Tempus AI: the growth is real, but at $70.43 the price already pays for ten more years of it, so hold and do not add

In the third quarter Tempus grew revenue 23%, its oncology test volume grew 32%, and adjusted EBITDA went from $1.5 million a year earlier to $25.6 million. In the same quarter it paid $48.1 million in stock-based compensation, which adjusted EBITDA leaves out. For the full year the company expects about $65 million of adjusted EBITDA and about $215 million of stock pay. The stock closed at $70.43 on November 5, up about 61% in three months.

My Base value of $21.76 is 69.1% below the price of $70.43; put the other way, the stock trades at 3.2 times it. The Bear value is $6.82 and the Bull value is $56.63, which is still 19.6% below the price. I assign no probabilities to the cases. The verdict is Hold, do not add, with a buy range of $10.04 to $15.85.

Valuation route: a fast-growing diagnostics and data company that is profitable before stock-based pay and loss-making after it. The valuation is a ten-year free cash flow model in which every dollar of stock pay counts as a cost, built on a pro forma basis with the pending all-stock Personalis acquisition. A revenue multiple against Natera and Guardant is shown as a check, not averaged in, because revenue multiples say little about the value of a company that has not yet produced free cash flow. Values are present fair-value estimates, not 12-month price targets.

What Tempus is

Tempus, founded in Chicago in 2015 by Eric Lefkofsky, sequences tumors and inherited DNA for doctors and sells the resulting, de-identified data to drug companies. The two halves feed each other: every test adds to a library that the company says now holds more than 600 petabytes, and its software is connected to about 8,000 healthcare providers.

Diagnostics is three quarters of revenue. Oncology tests (tissue, blood and minimal residual disease, or MRD, tests that look for traces of cancer after treatment) are the growth engine. Hereditary testing grew mainly through the 2025 purchase of Ambry Genetics. Data and applications is the other quarter: data licensing to pharmaceutical companies, trial matching and AI tools, at a gross margin of 71.5% against 62.4% in diagnostics.

The latest results

The quarter was good where it matters most. Oncology volume growth accelerated, MRD tests rose about 300% to roughly 12,000, and data licensing bookings passed $550 million for the year after a new deal with Bristol Myers Squibb. Revenue guidance for 2026 went up to $1.60 to $1.605 billion, about 25% growth, and the stock rose about 10% on the day.

Hereditary testing is the soft spot. It grew 3% as it laps an unusually strong 2025, and management says unit growth returned to low double digits in September and October. For 2027 the company has two visible levers: a price update for its main tissue test from January 1, worth about $80 million of revenue, and a larger MRD sales push once each test starts earning a gross profit, which management expects around the middle of the year.

Which numbers to trust

• Adjusted EBITDA leaves out the largest cost after salaries. Stock-based pay (SBC) will be about $215 million in 2026, 13.4% of revenue, more than three times the adjusted EBITDA of about $65 million. I deduct all of it. After stock pay, Tempus is still losing about $150 million a year at the EBITDA level.
• Net income is flattered by the Personalis stake. Over nine months Tempus booked a $108.9 million gain on marketable equity securities, mostly its 12.5% stake in Personalis marked up toward the $16.25 deal price. The nine-month net loss of $135.1 million would be much larger without it.
• Cash flow is close to break-even, not positive. Operating cash flow was -$71.6 million over nine months and -$108 million over twelve, with receivables growing faster than revenue.
• Tax is not a near-term cost. Tempus has years of losses to offset future profits. I assume a $2.0 billion pool of usable losses (my estimate) and a 24% rate after that.

The good news in that list: cash runway is not the problem. With stock pay added back, my Base burns only about $40 million of cash in total before turning positive. The cost of growth is paid in shares, not in cash, and that is exactly why it has to be counted.

Debt, shares and the Personalis deal

Tempus has no bank debt. It owes $750 million of convertible notes due 2030 at 0.75% (convertible at $84.19) and $460 million due 2032 at 0% (convertible at $69.26), with capped calls that offset dilution up to $111.20 and $98.94. A convertible note held by Google, carried at $176.0 million and payable up to half in stock, adds about $200 million with interest (my estimate). Against that sit $637 million of cash and short-term securities and about $90 million of other investments. I treat the notes at face value as debt; all three of my scenario values are below the conversion prices.

The share count is the bigger issue. Weighted shares rose from 174.9 million to 181.3 million in a year, and on October 29 there were 182.4 million (177.4 million Class A and 5.0 million Class B, the super-voting shares held by Lefkofsky). Then comes Personalis: Tempus is buying the rest of the MRD specialist for $16.25 a share, in stock unless it elects to pay up to half in cash. Personalis shareholders vote on November 30 and closing is expected in December. At $70.43, paying for about 102 million Personalis shares and award units in stock means about 23.6 million new Tempus shares (my estimate from the filings), roughly 13% more. Personalis brings about $150 million of cash at closing and a business with quarterly revenue of $25.6 million and a nine-month net loss of $95.5 million.

I value 206.0 million shares, the current count plus the Personalis shares, and net claims of $532 million. If Tempus pays half in cash with new debt, the Base falls from $21.76 to $18.80.

Primary valuation: ten years of free cash flow

Five years is not enough for a company that turns cash-flow positive in about 2030, so the model runs to 2036. The Base starts from 2027 revenue of $2.01 billion, including Personalis, then grows 21% in 2028 and slows gradually to 5% in 2036. That is about 13.8% a year for the decade, against management's ambition of 25% or more.

The adjusted EBITDA margin climbs from 4% in 2027 to 26% in 2036, and stock pay falls from 12.5% of revenue to 6.5%. So EBITDA after stock pay reaches 19.5% of revenue, a margin that would put Tempus among the most profitable diagnostics companies. Capital spending including software is 3.3% of revenue and working capital takes 12% of each extra dollar of sales. I discount at 10.5% a year, with 3.5% growth after 2036.

The present value of the rest of 2026 is -$29.1 million, of 2027 to 2036 $1,051.2 million and of everything after 2036 $3,992.7 million. The terminal value is 79.6% of the total, high for a ten-year model, but that is what a business looks like when its cash flow only starts in year four. Enterprise value is $5,014.8 million. Less $532.3 million of net claims, that leaves $4,482.5 million, or $21.76 a share.

The discount rate matters most, and I tested it:

Even the friendliest corner of the table, $29.81, is less than half the price.

Secondary check: what peers trade at

On revenue multiples, Tempus does not look expensive. In late September, Natera's enterprise value was about 19 times its 2026 revenue guidance and Guardant's market value about 18 times. Both grow faster, about 34% and 37%, and Natera expects positive cash flow this year. Tempus trades at 9.0 times trailing revenue and 7.5 times my 2027 figure. A 7.0 times multiple on 2027 revenue would give $65.71 a share, close to the price.

That is the market's logic, and it explains the gap. At the 19.5% long-run margin in my Base, 7.0 times revenue is about 36 times EBITDA after stock pay. That multiple only works if growth stays very high for a very long time, which is the question the cash flow model answers instead of assuming. So I show the multiple as a reference and do not average it in.

The growth and margin inputs are the same in both columns. The whole difference is the price paid today for cash flows that start after 2030.

History

The multiple has swung between 6.5 and 13.4 times within a year, and the 52-week range is $40.77 to $89.09. This is a stock that trades on sentiment about AI and genomics as much as on results, which cuts both ways for anyone waiting for a lower price. Earnings-based checks such as the growth-adjusted price to earnings ratio do not work here, because earnings and free cash flow per share are still negative.

Valuation triangulation

• Cash flow model (the only weighted method): $21.76.
• Peer-style revenue multiple, shown as a check: $38.92 to $93.79 across the cases, $65.71 in the Base.
• Base value: $21.76.
• Even the multiple does not produce a buy: $65.71 less a 42.5% margin of safety is $37.78, well below the price. The choice of method changes how expensive the stock looks, not the decision.

What is the market already pricing in?

This is the core of the analysis. To justify $70.43 with my Base margins, Tempus would have to grow revenue about 27.5% a year for ten years, from $1.65 billion this year (with Personalis) to about $18.7 billion in 2036. With the margins of my Bull case, it still needs about 21% a year for a decade. Very few companies of this size have done that. Tempus might, but the price leaves no room for a slowdown.

Where a return would come from: if the market keeps paying 7.5 times forward revenue, the Base gives about 17.3% a year to the end of 2029, including dilution from stock pay. That depends entirely on the multiple holding while growth slows from 21% to 17%. My own Base, valued at the end of 2029, is worth only 2.2 times forward revenue.

Bear, base and bull scenarios

• Bear: hereditary testing stalls, data deals prove lumpy and MRD stays loss-making longer, so growth falls to 14% in 2028 and 5% by 2036. Margins still rise because less is reinvested, to 21.5% adjusted and 14.5% after stock pay. Revenue reaches $4.2 billion in 2036. Long-term growth 3%.
• Base: as described above, $6.0 billion of revenue in 2036.
• Bull: management roughly delivers its 25% ambition for four more years, MRD becomes a second engine and data keeps lifting the mix. Revenue reaches $9.1 billion in 2036, adjusted EBITDA margin 32% and stock pay 5.5% of revenue. Long-term growth 4%.

The discount rate is 10.5% in all three, so the risk sits in the cash flows. A revenue multiple of 4.5, 7.0 and 9.5 times would give $38.92, $65.71 and $93.79; those figures are references, not part of the values.

How the buy price is set

Confidence is C, which carries a margin of safety of 42.5%. Applied to the Base, that gives $12.51. But a one point change in the discount rate moves the Base by 20% to 27%, more than any other single input, so a single buy price would look more precise than it is. Using the Base at 11.5% and 9.5%, the buy range is $10.04 to $15.85. I would treat only the lower end, $10.04, as a clear buy.

Other single-input effects on the Base: three points of long-term margin change it by about $5.12, two points of growth in every year by $3.40 to $3.89, one point more stock pay every year by -$2.17, and each $100 million of extra net debt by -$0.49. If stock pay were ignored completely, which I do not accept, the Base would be $37.39 and the Bull $76.42.

For scale: at $30 the expected return under my Base is about 9.2% a year, at $45 about 7.7%, and at today's $70.43 about 6.5%, below the 10.5% cost of capital.

Dividend

Tempus pays no dividend and does not buy back shares. It issues them: through stock pay, through an at-the-market offering of $195.5 million in 2025, and now for Personalis.

Reasons to own Tempus

• Oncology volume growth accelerated to 32% in the third quarter, and MRD tests grew about 300%.
• Data and applications grew 33% at a 71.5% gross margin, and data licensing bookings passed $550 million this year.
• Adjusted EBITDA is guided to about $65 million in 2026; the third quarter alone made $25.6 million, against $1.5 million a year earlier, and third quarter operating cash flow was positive at $9.2 million.
• 2027 has a built-in lever: about $80 million of revenue from the January price update for the main tissue test.
• The financing is cheap: $1.21 billion of convertible notes at 0% and 0.75%, no bank debt and enough cash to fund the Base without new money.

What could go wrong?

• Dilution. Stock pay of about 13% of revenue, plus about 13% more shares for Personalis. If stock pay does not fall as a share of revenue, the Base is too generous.
• Reimbursement. Much of the value depends on Medicare and insurers covering genomic and MRD tests at current prices.
• Competition. Natera leads in MRD and Guardant in blood-based tests; both are larger in those markets and growing faster.
• Related parties. Part of the data revenue comes from Pathos AI and the SB Tempus joint venture, both reported as related parties, and Pathos paid $17.5 million of receivables in preferred stock this year.
• Control. Lefkofsky's Class B shares carry super-voting rights, and at least one law firm is investigating the fairness of the Personalis price for its shareholders.
• Refinancing. If the stock is below $84.19 in 2030, the $750 million notes must be repaid in cash.

Management execution

• 2026 revenue guidance was raised again, to $1.60 to $1.605 billion, and the third quarter beat expectations.
• Management has a clear rule for 2027: reinvest about two thirds of extra gross profit and let one third reach EBITDA, and it quantified the $80 million price uplift.
• Bank debt was replaced with convertible notes at 0% to 0.75%, with capped calls to limit dilution.
• The headline profit measure excludes stock pay worth 13% of revenue, and net income includes investment gains.
• Per-share results are still negative after more than a decade, and Personalis costs about 15 times its revenue run rate.

On my ten-point management check (two points each), Tempus scores 8 of 18 scored points; the buyback point does not apply. Operations and financing are handled well, per-share economics and governance are not.

Stock Analyza scorecard

The valuation score is 0 of 100.

A zero does not mean the business is bad. It means every price test fails at once: the price is above the Base, far above the Bear and even above the Bull. Business quality and valuation are shown separately and are not averaged.

Confidence score

The grade measures how well the model is identified, not the chance of a loss.

What I would watch from here

The Personalis vote on November 30 and the form of payment come first. The full-year report around February 23, 2027 then has to show four things: 2026 adjusted EBITDA of at least $65 million, 2027 revenue guidance of at least $2.0 billion including Personalis, stock pay guided at or below 12% of revenue, and a plan for positive operating cash flow for the year.

The three most important thesis breakers

1. Revenue growth stays above 25% through 2029 while stock pay falls below 9% of revenue, which leads toward the Bull and would make the price easier to defend.
2. Growth falls below 18% in 2027 as hereditary and data slow, which leads toward the Bear.
3. Stock pay stays above 12% of revenue through 2027, which keeps the per-share economics negative however fast revenue grows.

Stress test and independent check

I rebuilt the Base cash flow model in a second, separately written script, along with the comparison of methods, the history multiples, the buy range and the price ladder, and the results match to the cent. That checks the arithmetic, not the forecasts.

The weakest point is the length of the forecast: almost 80% of the Base value sits after 2036, and one year of guidance is a thin base for ten. The strongest bull argument is that Tempus combines testing volume with a data business that has no clear equal, and peers trade at twice its revenue multiple. The strongest bear argument is that the business has never covered its stock pay, and the price already assumes it will do so at a scale few companies reach.

Final verdict: hold at $70.43, do not add

Tempus is doing what a growth company should: volumes up, margins up, cash burn nearly gone before stock pay. The problem is the price, not the quarter. At $70.43 an investor pays for my Bull case and then some, and has to keep believing in 20% plus growth until the mid 2030s.

Verdict: HOLD at $70.43, do not add. Base $21.76, 69.1% below the price, Bear $6.82, Bull $56.63, Confidence C (13.75/25). Required margin of safety 42.5%, buy range $10.04 to $15.85, maximum buy price $10.04.

New money waits. For an existing holder the original cost is irrelevant; what matters is whether you believe in a decade of 20% plus growth, because that is what $70 already assumes. The quarter that would change my mind is one where growth stays above 25% and stock pay finally falls below 10% of revenue.

Sources

Primary sources are Tempus's third quarter 2026 results release (Form 8-K, November 2, 2026), its Form 10-Q for the third quarter of 2026, the shareholder letter and third quarter earnings call (November 2, 2026), Personalis's third quarter 2026 results and Form 10-Q (November 4, 2026) and the merger registration statement (October 2026). Most were accessed through Bigdata.com; all filings are on SEC EDGAR.

Peer guidance comes from Natera's third quarter 2026 release and Guardant's second quarter 2026 release. The reference market price is $70.43, the November 5, 2026 close; historical enterprise values come from company fundamentals data on Bigdata.com.