Tempus AI, Inc. closed 2025 with numbers that finally make its long-running narrative feel internally consistent. This was the year where the company stopped looking like a genomics-heavy diagnostics business trying to learn software tricks and started to resemble a data-and-AI platform that happens to be deeply embedded in healthcare. Fourth-quarter revenue reached $367.2 million, up 83% year over year, and even after stripping out the Ambry acquisition, organic growth of 33.5% is hard to wave away as financial engineering. The core point here isn’t just scale, it’s trajectory: Tempus is growing fast while simultaneously tightening the operational story around what actually compounds over time.
Diagnostics remains the gravitational center of the business, and in Q4 it showed real muscle. Revenue of $266.9 million represented 121.6% year-over-year growth, driven by oncology volume growth of 29% and hereditary testing up 23%. These are not vanity metrics. In diagnostics, volume growth means physicians are ordering tests repeatedly, workflows are sticking, and payer relationships are at least stable enough not to choke adoption. The more interesting signal inside diagnostics is MRD. Roughly 4,700 MRD tests in the quarter, up 56% sequentially, suggests Tempus is moving beyond episodic testing into longitudinal monitoring. That shift matters because MRD naturally lends itself to repeat usage, richer datasets, and tighter clinical lock-in. Once MRD becomes routine, it quietly changes the lifetime value math.
The second engine, Data and Applications, is where Tempus tries to justify its platform multiple, and Q4 made that argument more credible. Revenue of $100.4 million grew 25.1% year over year, but the real headline sits inside Insights, the data licensing segment, which grew 69.5% when you exclude the one-off AstraZeneca warrant impact from the prior year. That detail sounds technical, but it’s doing real work: management is telling you the growth is demand-driven, not accounting-driven. When paired with 126% net revenue retention and over $1.1 billion in remaining total contract value, the picture that emerges is one of customers expanding usage rather than churning through pilots. In enterprise data businesses, that’s where the flywheel either starts spinning or stalls out.
Margins add another layer to the story. Q4 gross margin came in at roughly 65%, with full-year gross margin around 63%. For a company that still runs physical labs and handles wet-lab reality, that’s a respectable base. More importantly, adjusted EBITDA turned positive in Q4 at $12.9 million and narrowed to a $7.4 million loss for the full year, even after digesting acquisitions like Paige AI and OneOme. GAAP losses are still substantial, $245 million for the year, but the direction of travel is unmistakable. Stock-based compensation remains heavy, which dilutes the purist narrative, but that’s not unusual for a company trying to hire scarce AI, bioinformatics, and clinical talent at scale. The question is no longer whether losses exist, it’s whether they’re shrinking for the right reasons.
Cash flow and financing make the subtext explicit. Tempus ended the year with roughly $760 million in cash and marketable securities, yet operating cash flow was negative $218 million and investing cash flow negative nearly $400 million, largely driven by acquisitions. This business is not self-funding yet. It is very deliberately buying time and capability, leaning on equity issuance, convertible notes, and credit facilities while markets are still receptive. Interest expense of over $70 million in 2025 is a reminder that leverage is now part of the structure, not just a footnote. If the company misses its profitability inflection, that cost will matter quickly.
Operationally, the product and partnership announcements line up cleanly with the financials. Paige Predict, which applies AI to standard H&E slides to infer 123 biomarkers across 16 cancer types, is a classic Tempus move: extract more signal from existing clinical material, reduce friction when tissue is scarce, and quietly expand the data surface area feeding the platform. The Immune Profile Score results, showing better prediction of immunotherapy outcomes than conventional biomarkers, reinforce the claim that Tempus isn’t just aggregating data, it’s learning from it. Strategic collaborations with NYU Langone and Northwestern Medicine push the same theme further, anchoring Tempus inside health systems where longitudinal data and real-world evidence can accumulate over years, not quarters.
Guidance for 2026 is where the optimism gets stress-tested. Revenue guidance of $1.59 billion implies about 25% growth, a deceleration from 2025 but still strong at this scale. More telling is the expectation of roughly $65 million in adjusted EBITDA for the full year. If achieved, that marks a real transition from “approaching profitability” to “operating leverage is now visible.” The risk, as always in healthcare AI, sits in execution details: reimbursement dynamics, competitive pressure in genomic testing, integration complexity across acquired platforms, and regulatory drift around AI-driven decision support. None of those are abstract risks, but none of them appear to be derailing momentum yet.
The cleanest way to frame Tempus after this report is simple, even if the business isn’t. Diagnostics supplies the data exhaust, data becomes insight, insight becomes software-like revenue, and the whole system improves as it scales. Q4 and full-year 2025 don’t prove the model is finished, but they do show it’s working well enough to justify the ambition. The next year will decide whether Tempus becomes a durable precision-medicine platform or remains a very impressive, very capital-hungry hybrid. Either way, the story is no longer theoretical, and that alone changes how the company should be read.