Tempus AI: A Valuation Paradox as Foundation Model Performance and Tumor-Only FDA Approval Reframe the Growth Story
Inaugural Investor Day, Chicago — May 29, 2026
Tempus AI held its first-ever Investor Day on Friday, delivering a dense slate of updates that, taken together, materially change how investors should think about the company's near-term revenue trajectory and its longer-term positioning as the connective tissue between oncology data, drug development, and clinical AI. Two developments landed simultaneously: the FDA approved the tumor-only version of the company's flagship xT 648-gene solid tumor assay, unlocking what management estimates is roughly $200 in incremental ASP across a test volume that spans hundreds of millions of dollars of revenue. And separately, the company disclosed early performance results from its large-scale multimodal foundation model built in partnership with AstraZeneca — results that, by management's own account, are tracking at or above expectations for a model that has been in pre-training for less than a year.
The FDA Approval That Changes the ASP Math
The tumor-only xT CDx approval is not a minor regulatory footnote. Until this approval, Tempus had FDA clearance only for the tumor-normal version of xT, meaning patients for whom blood could not be collected — a clinically common scenario — were outside the FDA-approved reimbursement pathway. Chief Commercial Officer Mike Yasiejko explained the practical significance: "There are many cases where we're able to capture that blood. In other cases, we may not for a number of reasons. And so now this will allow us to service all of those patients that are coming in through the clinic." CEO Eric Lefkofsky was more direct on the economics: "That's $200 of lift across a massive number of tests. I don't know the number, but it's probably $750 million of gain. So that's a real number."
The xF liquid biopsy assay has also been submitted to the FDA, though CFO James Rogers confirmed it will not impact 2026 ASPs. The more meaningful contribution arrives in 2027, at which point both approvals in tandem with ongoing commercial coverage expansion underpin management's confidence in a 25% three-year revenue CAGR for the diagnostics business — a target that would put the segment at roughly $1.9 billion. For context, the Street has been modeling closer to 20% through much of the forecast period, a discrepancy Lefkofsky addressed with some impatience: "There is no street model that I would look at and be like, oh, they know something we don't know. We have more information."
Volume Growth Running at 28% — and Holding in April and May
Clinical oncology volumes grew 28% in the first quarter, continuing an accelerating trend from 2025. Rogers told investors that April and May order data is tracking at a similar pace, a meaningful real-time data point given that MRD and CGP volume can be lumpy. Two structural tailwinds support the durability of this growth. First, CGP penetration among oncologists remains incomplete, with independent estimates suggesting only 40% to 50% of doctors who should be ordering these tests are actually doing so under current NCCN guidelines. Second, the market is migrating toward more comprehensive ordering — DNA plus RNA, solid tissue plus liquid biopsy — and Tempus is disproportionately positioned to benefit given that its RNA xR assay is now generating the majority of orders as a companion to xT.
The clinical case for RNA is increasingly difficult to contest. Tempus has published data showing that 21% of patients who receive both DNA and RNA results find FDA-approved targeted therapies that would have been missed by DNA alone. A specific example: 40% of rare NRG1 fusions — a class of alteration with dedicated therapeutic options — are detectable only through RNA. The combination of xT and xF liquid biopsy similarly adds approximately 9% incremental actionable variants over solid tissue alone. These are not marginal improvements in a field where a single actionable alteration determines whether a patient receives an effective targeted therapy or cycles through chemotherapy.
The Immune Profile Score: A Preview of What the Foundation Model Will Do at Scale
One of the highest-signal presentations of the day came from Chief Medical Officer of Oncology Ezra Cohen, who detailed the commercial performance of the company's Immune Profile Score algorithm, which predicts immunotherapy response across solid tumors using Tempus's longitudinal multimodal dataset. IPS now carries an attachment rate above 40%, meaning more than four in ten physicians ordering a Tempus test are electing to add this algorithmic layer. The clinical impact is material: IPS reclassifies roughly 20% of patients in each direction relative to traditional biomarkers like tumor mutational burden — identifying responders who would have been missed and non-responders who would have been inappropriately treated.
Lefkofsky framed IPS as the prototype for everything the foundation model will eventually produce at scale: "We've unlocked roughly 20% of patients who you wouldn't think would respond to immunotherapy that will, and 20% of people you think should respond to immunotherapy that won't. And that journey is what we expect to happen across all biomarkers and all diagnostics, not just in cancer for the disease areas." The ambition is to compress what currently takes a specialized computational biology team six to twelve months into a model-generated insight that can be validated and deployed on a weekly basis.
Foundation Model: Early Performance Clears the AstraZeneca Hurdle
Tempus disclosed new results from its foundation model this morning — a large-scale multimodal model built on roughly 500 petabytes of de-identified data across DNA, RNA, clinical notes, imaging, and digital pathology, trained on a cluster of approximately 1,080 compute nodes running at near-full capacity for 90 days. The model was developed under a roughly $200 million co-investment from AstraZeneca, with the partnership structured around specific performance benchmarks.
The primary benchmark was whether the foundation model could replicate the predictive accuracy of highly tuned, indication-specific models — including in settings where no prior predictive model existed. Lefkofsky confirmed the model has cleared those hurdles: "The fact that it performed this well this quickly, I think, is an indication of what is to come." The illustrative example disclosed publicly involves EGFR-mutant lung cancer, where the model demonstrates the ability to stratify patients beyond known co-mutations like TP53 — identifying additional signals predictive of standard-of-care response that conventional biomarker analysis does not capture.
A half-dozen pharma and biotech companies already have access to one or more of these models for active R&D use. Kate Sasser, who oversees the MRD and scientific strategy, noted that the broader pipeline of models is expanding rapidly: "When we say model, we're really moving quickly towards many models. You can imagine an ecosystem of these models. And then layering on top of that, things like agents that can leverage the models, tools and capabilities."
Data Business: $1.1 Billion TCV, 126% Net Revenue Retention, and Almost No Competition
The data and applications segment grew 41% in the first quarter, with the core Insights business — data licensing and model building — growing even faster. Net revenue retention for 2025 was 126%, total contract value ended the year above $1.1 billion, and $350 million of that TCV is allocated to 2026 alone. The company is working with 19 of the 20 largest pharmaceutical companies globally, up from 35 total data customers in 2020 to approximately 240 in 2025. Revenue concentration has fallen sharply: the top five clients accounted for 85% of data revenue in 2020; that figure is now 59% and declining.
Ryan Fukushima, who leads the commercial data business, walked through three representative case studies that illustrate the ROI framework Tempus uses in customer conversations. A global biopharma company seeking novel immunotherapy biomarkers received a 5,000-patient dataset combining pre- and post-treatment biopsy samples with DNA and RNA — a dataset that did not previously exist anywhere — and uncovered four novel drug targets, with the customer estimating a 30x to 50x return on spend. A second company used Tempus data to stress-test inclusion and exclusion criteria for a Phase III colorectal cancer trial, deriving what Tempus estimates as greater than $500 million in net present value from the risk reduction alone. A third pharma company used the platform to make a first-line versus second-line positioning decision for an ADC, ultimately proceeding to a go decision with validated patient subgroup stratification.
Lefkofsky was characteristically unambiguous about the competitive position: "I think per vacate, we said like there will come a time when no Phase IIIs ever fail. And some company like Tempus will be responsible for that." He added that in multiple retrospective analyses of large, failed Phase III trials, Tempus has concluded it could have predicted the failure in advance — in some cases from digitized H&E slides alone.
Hub, Paige Preview, and the Clinical Distribution Stack
The company provided a live demonstration of Hub, its physician-facing ordering and results platform, and its embedded AI assistant Tempus One. The platform aggregates molecular results across DNA, RNA, liquid biopsy, IHC, and algorithmic diagnostics alongside clinical data pulled from EHR connections, allowing physicians to query the system in natural language. The integration with Epic is scheduled for this summer. A companion product called Paige Preview uses digital pathology to predict molecularly relevant findings — MSI-high status, EGFR mutations, FGFR alterations — directly from H&E slides, enabling provisional therapy selection while NGS confirmation is pending. Separately, a product called Paige Predict addresses QNS failures, which occur approximately 7% of the time across the industry. Lefkofsky was blunt about the competitive significance: "This QNS thing is not small. There's no way to solve that problem. We have that problem. All of our competitors have that problem."
The ECG algorithmic business in cardiology, while not yet generating meaningful revenue, is operating at scale across 140-plus hospitals and touching millions of patients annually. A reimbursement code currently exists for part of the relevant population at $128 per ECG; management expects the code to expand to the full population. Lefkofsky's math on the eventual market size is pointed: "I would imagine that we will, for sure, be running some kind of algorithmic diagnostic on all ECGs in the future. And somebody will generate $1 billion or $2 billion of revenue just from that product alone."
MRD: Promising but Still Deliberately Gated
The MRD portfolio remains the most nuanced part of the Diagnostics business. On the tumor-informed side, the Personalis partnership is producing best-in-class sensitivity through whole genome sequencing of tumor tissue tracking up to 1,800 variants. Personalis has moved from zero to three Medicare coverage decisions in approximately one year, with more expected. However, the program is explicitly volume-gated: Tempus has only approximately 10% to 15% of its field force promoting MRD, and management is deliberately pacing order flow to match Personalis's laboratory capacity and financial profile. Lefkofsky acknowledged the constraint directly: "If we sent them 20x the volume of orders we were sending them today, how much cash would they burn? I don't think that math is sustainable."
On the tumor-naive side, Tempus is migrating from a first-generation assay with limits of detection around 500 to 1,000 parts per million to a second-generation version approaching market-competitive sensitivity levels near 100 ppm. The v2 assay is expected to replace the CRC product and then expand to pan-cancer rather than proceeding indication by indication. Given that tumor-naive MRD currently represents approximately 2% to 3% of total volume, it is immaterial to near-term financials but is the platform's most significant longer-term optionality. Lefkofsky was candid about the broader uncertainty: "I don't think you can just look at the trend lines today and be like, oh, they're going to continue for the next 10 years. I think there's going to be some movement there to figure out how to rationalize what's going on."
Hereditary and Rare Disease: Ambry Integration on Track, Whole Genome Launching This Summer
The Ambry Genetics integration continues to contribute meaningfully to the hereditary cancer testing segment, with growth moderating in the first half of 2026 as the business laps the share gains of 2025. Management expects acceleration in the back half. The most distinctive commercial asset is the Care platform, which automates identification and outreach to unaffected high-risk patients — a population Tempus estimates at over 70 million people in the United States, of whom only 1.5 to 2 million are currently tested annually. A pilot site identified 5,000 breast cancer patients with unfulfilled germline testing obligations within a 60-day lookback window. The RNA diagnostic layer embedded in Ambry's hereditary panels improved diagnostic yield by nearly 9% in a 40,000-patient published study — a material differentiation in a market where most competitors do not offer RNA.
A clinical whole genome sequencing test for rare disease is launching this summer, expected to improve diagnostic yield from the current one-third of patients to above 40%. Longer term, management views rare disease as a natural fit for Tempus's core competency of connecting molecular data to clinical context — the diagnostic odyssey for rare disease patients averaging five to seven years is precisely the kind of problem that benefits from Tempus's longitudinal data infrastructure.
Financial Framework: Path to Positive Free Cash Flow by Year-End
Guidance for 2026 calls for revenue of $1.59 billion to $1.60 billion, representing approximately 25% year-over-year growth, and $65 million in adjusted EBITDA. A convertible note transaction completed several weeks ago retired the remaining term loan, generating approximately $30 million in annual interest savings and setting up positive free cash flow by year-end. The three-year framework calls for 25% compounded revenue growth, with incremental gross profit dollars reinvested at a two-thirds/one-third split — two-thirds back into the business, one-third dropping to the bottom line — flipping to the reverse ratio after year three as the investment base matures.
On the valuation disconnect, Lefkofsky did not avoid the topic: "If we took the data and apps business public tomorrow, I would suspect we'd trade higher than the entire market cap of Tempus and could trade at 2x it. So it's not — I mean it's — one could argue it's got negative value." The company's compute infrastructure, he noted, is roughly equivalent to all of pharma combined — a claim that includes a cluster of GB200 nodes alongside the H200 cluster dedicated to the foundation model, plus additional operational compute that exceeds either cluster individually. Whether the market begins to ascribe appropriate value to that asset base may ultimately depend less on what Tempus discloses and more on whether the large pharma companies signing multiyear data agreements start talking about those commitments publicly in ways that force sector-generalist investors to pay attention.