Tempus AI grows revenue 36% to $348M yet GAAP losses widen to $126M

---

There is a particular kind of company that the market loves precisely because it is hard to value, and Tempus AI is its purest current specimen. It draws blood, sequences tumors, runs hereditary panels, and ships clinical diagnostic reports — boring, regulated, reimbursement-bound laboratory work that companies have done for decades. But it wraps that work in the most expensive two letters in the English language. The "AI" in Tempus AI is not a product line; it is a multiple. And the company's first-quarter 2026 results, reported on May 5, 2026, are a near-perfect laboratory specimen of the gap between what a story company says it is and what its income statement says it does.

Start with the facts the bulls lead with, because they are real. Revenue was $348.1 million, up a little over 36% year over year, beating the consensus estimate of roughly $345.4 million. The Diagnostics segment did $261.1 million, growing 34.7%, with oncology test volume up 28%. The Data and Applications segment — the part that actually sounds like artificial intelligence — did $87.0 million, up 40.5%, with the Insights data-licensing line growing 44.1%. Management raised full-year revenue guidance to $1.59–$1.60 billion and reiterated an expectation of roughly $65 million in adjusted EBITDA for 2026. On a slide deck, this is a high-growth, AI-native, soon-to-be-profitable healthcare platform.

Now turn the page to the GAAP income statement, where the costumes come off.

The loss is not shrinking — it is widening

The single most important number in the quarter is the one management does not put in the headline. Tempus posted a GAAP net loss of $125.9 million in the first quarter of 2026, versus a $68.0 million loss in the same quarter of 2025. That is not a narrowing loss on the path to profitability. That is a loss that grew by roughly 85% year over year, in a quarter where revenue grew 36%. Read those two growth rates next to each other slowly: the losses are compounding faster than the sales.

This matters because the entire investment thesis on a company like Tempus rests on operating leverage — the promise that as revenue scales, the cost base does not scale with it, and the company crosses into durable profitability. A genuine operating-leverage story shows the loss shrinking in absolute terms even as it spends to grow. Tempus is showing the opposite at the GAAP line. The company will tell you, correctly, that the quarter included $32.3 million of unrealized losses on marketable securities — a mark-to-market item, not an operating cost. Fair enough; strip it out and the underlying loss is still meaningfully larger than a year ago. The non-operating noise softens the story. It does not reverse it.

There is a tell in how a company talks about its own losses. Tempus directs investors to adjusted EBITDA, where it expects to be modestly positive for the full year. That is a legitimate metric for a scaling business. But the distance between "roughly $65 million of adjusted EBITDA" and "$125.9 million of GAAP loss in a single quarter" is the entire argument, and it is filled almost entirely with two things: stock-based compensation and the cost of buying growth.

Fifty-six million dollars of pay that never touches the cash line

In the first quarter alone, Tempus recognized $56.3 million of stock-based compensation expense and related employer payroll taxes. Annualize even a flat version of that and you are looking at a pay program that, in equity terms, runs in the order of magnitude of $200 million-plus a year against a company guiding to $65 million of adjusted EBITDA. Stock-based comp is the single most abused adjustment in growth-company accounting, because it lets a company claim its employees as nearly free. They are not free. Every share issued to an employee is a share of future ownership transferred away from the investor who bought in at the prevailing multiple.

Adjusted EBITDA, by construction, adds stock comp back — it treats $56.3 million of quarterly compensation as if it never happened. So when Tempus tells you it expects to be adjusted-EBITDA positive in 2026, it is telling you it expects to be profitable in a universe where a meaningful slice of its payroll is invisible. In the GAAP universe, where that payroll is counted and dilution is real, the company lost $125.9 million in three months. An investor should ask a blunt question: at what revenue level does Tempus actually generate GAAP profit, counting the equity it hands out? The company has not given a date for that, and the first quarter of 2026 — with losses widening — did not bring it closer.

Bought growth wearing organic clothes

The 36% revenue growth headline deserves an asterisk that the market has largely forgotten. On February 3, 2025, Tempus closed its acquisition of Ambry Genetics for $375.0 million in cash plus roughly 4.8 million shares of stock. Ambry is a hereditary-testing business that contributed about $102.6 million of revenue in the third quarter of 2025 alone. For the full year 2025, the consolidated Tempus-plus-Ambry business was guided to roughly $1.24 billion — described by the company itself as about 79% annual growth — a figure that is impossible to reach on organic momentum and that the acquisition manufactured.

This is the oldest move in the growth-company playbook: buy a business, bolt its revenue onto yours, and let the market price the combined top line as if it were one secular growth curve. By the first quarter of 2026, Ambry is folded into the base, so the 34.7% Diagnostics growth still partly reflects the residual lift of an acquired revenue stream maturing inside the reported numbers. None of this is improper accounting. It is, however, the difference between a company that is growing because its product is winning and a company that is growing partly because it wrote a $375 million check. Those two things command very different multiples, and Tempus is being priced as the former.

The AI in the name versus the AI in the revenue

Here is where allegation and fact must be kept carefully separate, because the most damaging claims against Tempus are allegations, not adjudicated findings. In May 2025, short-seller Spruce Point Capital published a "Strong Sell" report alleging, among other things, that Tempus "misrepresented the importance of its artificial intelligence business." Spruce Point asserted that AI-related revenue in 2024 was roughly $12.4 million — less than 2% of the company's total revenue of about $693.4 million that year — and argued this implied "a disconnect between the company's branding and its financial reality." Spruce Point estimated 50–60% downside.

These are Spruce Point's allegations. They are not proven facts, and Tempus disputes the report's characterizations; Wall Street largely shrugged, and the stock recovered roughly 15% in the days after the report. An investor should weigh a short-seller's claims skeptically — Spruce Point is itself positioned to profit from a decline. But the underlying tension the report points at is visible in Tempus's own segment disclosure. The company carries "AI" in its legal name and at the center of its narrative, yet the bulk of its revenue is conventional diagnostic testing reimbursed by payers, and the highest-margin, most "AI-native" line — Insights data licensing — is the smaller Data and Applications segment. The story is named after the smaller business.

Who is on the other side of the data deals

The forensic question that should keep a Tempus bull awake is not how fast revenue grows but who is buying. Spruce Point's report flagged related-party concerns around a deal with Pathos AI — alleging that Pathos, attached to roughly $200 million of total contract value, was "founded, managed, and funded by Tempus leadership and Board-related investment firms." Again: allegation, not adjudicated fact, and Tempus contests the framing. But related-party revenue is exactly the kind of thing that inflates a data-licensing growth rate without representing arms-length market demand. When a company's own insiders or affiliated funds stand up an entity that then becomes a customer, the revenue is real cash but the demand signal is hollow — you have not proven the market wants the product; you have proven your friends will buy it.

This is why data-licensing revenue, the most flattering and highest-margin line in Tempus's mix, deserves the most scrutiny, not the least. Total-contract-value announcements are not GAAP revenue; they are multi-year bookings that may or may not convert. A 44.1% growth rate in Insights is impressive only if the dollars come from independent pharmaceutical and biotech buyers competing for a genuinely scarce dataset — and far less impressive if a slice traces back to entities in the company's own orbit. The disclosure does not let an outside investor cleanly separate the two, and that opacity is itself the risk.

A founder whose past is part of the bet

Tempus is run by Eric Lefkofsky, a serial entrepreneur whose prior ventures include Groupon — a company whose post-IPO history is a cautionary tale in aggressive accounting and a collapsing stock. Spruce Point's report leaned hard on Lefkofsky's track record, raising governance concerns about prior companies associated with his name. This is not a financial metric and it is not dispositive; plenty of repeat founders have mixed records and go on to build durable businesses. But when a company's reported profitability hinges on non-GAAP adjustments, when its data-licensing growth involves related-party questions, and when its losses are widening rather than narrowing, the quality of governance is not a soft factor. It is the load-bearing wall.

Priced for a perfection it has not delivered

Strip everything down and you have a company doing roughly $1.6 billion of annualized revenue, losing money on a GAAP basis at a widening rate, paying out equity compensation at a scale that swamps its own adjusted profitability target, and carrying a market valuation that prices it as a secular AI compounder rather than a fast-growing but unprofitable diagnostics lab with an acquired revenue base. The stock traded around $55–$56 after the Q1 print. At that level, the market is paying a high multiple of revenue for a business whose path to GAAP profit has no public date attached.

That is the asymmetry. If Tempus executes flawlessly — converts its data backlog, scales diagnostics margins, grows into its valuation, and crosses into real profit — the upside is meaningful and the bulls are vindicated. But the price already assumes much of that. The downside, if loss growth continues to outrun revenue growth, if related-party questions migrate from allegation toward scrutiny, or if the AI premium compresses toward a diagnostics multiple, is a long way down. You are not being paid much to take that bet.

What the bulls genuinely get right

It would be dishonest to dismiss Tempus as mere narrative. The bull case has real load-bearing facts, and they deserve a fair hearing. First, the growth is genuinely fast and broad-based: 36% total revenue growth with 34.7% in Diagnostics and 40.5% in Data and Applications is not the profile of a stalling business — both engines are accelerating, not just one. Second, the MRD (minimal residual disease) franchise is scaling explosively, with volume up roughly 500% year over year to about 6,500 tests; MRD is one of the most clinically and commercially important frontiers in oncology, and an early lead there is genuinely valuable. Third, the company's adjusted EBITDA trajectory is improving sharply — from roughly negative $104.7 million in 2024 toward modestly positive in 2026 — which, even discounting the add-backs, reflects real operating leverage in the underlying business. Fourth, Tempus holds a substantial cash position (on the order of $764 million reported in late 2025) and low-cost convertible debt, giving it years of runway to fund losses without an emergency raise. Fifth, the data moat is not fictional: a multimodal library linking genomic, clinical, and imaging data across millions of de-identified oncology patients is genuinely hard to replicate, and pharma demand for exactly that kind of training data is real and growing. And sixth, the Merck collaboration and other pharma partnerships referenced alongside 2026 guidance are exactly the kind of independent, arms-length demand that would answer the related-party critique if they scale. A bull can look at this company and see a genuine genomics-data platform in the early innings of monetizing an irreplaceable asset. That case is not stupid. It is simply expensive, and it is being asked to carry a GAAP loss that is getting bigger, not smaller.

The denominator nobody discounts

There is one more frame worth applying: dilution as a hidden denominator. When a company funds itself partly by issuing $56 million of stock per quarter to employees and 4.8 million shares to acquire a competitor, the per-share value of every future dollar of profit erodes even if the business performs. Investors fixate on the revenue line and the adjusted-profit line because those are the numbers management presents. Almost nobody runs the share count forward. But the share count is the denominator under everything, and at Tempus it is growing in two directions at once — comp and acquisition. A business can grow revenue 36% a year and still deliver mediocre per-share returns if the share count compounds alongside it. That is the quiet tax the AI premium does not price.

Reimbursement is the floor nobody sees

For all the talk of algorithms and multimodal data, the largest part of Tempus's revenue — the Diagnostics segment at $261.1 million — lives and dies by reimbursement. Oncology and hereditary testing get paid because payers, including Medicare, agree to pay for them at specific rates set through coverage determinations and fee schedules that the company does not control. This is the unglamorous bedrock beneath the AI story, and it cuts both ways. On the upside, reimbursed clinical testing is genuine recurring demand rooted in patient care, not hype. On the downside, it means the most "secular-growth" part of the company is structurally a price-taker, exposed to coverage decisions, prior-authorization friction, and the perennial risk that a payer reprices a test downward. A platform that compounds toward omniscience and a lab that bills CPT codes are different businesses with different margins and different risk profiles, and Tempus is both at once. The market tends to price the first and forget the second. When growth eventually decelerates — and the law of large numbers guarantees it will, as a $1.6 billion revenue base cannot sustain 36% indefinitely — the question will be whether the underlying diagnostics economics can carry a valuation built for a data company. That test has not yet come, because growth has not yet slowed. It is precisely the moment of deceleration, not the moment of acceleration, that reveals what a business is actually worth.

Cash runway buys time, not a thesis

The bull's strongest defensive point is the balance sheet: roughly $764 million of cash reported in late 2025 plus low-cost convertible debt of around $468 million following the Ambry deal. That is real and it matters — it means Tempus is not staring at a financing cliff and can fund years of losses without diluting shareholders at a distressed price. But a long runway is not the same as a destination. Cash buys time for the operating-leverage thesis to prove itself; it does not prove the thesis. If the GAAP loss keeps widening, the runway simply lengthens the interval over which investors wait to learn whether the model ever generates real per-share profit. Convertible debt, meanwhile, is a future claim on equity dressed up as cheap financing today — another quiet line in the dilution ledger that the AI premium does not discount.

The kicker

The most honest way to describe Tempus is not "fraud" and not "the next great healthcare platform" — both are stories, and the truth is more boring and more uncomfortable than either. It is a fast-growing diagnostics company with a real and valuable data asset, an expensive habit of paying its people in stock, an acquisition stitched into its growth rate, a short-seller's unproven but unrefuted questions about who buys its data, and a GAAP loss that got bigger in the exact quarter the bulls were celebrating. The AI in the name is doing an enormous amount of work, and at $55 a share the market is paying full price for the costume while the income statement quietly insists on being a laboratory.

The numbers that matter most at Tempus are not the ones in the press-release headline but the ones beneath it: a $125.9 million quarterly GAAP loss that grew 85% year over year while revenue grew 36%, $56.3 million of quarterly stock-based pay added back to manufacture adjusted profitability, and a data-licensing growth rate whose independence from related parties the company has never cleanly let outsiders verify — and until those three lines bend, the AI premium is a bet on a story the income statement has not yet agreed to tell.