Tempus AI CEO on the “insanely large” AI healthcare opportunity

Tempus AI has only been public for a year. But it’s a been a big, volatile year. And its big swings — it’s been up as much as 120% and down by nearly 40% at times — have put it on the radar screen of some retail traders.

The stock is up more than 70% since its IPO and has doubled in 2025. Since the beginning of June, it has rallied more than 20%, giving it a market value of nearly $12 billion. That makes Tempus AI, founded in 2015 with partial backing from SoftBank, more than the equal, at least in the market’s eyes, of household name blue chips like Moderna, Molson Coors, and JM Smucker.

But sustained quarterly losses — Tempus has lost more than $770 million in its six quarters as a public company — as well as recent stock sales by top insiders have also attracted attention from short sellers. Their share of Tempus’ public float has nearly doubled over the last three months. In May, short seller Spruce Point released a searing analysis of the company, sending shares down nearly 20% in its worst day ever.

We sat down with Tempus AI CEO Eric Lefkofsky, a serial entrepreneur who was cofounder and former CEO of Groupon, just a few days before Spruce’s scathing report to speak about what exactly Tempus AI does and why he’s so excited about the opportunities to apply AI technologies in healthcare. As our discussion occurred before the Spruce Point report was released, we didn’t address its allegations in our interview.

This interview has been edited for clarity and length.

Sherwood News: I’ve got a general sense of what Tempus does. You’ve got a sort of medical testing business, mostly genomic sequencing for cancer patients, and a data business, where you license data that you’ve collected through your testing business. How does AI fit in?

Eric Lefkofsky: Basically, we have two main businesses. We have a genomics business where we essentially sequence patients, whether they’re cancer patients or cardiac patients or patients that are at risk of getting a disease. And then we have largely a data business where we license de-identified data pretty broadly to researchers and particular people that develop drugs.

AI is powering both of those businesses in different ways. The purely AI products we monetize, those that are algorithmic in nature, are really quite small in terms of revenue because our two larger businesses, genomics and data, make up the vast majority of our revenues. We do license some purely algorithmic models, but it’s small compared to genomics and data.

We’re essentially focused on this idea of AI-enabled diagnostics. How do you use the benefits of artificial intelligence to help navigate patients to the optimal therapeutic and help navigate researchers to the best research?

The way we do that is we aggregate vast amounts of multimodal healthcare data — in particular, really rich molecular data. We want to sequence patients and generate really rich molecular data, and then we want to know who those patients are, what drugs they took, and how they responded, which we get from clinical data. By tracking patients clinically, and by generating lots of molecular data for those patients, we can start to see patterns.

So underneath our diagnostic business and our data business, there’s this big engine churning through all this multimodal data — which is more than 300 petabytes of data — trying to look for those patterns, look for the insights that we can either embed into our diagnostics or embed into our data business.

Sherwood: And that AI work is something that you’re doing now? Or something that you plan to do?

Lefkofsky: Doing now. We’ve been an AI company for a long time. If you go on the Wayback Machine and look out our website in our earliest incarnations, we’re talking about big data and machine learning and natural language processing. These are all the precursors to the kind of large language models that we talk about today.

When these large language models began to really take off, largely with OpenAI and ChatGPT, we could see the the exponential benefit of these new technologies. So we pivoted large amounts of our technical team — we have about a thousand software engineers and Ph.D.s in data science, a very large technical team — we pivoted a large amount of that to working on our own agentic platform, which essentially seeks to take these large language models and make them work on healthcare data, where they weren’t designed to work. We’ve been at that now for a few years.

Sherwood: But that work right now is not a major part of the revenue that you’re producing. Your revenue comes from licensing your datasets and from the sequencing, the assays and things that you sell. Is that fair?

Lefkofsky: It’s hard to kind of rip the AI out. When we sequence patients, we have AI benefits that produce the results we deliver, right? Same thing in the data business. For example, our largest data deal we’ve ever sold or licensed was to basically build a foundation model in partnership with AstraZeneca.

Sherwood: So AI is enmeshed in the two lines; it’s not like there are three business lines and one of them is AI.

Lefkofsky: Exactly. It will never be a separate line item. It will always be just a bigger and bigger component of our two main line items.

Sherwood: You’re not GAAP profitable at the moment. How are you working toward that level? Is that way off in the distance?

Lefkofsky: We’ve told people, in various shareholder letters we’ve written and quarterly reports, that we seek to be EBITDA positive this year. So the first metric we want to get over is to be EBITDA positive. Shortly thereafter will be cash flow positive, because those things trail each other by a few quarters. Then, eventually, you work your way into GAAP earnings per share positive.

But I think the good news for a company like ours is, we’ve got pretty significant top-line growth, right? The overall business is growing north of 30%. Our data business in particular grew 58% last quarter.

These are very high-margin businesses. Our genomics business generates a close to 60% margin; our data businesses, a close to 80% margin. The whole thing blends into really attractive margins. So when you have something producing, let’s say, 70% gross profit, and it’s growing at more than 30%, you generate a lot of gross profit dollars.

Sherwood: Let’s take a step back. Talk to me about how you ended up in this world. I know your background a bit from Groupon, and as an investor. You were involved in media. You have a law degree. But this is some deep science. How did you end up here? Why do you feel like you’re a good person to be driving this kind company?

Lefkofsky: I got into tech at the end of the internet boom, so I’ve been in tech for about 25 years. I’ve essentially started six companies over those 25 years.

I’m fortunate that all six have worked. I started a company called InnerWorkings, which went public. I started the company Echo Global, which went public. I started on the Mediaocean that was sold to Vista. Our fourth was Groupon. Fifth, Tempus, and sixth, a company called Pathos.

So, I think what makes me ideally situated to tackle the problems Tempus is tackling is that I tend to be singularly problem-focused. It’s not that I think I’m the world’s best at building businesses, but I think I’m very good at being maniacally focused on solving problems that might be hard to solve and other people walk away from. And I just keep chipping away at it piece by piece and make progress.

Each problem I’ve tried to solve has been something I personally cared about or became intimately aware of. In the case of Tempus, my wife was diagnosed with breast cancer about 11 years ago. Here I am running Groupon and all of a sudden I’m dealing with that situation. And it was maddening that people looking to save money on a pizza had more technology at their disposal than people who were trying to save lives.

I just had this moment where I was like, “This is crazy and someone should try to go solve this problem.” So we began trying to solve that problems almost 10 years ago, and we’ve made a lot of progress.

Sherwood: What markets are growing the most for your company?

Lefkofsky: We’re lucky to be in two high-growth sectors. Genomic sequencing continues to be a high-growth sector. More and more people are being sequenced when they’re at risk of disease, when they have disease, certainly in oncology. We sit in the middle of a lot that growth and all those tailwinds.

Then in terms of AI and data, those spaces are still relatively small and there’s just enormous opportunity for growth there. I mean, we spend hundreds of billions of dollars a year on biotech and pharmaceutical R&D. By all accounts, an enormous amount of that is inefficiently spent.

I think people are looking for different solutions, and those solutions are going to have to be data- and AI-driven. There’s no other option. And companies that can really bring those solutions to the market are going to do well. That’s why our data business is growing so quickly.

Sherwood: One thing I wanted to ask you about: you sold some stock at the start of the year. They weren’t 10b-5 plan sales. Can you talk just a little bit about that?

Lefkofsky: They’re non-10b-5 sales because my 10b-5 has me selling stock over something like the next 20 years. I intend to be a very long-term shareholder and a very slow seller, as I have been in other places.

But I did have two stock sales at the beginning of the year. One had to do with the fact that I was granted a bunch of PSUs — performance stock units — and the company has no choice but to sell some of them to pay the taxes on those grants. So the company sold stock to pay the tax.

The second is because I’m a limited partner in a fund with other people. That fund had to sell its stock because it doesn’t hold public company stocks, so part of that was attributed to me as well.

Sherwood: So you don’t see those sales continuing much in the next...

Lefkofsky: That thing’s done. They had a little bit of stock and that’s over. I have no more tax obligations that I know of that are out there.

Sherwood: OK, thanks for that. That’s one thing I always watch, is insider sales.

Lefkofsky: It’s complicated when you think about it — we digress for a moment here — but when you look at a tech company that doesn’t go public for a decade and then it goes public, I think you generally have some people who want some amount of liquidity.

Even from my own team, it’s hard to look at what people do right after the lockup expires in an IPO as much of a signal. When you get a year out or so, and you get to a steady state, then the question is: do people believe or not, and how long of a holder are they going to be?

Sherwood: Peter Lynch had this line that goes something like, there are a zillion different reasons why executives sell stock, but there’s only one reason why they buy it. That’s because they think it’s going to go up. And the research shows that insider or executive buying is actually the signal, and that there’s not much signal to be found inside executive selling.

Tempus AI shares have done well since you gone public. How do you think about how your company relates to the market broadly? There’s a fair bit of froth, I think it’s fair to say, in the tech sector broadly. Is that something you care about?

Lefkofsky: I’ve said this before, and again, I have no idea if Tempus is going to be one of these companies. I don’t know. But when you look at the US healthcare sector, there’s nothing like it.

You’re talking $5 trillion of expense, and by all accounts, at least a third of that is wasted. So you’re talking $1.5 trillion of free cash flow wasted. That’s like ten Googles, ten online ad spaces, right? It’s just so insanely large. And I think the companies that bring AI to healthcare will get very, very big, regardless of any economic climate. Economic climates are irrelevant. I think it’s just that big.

Sherwood: Well, listen, thanks. You’ve been generous with your time. I really appreciate it. It’s nice to meet you.

Lefkofsky: You too. Be good.