increase of 20% year-over-year, driven by oncology revenues, which grew 28% with volume growth of 32%, as we continue to see some of the strongest unit volume growth we have seen in years. Oncology revenues were $172.6 million, and hereditary cancer testing revenues were $111.3 million. Data and apps revenues were $108.1 million, increasing 33% year-over-year as that business continues to outperform, led by Insights, our data licensing segment, which grew 29% and -- these results reinforce the durability of our model as our best-in-class diagnostic platform enhances our data and AI offerings, distinguishes our business from our many competitors and helps drive the performance we have delivered consistently quarter after quarter.

There were also several notable highlights since our last quarterly call. We entered into a strategic collaboration with Bristol-Myers Squibb, or BMS whereby we agreed to license BMS oncology data and models for up to $137.5 million over a 5-year term. Additionally, BMS, Pathos and Tempus have entered into a collaboration to build data sets and models for Alzheimer's disease. Through this collaboration, Pathos is paying Tempus $55 million over a 5-year term.

As a result, over the next approximately 5 years, we expect to generate $192.5 million of additional data and applications revenue from these agreements alone. This is 1 of our largest strategic collaborations to date, and it's nice to see BMS join the likes of AstraZeneca, Merck, GlaxoSmithKline and BioNTech. As I've said before, if we signed just 1 of these big collaborations, we would be content.

We've signed 3 this year alone. We announced a multiyear strategic collaboration with Moderna and Merck to support the potential commercialization of their novel first-in-class individualized neoantigen therapy, or INT, being evaluated in combination with KEYTRUDA in patients with completed resected Stage 2 B4 melanoma and other cancer types. Given the exceptional response of the clinical trial data so far and the potential for this drug to be transformational, we are excited by the future of this program and thrilled to be partnering with Moderna and Merck.

We cleared regulatory review of our pending acquisition of Personalis and a personal shareholder meeting to approve the transaction is scheduled for November 30. We are more convinced than ever of the synergies between our combined companies. We have been the exclusive distributor of personalysis tumor-informed MRD assay next since 2023, which we believe is a best-in-class assay given its ultrasensitivity, MRD represents an estimated $20-plus billion market and is 1 of the fastest-growing segments in oncology diagnostics.

It is transformative for cancer care, allowing clinicians to detect disease recurrence earlier than traditional imaging enabling more informed treatment decisions when cancer recurs, bringing personalysis testing portfolio under 1 roof will accelerate commercial adoption, round out our overall portfolio and strengthen the multimodal data flywheel that continues to differentiate our business. We generated $25.6 million in adjusted EBITDA and ended the quarter with $854.4 million in cash and marketable securities. We had long suspected that with gross profit dollar growth outpacing our investments, we would eventually turn the corner and be able to generate meaningful EBITDA.

This quarter, we made the turn. Despite our increasing investments in diagnostics, we have always been a tech company. Our entire business model was built around the idea of collecting vast amounts of real-time, multimodal de-identified data to both improve clinical decisions and enhance research. A decade later, those efforts have allowed us to amass more than 600 petabytes of rich data, adding more than 50 petabytes last quarter alone. -- along with the technology infrastructure that is connected to approximately 8,000 providers, enabling us to use that data to train models and generate insights.

The scale of our business and the breadth of our portfolio has allowed us to build a significant data mode by compounding a decade of investment in our AI capabilities, we deliver solutions that grow more powerful with every patient touch point. cementing our position as an indispensable engine for health care innovation. As the world gets hungrier for the benefits AI can deliver, namely, allowing people to live longer and healthier lives Tempus is uniquely positioned as we have strategically amassed the data needed to power these models and the distribution platform to deliver them to patients.

As a result of our financial performance this quarter, we are increasing guidance to $1.60 billion to $1.605 billion in 2026. The -- we continue to expect 2026 adjusted EBITDA to be approximately $65 million. And with that, we will open it up to questions.

Operator

[Operator Instructions]

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

질의응답

Kyle Mikson

Congrats on the quarter. So 1 question on hereditary. You're hoping that return to stable growth next year, but just given these proposed financial schedule rates that Dr. and Hamaca. 'm just wondering how you're thinking about potential headwinds there given and then obviously, ASV tailwinds on the oncology side. So maybe just talk about diagnostics outlook for '27?

Eric Lefkofsky

Yes. I can start. So I think the overall hereditary business seems to be poised to do kind of much better as we move forward than it has over the last few quarters, in large part because it's both lapping excessive growth last year, which we've talked about relative to the some of the market dynamics where you had a big player in the market that was having some financial issues, a bunch of volume moved over.

And so instead of kind of nice even whatever year 15% growth year after year, you have this like jump up to 30% growth that we're now lapping. We've seen some good sign of that in September and October were unit growth rates have returned to low double digits. So we feel overall like that business is moving in a better direction. And whatever macro headwind there is in terms of guidance is kind of small compared to the just massive unaddressed TAM relative to the fact that current guidelines and current reimbursement would cover about 70 million lives and we run in the U.S. and we run something like 2 million tests a year.

So I think we still have a very underpenetrated market. I suspect as more and more literature comes out, finding clear links between germline variants and diseases that can be prevented or managed earlier, you'll see this category continue to be stable and grow. And Jim can cover ASPs on the oncology side.

James Rogers

Yes. So on ASPs for hereditary, just as a reminder, obviously, hereditary cancer based on the majority of that revenue, but we do have a small but growing rare business. And so we don't anticipate material impacts from PAMA on the hereditary cancer rates in that we've long said that the hereditary rates for us will actually probably increase slightly over time as rare becomes a bigger kind of piece of the overall pie.

So that's the store ISPs for hereditary. And then on the oncology side, as we talked about, given the expansion of our FDA approval to include kind of tumor only for our XT test, that updated pricing will kick in on January 1 and allow us to move the remaining XT volume over. That gives us about $80 million of uplift in oncology ASP in 2027. And then we still have XF kind of going through the FDA process.

And once that kind of finalize that process and hopefully gets approved, then there'll be even a larger increase as we migrate XF to the FDA-approved version of the assets. So lots of tailwind from ASP expected on the oncology side beginning early in '27, and we don't anticipate much impact on the hereditary AC.

Operator

Your next question comes from the line of Sarah James with Cantor Gerald.

Sarah James

I was hoping you could talk a little bit about the headwinds and tailwinds in '27. So can we assume that CGP volume holds in the low 20s and D&A might be closer to $30 million than $25 million you've still got the incremental gross profit, 2/3 of it even after Personalis being reinvested. Can we get to the 25% long-term range and thinking out further once we get the benefit of XTM excess uplift to '28 be a year above that 25% CAGR.

Eric Lefkofsky

Yes. I mean I think -- so first of all, we feel like a business at our size growing at 25% a year over a long term is incredibly healthy. So we've always said we're more interested in long-term 25% growth than we are in kind of short-term 30% growth that then decelerates. So I mean, when you talk about a business that's going to do a few billion dollars in revenue that's growing at 25%, that means like literally that every year, you're adding $500 million or more of new revenue, and then that becomes $1 billion of new revenue and more and more. So we're into the law of large numbers.

We're fortunate that our 2 main businesses, our oncology sequencing business is experiencing strong unit growth, about 32%. It's actually up from last quarter. It's got obviously rising ASPs, both because our solid tumor portfolio is about to be reimbursed at the ADLT rate and then when you get into the later part of the year, you'll have the -- could have FDA approval where you're going to get an even bigger step up. So we've got really strong unit growth, rising ASPs and our data business is performing even better in the aggregate at about 33% growth.

And obviously, we've announced a string of very big deals with Merck and BionTech and now BMS. And so it's really the core businesses are super strong. But we have other businesses like our rebate growing slower. And so in the aggregate, we feel like this kind of 25-plus percent growth rate is achievable and achievable long term. Is it possible that you get into '27 or '28 and have moments where the business is growing significantly faster, either because of ASP lift or because the data business just is running really hot, Yes.

But even if it is, even if we deliver a quarter where the aggregate growth is materially higher, I will say the same thing, which is it's great, I'm glad, but at the end of the day, if you can sustain kind of 25% growth over a decade over another decade, you will build a really large business. And we think we're building a really large business.

Operator

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

Mark Massaro

Nice continued growth in MRD, you're tracking about up 3,000 units per quarter, recognizing that only 15% of your sales force is selling today and the shareholder meeting is sort of around the corner after Thanksgiving. Can you just give us a sense for once personalysis in your portfolio, to what extent you plan to ungate and let the remaining 85% of reps sell the test. Is that something that we could expect in the first half of 2027? Or is that -- are there other criteria you need to see later in the year?

Eric Lefkofsky

I think it's going to be largely a function of -- it's more a function of reimbursement than it is a function of closing the deal. We have to be sensitive that at the present moment when personnels runs these tests they lose money. And once we acquire them, we'll be in the same boat. Granted, we are in this really nice place where given the categories they've got reimbursement for in 2026. They're in this rising ASP landscape where they're getting much closer to being breakeven on an individual test basis.

And we think with some of the approvals coming, obviously, in categories like colorectal cancer, we think we will turn that corner where on a per test basis, they're generating basically positive gross profit. That's a big unlock to really ramp up. Now that said, we're still ramping now. We're adding sales members to our team in the aggregate across all of our different assay categories, and we will in '27 dial up MRD in sales efforts.

But the big unlock, which I suspect will be, I don't know, it could be midyear, it could be Q2, it could be Q3. who knows. But somewhere, I suspect midyear, you'll be in this boat where they will turn the corner and be kind of positive -- it was profit dollar contribution on a per test basis, and then we'll feel much better about really starting to put the pedal to the metal. So I think over the latter half of '27 into '28, I would suspect growth will continue to be really strong.

Operator

Your next question comes from the line of Bradley Bowers with Mizuho.

Bradley Bowers

Just wanted to maybe ask another 1 on MRD while we're here. Definitely understand the timing with respect to how much you're getting paid per test and colorectal is a big unlock -- what do you think about the competitive landscape that you're launching into? Obviously, Tempus has a lot of things that it offers to physicians that would allow the test to win kind of at the physician level, but maybe just want to hear about the MRD offering kind of as it's fully rolled into Tempest and how that compares to competition and how we can maybe think about market share expectations for the test.

Eric Lefkofsky

Yes. I'll jump in. So the assay first of all, is best-in-class. I mean it's -- in terms of its performance, it performs as well as any assay out there. Obviously, we take a better but it's a super well-performing assay. So we're never meeting with oncologists, whether it's academic medical center or community practitioners who were saying, God, is wish you had a better test. They love the test. So the real unlock to volume is just that the full Tempus sales machine is kind of unlocked to sell MRD at scale.

And that, again, is a function of do you have unit economics that support that, and we're getting very close. I think once that's in place, the same benefits that drive our CGP business, which is this kind of best-in-class technology stack, deep integration to thousands of hospitals a broad compendium of tests that all are interconnected to clinical insights and algorithmic diagnostics, that all just literally feeds perfectly into MRD.

And our clients, our physicians are interested in us offering more and more, more broadly. And so I think there's really no competitive impediment to us getting to real scale. And I would suspect I think I've said this before, that at the end of the day, what Tempus has done that's, I think, unique is no other company like ours is #1 or #2 in more than 1 category. We are #1 or #2 in solid. We are #1 or #2 in liquid. We are obviously #1 or #2 in hereditary cancer profiling, and we will be #1 or #2 in MRD, I think.

There's just no reason for me to believe that market won't evolve the same way. When we entered solid tumor profiling, Foundation Medicine was every bit the kind of lies that Natera is today and the same thing in liquids. So I suspect we will do quite well in MRD because we wrap all this technology and all this AI and all this intextualization around these tests, and we just become a test of preference for our physicians.

Operator

Your next caller comes from the line of Ryan MacDonald with Needham.

Ryan MacDonald

Eric and Jim, maybe to talk about the success you're seeing in the data business, obviously, amazing year so far with the $500 -- over $550 million of bookings throughout the year. I guess as you look at sort of maybe the state of the pipeline and maybe of the business and sort of demand cycle.

Are you seeing any changes in terms of pull forward of demand -- and then as we think about sort of the components of the pipeline, what's the mix that you're seeing sort of within oncology, neuro, cardiology, some of these areas? And Jim, how does this sort of make you feel about the level of visibility going into 2027 to sort of hit those 25% to 30% targets that you're looking for in the business?

Eric Lefkofsky

Jim, do you want to take it?

James Rogers

Yes. I mean I'd say that the pipeline -- we've been talking about this now for several quarters. That's a kind of towards the tail end of last year, we saw kind of an uptick in engagement with our biopartners, we started kind of commencing out of that, and I think you've kind of seen it playing out in the deals that we've announced thus far this year. The beauty of these deals are as we kind of march our customers up from being kind of episodic purchasers to these large strategic collaborations, it obviously just enhances the level of visibility that we have for that business.

So you're no longer trying to get deals signed, you had them in kind of under contract for kind of 5, 6 years, very long-term deals that give us that visibility. And so as we enter '27, we feel like we're in a really strong position to see that growth, not only in '27, but '28 '29 as well.

Eric Lefkofsky

Yes. And I'll just jump in. I mean, I think the -- the nice part about the data business that we've been able to build is we used to -- our business about 4 or 5 years ago was much more lumpy and episodic in terms of the way we did these deals. But over the last half a decade, we've really migrated people to long-term subscriptions where people are basically connecting to our lens platform.

They have access to the totality of our data they're able to, based on the level of subscription they sign up for, pick a certain number of files that they can download annually and do work with more and more, as we've mentioned, they're now connecting that data to kind of private GPU clusters that we set up for them where they can build models that they can keep in our environment to basically do a bunch of their AI modeling work. And so it just -- it becomes much smoother long term, sticky. And so it's nice to kind of be going into -- we'll be going into 2027 and like the strongest position we've been in for a long time.

James Rogers

And 1 more thing that I'd add is 1 of the exciting things about the BMS collaboration is the work that we're doing in Alzheimer's. We've long talked about kind of expanding beyond oncology, although oncology drove most of the revenue today, but the data sets that we've invested in over the last 10 years well beyond oncology in other disease areas and seeing that demand pick up and expand beyond oncology is something we're really good except on.

Operator

Your next question comes from the line of EV Kozlowski with Goldman Sachs.

Evie Koslosky

So given the strength, I think you mentioned in the AI modeling portion of the Insights business, -- how does this change the stickiness of the relationships with your customers? And then maybe touch on what percentage of the Insights data customers also utilize AI modeling and then how you expect this to evolve into next year?

Eric Lefkofsky

Yes. I think the stickiness is essentially a function of how we've designed the data business. So as I mentioned a second ago, -- if you go back, we started licensing data, maybe, I don't know, roughly 6 or 7 years or something like that. And back then, you would be like, "Oh, I want 1,000 files of cancer patients that look like this and we would identify that data and then you would download that data in your environment. And so we didn't know what you were doing with the data.

We didn't know how -- what kind of value you're getting from it. We didn't -- we hope that you would call back 6 months later, a year later, want more data, but it was very transactional. Over time, we basically said, look, we built this very large environment. We have something like 700 software engineers that have been working here for almost a decade.

So we built this very large multimodal technology environment to interrogate this kind of data, and we started saying to our clients, why don't you do that work in our environment. It's just -- we've just invested way more in it than you have -- and so people began doing more and more of their analytic work, generating insights from the data in our lens platform. And then over time, we move to this kind of concept of, okay, we'll actually put all the data in lens you can build cohorts of interest in lens, pull down the data, just the exact data you want, the exact patients you want, which typically have criteria that would almost match a trial based on inclusion exclusive criteria.

We're going like very deep characteristics of a cohort, pull that data into a workspace within our environment. Build kind of either do your analytics or build models and then you're off and running. And obviously, that latter part building models is new. You used to be able to just kind of do analytics and then kind of take the analytics out or download the files directly in your own environment. But now we're connecting lens to 8 GPUs, 16 GPUs, 32 GPUs, some kind of cluster like that.

Typically denominations to date, where clients are signing up for like 3 years of 32 GPUs or 3 years of 64 GPUs and they will build -- they're building models in our environment that either have to stay in our environment or you pay an egress free to pull them out. And so it's gone from like not sticky to crazy sticky. And so we just -- in light of how people are using our data today, we just don't know how they would in mass get off unless somebody else builds the same set of products we've built, which we don't see.

Operator

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

Subhalaxmi Nambi

In your letter, you specifically called out strength in excess of order acquisitions being up 25% year-over-year. Could you give us any additional color on what is driving the strength? Is there any specific indication or academic setting, community setting EMR integration, any reason?

Eric Lefkofsky

I'm happy to start. So you have some kind of macro trends, which we've talked about a bit in the past that are just driving in the aggregate, some overall liquid volume. One of them is the benefits of concurrent testing, especially for metastatic patients is just high. So there's just been a ton of literature over the past few years about mutations, driver mutations you can find from solid tumor profiling or liquid profiling that you have to do both to find all of them.

And that -- and we published on this as well, but it can be 10% to 20% of mutations that would be therapeutically relevant that you won't find if you only do solid tumor profiling or you only do liquid biopsies. So you have this movement to kind of do both more often for later-stage patients that are high risk of mortality. And then you also have the benefits of kind of earlier testing serial testing certainly in certain indications where more and more, it's becoming important to be changing therapies every 6 months or every year and getting another insight from a test -- and so liquid is just a beneficiary of that.

We've seen it, I think, overall in the market, Garden, who's a leader in that space has posted really strong unit volume as well. And we just see this as a as kind of a near-term trend of really robust growth in liquid. Up until now, our growth rates have been relatively consistent. Sal's been growing really strong. as a liquid. And we've historically called out that it's not meaningful to talk about the difference between the 2 of them.

That's still the case for Q3. Our solid portfolio grew really quickly. But we did see a sign here in September, which we called out that we saw this kind of uptick of whatever, 300, 400 basis points. If that continues into Q4 you will start to see the first signs of really differentiation between liquid and Solidware, even with really strong growth rates and solid and it was outperforming. And with that plus MRD, it could be a propel growth rates over the next several quarters.

Operator

Your next question comes from the line of Dan Brennan with TD Cowen.

Daniel Brennan

Sorry for the voice here. Maybe just on EBITDA, nice print. Consensus has margins expanding mix, I think, to 7% and then 1 in -- and there's always this push and pull for diagnostic companies, how much you reinvest back in the business versus let it drop through. So I'm wondering, you've got that slide in the deck showing EBITDA like a line going up. Like to 71%, is that like a reasonable ramp how you guys think about it?

And connected to that. I know you had the $2,600 realized oncology price that you also pointed to in the next few years to be meaningfully above what we have today. So I'm just wondering how much of that increase in oncology pricing, you guys would be inclined to drop through to the bottom line?

Eric Lefkofsky

I mean to start so we'll jump in.

James Rogers

Yes, more than happy to start and then you can jump in. So I think our view on EBITDA kind of remains the same. We're still relatively early days in a lot of the offerings. And so we've had this perspective of kind of reinvesting 2/3 or roughly 2/3 of any incremental gross profit dollars that we generate back in the business and then letting a third drop down to the bottom line.

So obviously, with personal is coming on board, we are already going to invest a lot in MRD. We will utilize some of the reinvestment dollars to absorb with the burn from personnel is -- but then after another couple of years, then that slips, right?

Then you're generating meaningful gross profit dollars such that you really only need to invest a third, you can let kind of 2/3 drop down to the bottom line. So -- we want to demonstrate that we continue to see leverage in the business, which we saw obviously a big step forward that in the quarter. But we're not in a position where we want to maximize EBITDA in the short term because we think the opportunity set across our entire portfolio is so large and that would be a poor choice that our -- given that where we're at in our life cycle.

Eric Lefkofsky

Yes. And I was just going to add, we have obviously kind of now you can start to see this really strong EBITDA trend line that is picking up. but personnel who -- assuming that obviously closes here the next whatever later in Q4, they're burning money. And so you have to look at both businesses as one, we'll have to absorb some of their losses. But even absorbing their losses, we still expect improvements in EBITDA, and we're on a really nice trajectory where we've got to decide how to invest the strong gross profit dollar growth we're generating quarter-after-quarter.

Operator

Your last analyst question comes from the line of Dan Arias with Stifel.

Daniel Arias

Eric or Jim, within data, can you just maybe catch us up on where trials might land for the year on revenues or growth? And then what the outlook might be there in the context of where AI is helping to cause I mean trial matching seems to be 1 of these areas where pharma folks are most encouraged about how things can get better with some of these newer capabilities.

So curious whether you see an inflection there in the next year or 2?

Eric Lefkofsky

Yes. We -- so we took our real-time or just-in-time clinical trial matching business, which we call time. We tried to build the marketplace for a while to see if like small biotechs could load in their trials and with this marketplace evolved. We didn't see that growth. And so we pivoted the business really about a year ago, maybe 18 months ago, to be more enterprise focused on our largest enterprise data clients.

So we're fortunate now that this network, which spans north of 1,000 oncologists across the country and has this really robust platform to be able to identify patients that are fit for a trial and get them on a trial produces what pharma needs badly, which is enrollments. We're a high generator of enrollments for pharma. And so it's a really nice asset. And -- but I don't think it's going to scale into a $1 billion business.

And so we had to essentially target it toward, hey, we're in these strategic partnerships with more and more large pharma clients, whether that's BMS or Merck or AstraZeneca or whatever. And these folks want to need help, and we're tightly integrated. And so our time network now is enterprise-focused where you as a client have to come on, load on all or most of your oncology trials, and that's how the baskets are built.

And we have a wait list of folks now that want into that network. So the kind of market dynamics have allowed that to become a really premium product. And I suspect it will continue to grow nicely. It's just not going to be a huge business.

On the other side of our clinical trial matching business, our CRO business, as we mentioned years ago, we've been kind of deemphasizing some of the just pure CRO work because there are other people that do that and do it well, and it is a business that is in significant flux relative to artificial intelligence and what that's going to mean for the administrative and logistic components of -- and you're obviously seeing a lot of these big frontier modeling companies do deals with big Glen De departments at big pharma trying to figure some of that out.

We're fortunate that we don't play in that space. It's not important to us. And so we're kind of -- we'll let other people tackle that.

Operator

I'll now pass the call back to Kendra for additional Q&A.

Unknown Executive

Thank you, Matt. For the first time, we're pleased to answer questions submitted by retail investors through the Robin Hood platform. Our first question, can you classify the opportunity for growth outside of oncology?

Can you detail cardiology opportunities and other therapeutic areas of interest like nephrology, immunology and neurology.

Eric Lefkofsky

Yes. So I mean, obviously, as you mentioned a few minutes ago, 1 of the biggest benefits to the BMS deal is that it's our first really large deal outside oncology. So all of a sudden, we're building a very large Alzheimer's data set, it's a $55 million deal to generate a data set of like 0.5 million patients with Alzheimer's and late-stage dementia -- and so it's really exciting.

We've done smaller projects historically in data apps to enter these new businesses, but this is really the first time someone is in making a very large investment in real-world data with Tempus to build these kind of agent platforms. And we're excited because we think it's the start of a whole new round of growth for our data business, and we suspect it's going to continue at scale.

On the diagnostics side, we obviously are big in oncology. We do a bit of neuropsych testing. We do a bit of some rare testing. It's more a function of where is the reimbursement for these tests -- and we, like other companies kind of have to follow where the reimbursement landscape goes before we can enter new categories. But we do have 1 of the broader portfolios today, pan disease because we're in are, we're in neuro, we do some cardiac testing, obviously, with big cancer portfolio. So we touch a bunch is these areas.

Unknown Executive

Our next question -- if investor appetite for AI-related companies weekends, how confident are you that operating cash flow and existing liquidity can fund your growth plans and the Personalis acquisition.

Eric Lefkofsky

Super confident. Jim, do you want to want to come in.

James Rogers

Yes. I mean I'd say that the nice thing and part of the reason our business model is different is we're not overly reliant on 1 product. And so we have a very strong diagnostics business obviously, growth rates have been really strong in oncology. We're entering this period of kind of increasing ASPs pretty dramatically, almost $750 over the next couple of years.

And a lot of those dollars to be used to kind of fund investments elsewhere. Similarly, on the data side, as we continue to sign up kind of these long-term agreements with biopharma, again, gives us that visibility over the next several years to kind of continued growth. And lastly, as we've mentioned on kind of previous calls, we're reinvesting heavily today. And certainly, those investments are meant to drive long-term growth. And so we scaled that level of investment as necessary as well. But to Eric's point, we're very confident in being able to look.

Unknown Executive

Our last retail question. Since data is the main input for diagnostics and using AI to interpret and generate an outcome based on the data -- how can Tempus ensure that it has a growing inflow of patient data and how can it ensure that it will have access to it in the future since it doesn't own the patient data.

Eric Lefkofsky

Yes. So first of all, our diagnostic business is important for the data business, no question. Obviously, we're fortunate that we are 1 of the largest sequencers of patients for next sequencing in the world. And so we have a very broad portfolio from hereditary profiling to comprehensive genomic profiling when a patient is being treated to MRD and multiple diseases that we would suspect there's going to be a constant flow of new diagnostic data entering our platform.

That said, if you look at the 50 million patients or so within Tempus data set today, more than 80% of those patients are outside oncology and even the patients within oncology, I think more than 80%, we didn't sequence right? They're basically patients where the data is coming in because we have many, many, many data partnerships with academic medical centers and community practitioners. We have different products we offer where we get data as part of that -- as part of the product offering in terms of closing care gaps or matching patients to clinical trials.

And so we have a constant stream of data coming in from more than 8,000 providers. And so we've kind of over the last decade very intentionally tried to make sure the pipes that bring data into our platform are highly diversified so that we never have to worry about is the data going to dry up. And just as we are have crossed this chasm where the business is generating really strong gross profit dollar growth and EBITDA and will generate tons of cash flow, it's the same thing on the data side. We just have so much data flowing from so many directions that we've hedged that risk.

Unknown Executive

Thanks, Eric, and thank you all for joining our call this morning. Have a great week.

Operator

This concludes today's call. Thank you for attending. You may now disconnect.