The Chicago company that decided cancer treatment should be informed by data - and then went and built the data.
Above: the Tempus wordmark, photographed mid-pivot. A logo so clean you'd never guess it sits on top of 500 petabytes of other people's worst days.
Adoctor in a community clinic types a patient's name. In seconds, a tumor's genomic profile, the imaging, the messy clinical notes, and the treatments that worked for patients who looked just like this one all arrive in one place. None of that used to be possible. Tempus AI made it ordinary.
Today Tempus is a public company - it trades on Nasdaq as TEM - sitting on more than 500 petabytes of molecularly grounded data and tens of millions of de-identified patient journeys. It runs genomic sequencing labs, ships an AI clinical assistant, and licenses de-identified data to the pharmaceutical companies trying to invent the next drug. It is, in the least romantic phrasing possible, an operating system for precision medicine. In the more romantic phrasing: it is trying to make sure no patient is treated as if they were the first.
The company reported $348.1 million in revenue for the first quarter of 2026, up roughly 36% year over year. That is the kind of growth that gets a health-tech company noticed. What got it built was something less tidy.
"Tempus has built the world's largest library of clinical and molecular data and an operating system to make that data accessible and useful, starting with cancer."- Tempus, on what it actually does
Here is the uncomfortable truth Tempus was built on: a cancer patient generates an astonishing amount of information - genomic, radiological, pathological, clinical - and almost none of it talks to the rest. It lives in PDFs, fax machines, proprietary lab formats, and the handwriting of busy clinicians. The result is that two patients with nearly identical tumors can get wildly different care, and nobody can easily learn from either one.
The dream of "precision medicine" - matching treatment to the molecular reality of a specific patient - had been promised for years. The obstacle was never imagination. It was plumbing. Someone had to do the deeply unglamorous work of structuring the mess.
"Every patient should benefit from the treatment of the patients who came before them."- The premise, stated plainly
Eric Lefkofsky is not the person central casting would send to fix oncology. He co-founded Groupon, the daily-deals company, along with Echo Global Logistics, InnerWorkings, and Mediaocean - a resume heavy on logistics and marketing, light on medicine. Then, in 2015, his wife was diagnosed with breast cancer.
Sitting through her treatment, Lefkofsky noticed how little data informed the decisions being made about her. For a man who had spent a career building software that organized information, this was both maddening and, evidently, an opportunity. He founded Tempus the same year. The name is Latin for "time" - which is either a tidy metaphor for a disease that runs on the clock, or simply a good word. Probably both.
The bet was specific: build the data infrastructure first, the diagnostics second, and let the two feed each other. Every test ordered makes the library richer; a richer library makes the next test smarter. It is a flywheel, and Tempus has spent a decade spinning it.
Lefkofsky's fourth company to go public. The first three sold coupons, freight, and print. This one sells the chance to be treated like a statistic - in the good way.- On serial founders and second acts
Eric Lefkofsky starts Tempus Labs after his wife's breast cancer diagnosis exposes how little data informs treatment.
Raises over $1.3B from Baillie Gifford, NEA, Novo Holdings, T. Rowe Price, Google and others while assembling the data library.
Prices at the top of its range, raising $410.7M at roughly a $6.1B valuation. Forms SB Tempus, a healthcare AI joint venture, with SoftBank in Japan.
Closes the ~$692M deal, expanding into hereditary and germline testing and accelerating revenue growth past 75% YoY.
Unveils an expanded agentic AI platform for drug development and lifts full-year 2026 guidance to roughly $1.6B.
Tempus does not sell a single magic box. It sells a stack. At the bottom is sequencing - the labs that read a tumor's molecular signature. On top of that sit the software products that make the data do something useful for a clinician at 4 p.m. on a Tuesday.
A generative-AI clinical assistant for oncologists. It uses large language models to pull insight from unstructured records and simplifies ordering and interpreting genomic tests - by voice, if you like.
The physician's command center: order diagnostics, review results, and interact with the full clinical context around a patient in one interface.
An agentic AI platform for exploring millions of de-identified records to spot patterns, find drug targets, and design smarter clinical trials.
Watches patient data against care guidelines and flags, in near real time, patients who have quietly drifted off the standard of care.
Pictured (in your imagination): four software products doing the job that used to require a fax machine, three phone calls, and a small prayer.
"The ability to have a voice assistant augment efforts in delivering precision oncology allows for direct clinical benefit for patients."- Tempus, on Tempus One
Skepticism is healthy, especially around the words "AI" and "healthcare" in the same sentence. So the relevant question is whether anyone uses this. The short answer: a large majority of U.S. academic medical centers, a broad network of oncologists, and the pharmaceutical companies that license Tempus data to find their next drug.
A growth curve that looks suspiciously like the flywheel they kept talking about. Figures are approximate and include guidance.
The partnerships tell the same story from the supply side. AstraZeneca and Pathos signed on for a multi-year collaboration worth around $200 million to build what Tempus calls the largest multimodal foundation model in oncology. SoftBank formed a Japanese joint venture, SB Tempus. Google is both an investor and the cloud underneath much of the data. None of these are companies that partner casually.
The diagnostics business feeds the data business, which feeds the diagnostics business. Boring to describe. Hard to compete with.- The flywheel, explained without a whiteboard
Precision medicine has been a conference buzzword for a long time. Tempus's contribution is to treat it as an infrastructure problem rather than a marketing one. The mission is unglamorous and exact: collect, structure, and analyze the world's clinical and molecular data so a physician can make a real-time, data-driven decision for a specific patient.
The work spans more than cancer. Tempus has extended into cardiology, radiology, and even pharmacogenomic testing for patients with depression - anywhere the gap between "data that exists" and "data a doctor can act on" is wide enough to fall into. Which is, it turns out, most of medicine.
"Tempus" is Latin for time. In oncology, that is not a brand flourish. It is the entire problem.- On naming
The early Tempus story was about sequencing and structuring. The next one is about foundation models - AI trained on multimodal medical data that can generate insight at a scale no individual lab could. The next-generation Lens platform and the AstraZeneca-Pathos model are bets that the library is now big enough to teach a machine something doctors don't yet know.
There are real risks. The company still posts net losses. Healthcare data carries privacy and regulatory weight that software companies rarely face. Competitors - Foundation Medicine, Guardant, Caris, Natera - are not standing still. Skepticism remains warranted. But the flywheel is spinning, and the data moat compounds.
Return to that oncologist with the laptop. A decade ago, that screen would have been mostly empty - a scattering of PDFs and a fax confirmation. Now it holds a patient's molecular story and the collected experience of millions who came before. The doctor still makes the call. Tempus just made sure the call is informed.