TEMPUS

In 2015, Eric Lefkofsky's wife was diagnosed with breast cancer. He sat in oncology appointments and watched doctors make life-or-death treatment decisions based on intuition, decades-old clinical studies, and whatever they happened to remember from medical school.

There was no system feeding them real-time data about what had worked for patients with the same genetic profile, the same tumor markers, the same treatment history. The most important decisions in medicine were being made with some of the least data.

Lefkofsky had already cofounded Groupon — a company that scaled to a $13 billion IPO before falling apart spectacularly. He wasn't short of money or connections.

What he saw in oncology wasn't a charity problem. It was a data infrastructure problem.

Hospitals had mountains of clinical data — patient records, genomic sequences, imaging results, lab reports — but it was all siloed, unstandardized, and essentially useless at scale.

He founded Tempus in Chicago with the explicit goal of building the world's largest library of clinical and molecular data, and then building AI tools on top of it to help doctors actually use it. The pitch was simple: sequence tumors, aggregate the data, find patterns across thousands of patients, and tell the oncologist what's worked before for someone who looks exactly like their patient.

The execution, obviously, was anything but simple.

Tempus has two revenue streams and they work together in a flywheel — though the flywheel took years to start spinning.

The first stream is genomic testing. Hospitals and clinics send Tempus tissue samples from cancer patients.

Tempus sequences the tumor's DNA, identifies genetic mutations, and sends back a detailed report telling the oncologist which targeted therapies might work, which clinical trials the patient might qualify for, and what the molecular profile looks like. This costs money — paid by hospitals, insurers, or patients — and it generates data.

That's the key part.

Every test that runs adds another patient's molecular and clinical data to Tempus's library. Do that enough times and you end up with what Tempus claims is the world's largest set of linked clinical and molecular data — over 200 petabytes across millions of de-identified patient records.

That library becomes the second revenue stream: data licensing. Pharmaceutical companies and research institutions pay Tempus to access the dataset for drug discovery, clinical trial design, and research.

They're not buying patient records — they're buying insights from an anonymized dataset that no single hospital could build on its own.

More recently Tempus has expanded into AI applications for radiology, cardiology, and mental health — essentially trying to apply the same playbook beyond oncology.

Tempus's core product is its genomic sequencing reports — specifically the xT assay, which sequences 648 cancer-related genes from a tumor sample and delivers a clinical report telling the oncologist what mutations are present and what targeted therapies or clinical trials might apply. It's the product that generates revenue and generates data simultaneously.

On top of that sits the Lens platform — Tempus's data analytics tool that lets researchers and clinicians query the aggregated dataset. Think of it as a search engine for clinical outcomes: ask what happened to patients with BRCA2 mutations who received a specific therapy, and Lens surfaces the answer from real-world data across all of Tempus's partner institutions.

Tempus AI is the newest layer — a suite of AI-powered applications for radiology (automated analysis of medical imaging), cardiology (ECG interpretation), and psychiatry (matching patients to therapies based on clinical profiles). This is where the company is trying to prove that the platform isn't just an oncology business — it's an infrastructure play for AI in medicine broadly.

The TIME trial network connects patients to clinical trials based on their molecular profile — which matters enormously because most clinical trial failures happen because the right patients weren't enrolled. Tempus is trying to fix the matching problem at scale.

The counterintuitive move was going to the hospitals first instead of the patients. Most health tech startups try to disrupt hospitals by routing around them.

Tempus embedded itself inside them.

They signed data-sharing agreements with major cancer centers — Northwestern Medicine was the first big one in 2016. The deal was: hospitals give Tempus access to de-identified clinical data, Tempus gives them back better tools and reports.

Both sides get something. Hospitals weren't being displaced — they were being upgraded.

That reduced the resistance that kills most health tech companies before they get started.

The network effect is real but slow. Each new hospital that joins adds data.

More data means the AI models get better. Better models make the reports more useful.

More useful reports make it easier to sign the next hospital. By the time competitors started catching on, Tempus had a multi-year head start on building the dataset — and datasets in healthcare are almost impossible to replicate quickly because of the regulatory complexity involved.

The IPO in June 2024 was another calculated move. Lefkofsky had seen what happened to health tech companies that stayed private too long and burned through cash without a clear path to public markets.

Going public gave Tempus a currency for acquisitions and partnerships, and it forced the kind of financial discipline that private companies often avoid.

Tempus has never made money. Not once.

The company reported a net loss of $214 million in 2023, and losses have been consistent since founding. The genomic testing business is expensive to run — sequencing isn't cheap, regulatory compliance is constant, and selling into hospitals requires large enterprise sales teams.

The data licensing business has higher margins but takes years to build because you need the dataset first.

There's also a credibility challenge. The pitch — that AI can meaningfully improve cancer outcomes — is compelling.

The proof is murkier. Validating that an AI recommendation led to a better patient outcome is methodologically hard, takes years, and requires exactly the kind of controlled data that's difficult to generate in real clinical environments.

Critics have asked whether Tempus is genuinely changing oncology or selling an expensive data platform with an emotional origin story attached to it.

The competitive pressure is intensifying too. Foundation Medicine (owned by Roche), Guardant Health, and Illumina all operate in adjacent genomic testing spaces with serious resources.

And on the AI side, every major tech company is now circling healthcare data as a strategic asset. Tempus's window to build an unassailable moat is not unlimited.

Tempus has raised over $1.3 billion from a mix of venture capital, strategic investors, and healthcare-focused funds. New Enterprise Associates (NEA) was an early backer and has remained involved through multiple rounds.

Google led a $110 million round in 2019 — which was a signal moment. Google investing wasn't just about the money; it was a validation that one of the world's leading AI organizations thought Tempus's data strategy was credible.

T. Rowe Price, Baillie Gifford, and Franklin Templeton participated in later rounds, which brought in institutional capital more comfortable with longer time horizons and larger check sizes.

These are the kinds of investors who backed Moderna before anyone had heard of it — patient, thesis-driven capital.

Eric Lefkofsky himself is one of the largest shareholders, having put significant personal capital into the company. That matters in health tech — founders who are financially committed alongside institutional investors tend to make different decisions than founders who've already taken money off the table.

The June 2024 IPO on NASDAQ under the ticker TEM raised additional capital and gave existing investors liquidity, while keeping Lefkofsky firmly in control.