Tempus AI, Inc.(TEM) · Healthcare AI & Precision Medicine

Tempus AI: A Long-Term Owner's Perspective

Other languages
Quick ReadPlain-language overview · read this first

Tempus AI sells de-identified oncology-testing data to pharma companies for R&D, running a "testing + data + AI" linkage. 2025 revenue was $1.272 billion, +83%; the story is real, the growth is real. Rating: Watch.

The problem is certainty, not the story. In 2025 operating cash flow was still a net outflow of $218 million, and owner earnings on a conservative basis were still negative, so the P/Owner Earnings valuation tool simply breaks down. At the current $43.93 and EV/Sales of 6.4x, it is cheaper than the equally cash-burning Natera/Guardant but pricier than the already-profitable Illumina: not deeply cheap, just not as expensive within a pricey peer group; the current price bets almost the entire optimistic scenario in.

The most fragile point is the terminal margin, not revenue: the 10-K admits that as foundation models grow stronger, pharma companies may no longer need to buy its data. Add acquisition-integration complexity, a dual-class structure giving the founder 57.9% voting power, and the dilution from newly issued $460 million of convertible notes, and the ideal buy range is $18-28, with the current price out of range.

Lead

A real business with real customers and real growth, but still paying for "future success": 2025 revenue of $1.272 billion (+83%), NRR of about 126%, and Remaining TCV above $1.1 billion, yet operating cash flow was still a net outflow of $218 million and free cash flow stayed negative. What the current $43.93 price lacks is certainty, not story: the valuation prices in near-optimistic assumptions. Rating: Watch, with a more attractive entry range of $18–28.

Full report

Prices in the article are as of publication; see the valuation band above for the live price.

Bottom Line Up Front

Because you did not fill in your "objectives" or "risk tolerance," I make the following judgment using a long-term, conservatively value-oriented framework; if you are a high-risk growth investor, your conclusion can be somewhat more positive than mine, but in my view it still falls short of a "Buffett-style buy."

Preliminary rating: Watch.

Core judgment: Tempus AI is not a hard-to-understand, scam-style story; it genuinely has a real business, real customers, and real growth, and it owns a rare "diagnostics + clinical data + pharma services + AI applications" linked platform. In 2025 the company's revenue reached $1.272 billion, up 83% year over year, of which Diagnostics revenue was $955 million and Data and applications revenue was $316 million; 2025 Net Revenue Retention was about 126%, and signed Remaining Contract Value exceeded $1.1 billion, showing that its data business has a degree of stickiness. At the same time, the problem is equally clear: as of the end of Q1 2026, the company still has not proven it can steadily generate real, distributable cash flow. In 2025 operating cash flow was a net outflow of $218 million, and estimated from capital expenditures and capitalized software spending, free cash flow remained clearly negative. At the most recent available closing price of $43.93, the company's market capitalization is about $7.86 billion; this price looks more like paying for "a very successful future" than for "a cash machine that already exists today."

Is there a margin of safety at the current price: not obvious. Viewed through a Buffett-style framework, what the current price lacks is certainty, not story. Buying TEM now essentially means betting that data assets will convert into higher margins, that acquisitions such as Ambry/Paige/Deep 6 will integrate smoothly, that dilution will stay manageable, and that regulation and reimbursement will keep improving. Once any one of these fails to materialize, the rate of return gets compressed quickly.

Suitable investor type: better suited to aggressive growth investors who can accept high volatility and high valuation elasticity and are willing to keep tracking operating metrics; less suited to traditional long-term value investors who put "predictable cash flow, steady capital returns, and margin of safety" first.

Biggest uncertainties:

  • Whether the "moat" from data and AI applications can actually translate into long-term high margins, rather than just high growth.

  • After acquisition-driven growth, when operating cash flow truly turns positive, and whether it can stay positive afterward.

  • Whether dilution, the dual-class share structure, and convertible-bond financing will make "the company bigger" while per-share intrinsic value grows more slowly than expected.

One-sentence conclusion: Tempus AI may be a company with great potential, but at today's price it still does not look like a value investment with a sufficient margin of safety.

The Business and the Industry

What is the core business. Tempus currently has two main product lines: Diagnostics and Data and applications. Diagnostics is essentially an oncology and genetics testing business, providing NGS, PCR, pathology, and molecular diagnostic services to physicians, genetic counselors, research institutions, and pharmaceutical companies; Data and applications takes the de-identified clinical, molecular, and imaging data accumulated in the diagnostic process and sells it to pharmaceutical and biotech companies for drug development, clinical-trial matching, analytics services, and AI applications. In its own 10-K the company explicitly states that these two product lines reinforce each other: more diagnostics means a thicker data asset, and the more successful the data business, the more it in turn drives usage of the diagnostics business.

Who are the customers, and how does it charge. Diagnostics' direct customers are mainly healthcare providers and biopharma companies. Its charging models include third-party payer reimbursement, contract research testing, and direct billing. Data and applications' main customers are pharmaceutical and biotech companies, with revenue coming from data licensing, AI-assisted clinical-trial matching, and analytics services. In 2025 the Data business's Net Revenue Retention was about 126%, indicating that existing customers as a whole are still expanding their orders; as of the end of 2025, Remaining TCV exceeded $1.1 billion, including about $300 million of optional add-on contract value.

Is the revenue recurring, stable, and predictable. In my view this company is not a purely "one-and-done" business, but it is also by no means Coca-Cola-style, highly stable repeat consumption. Within Diagnostics, oncology testing and genetic screening have a degree of continuous demand, but reimbursement, test-category iteration, and order cadence affect collections and revenue recognition. On the Data and applications side, contract stickiness is relatively high, but the company itself notes in the 10-K that this business is affected by pharma budget cycles, and data revenue is usually weighted toward recognition in the second half of the year. In other words, it has recurrence, but its predictability is still medium, not high.

Cost structure and dependencies. Diagnostics costs come mainly from lab personnel, consumables, depreciation, logistics, and third-party lab services; Data and applications costs come more from data acquisition, cloud computing, royalties, and delivery personnel. In 2025 the company's total revenue was $1.272 billion, against cost of revenue of $474 million, implying by the statements a gross profit of about $798 million and a gross margin of about 62.7%, continuing to improve from about 55.0% in 2024. The problem is that the gross-margin improvement has not translated into positive free cash flow. On dependencies, the company discloses that no single customer accounts for a material share of revenue, which is a positive; but it also discloses that for certain lab materials, equipment, and services it depends on a single supplier, which is an operational fragility.

Can this business be understood. In a single sentence, it can be understood as: first do the testing, then accumulate high-value medical data, then sell the data and algorithms to pharma companies and clinical settings. This is not mystical. But it also blends lab services, clinical reimbursement, a pharma data business, medical software, and AI algorithms, so its complexity is clearly higher than that of a typical consumer, insurance, or industrial company. My understandability score: 3/5. I can understand what it does, but I cannot estimate the stability of its cash flow 10 years out as easily as I can for a mature toll road, an insurance company, or a seasonings maker.

Industry stage and competitive landscape. At the industry level, this is not a declining industry; it sits at the intersection of precision medicine, molecular diagnostics, and healthcare data services, still in a growth stage. The long-term direction of demand is positive: individualized cancer diagnosis and treatment, genetic screening, real-world data supporting drug development, clinical-trial matching, and AI-enabled diagnostics all have a long-term demand base. But the other side of this industry is that technology substitution is fast, regulation is heavy, reimbursement is critical, and competitors are many. The competitors Tempus lists in its 10-K include, on the Diagnostics side, Roche/Foundation Medicine, Caris, Guardant, Natera, NeoGenomics, Quest, and Labcorp, and on the data and clinical-trials side, Flatiron, IQVIA, ConcertAI, ICON, PPD, and Syneos. In other words, Tempus sits in a good industry, but by no means an easy race track.

Industry attractiveness score: 3/5. The growth space is large, but the barriers of capital, technology, reimbursement, and execution are all high, and the moat is still forming dynamically.

Moat and Management

Moat assessment. Following Buffett's usual breakdown, I think Tempus's most credible moat is not brand, nor extreme low cost, but data scale + the ability to structure multimodal data + embedding into the clinical workflow + pharma partnerships. In its 10-K the company describes its core selling point as connecting lab results with clinical data so that testing "gets smarter the more it is done," and explicitly says a network effect exists between the two business lines. In 2025 the Data business's NRR was about 126% and Remaining TCV exceeded $1.1 billion, which also shows these data assets are not without commercial value. On top of that, it has a partnership agreement with GSK carrying a minimum committed purchase of $180 million, running through 2027, plus an option for an additional up to $120 million during 2028–2030.

But to call this a "durable, unbreakable" moat, I think the evidence is not yet sufficient. There are three reasons. First, the Diagnostics race is intensely competitive in itself, and the rivals Tempus lists are heavyweight. Second, the pharma data business is not run by Tempus alone; Flatiron, IQVIA, ConcertAI, and others are in the same profit pool. Third, in its risk factors the company explicitly warns that the development of foundation models could in the future weaken customer demand for its data products; if customers ultimately no longer need to buy de-identified data from Tempus, the economic rent of the data moat could narrow.

Accordingly, my judgment on the various types of moat is as follows:

Dimension Assessment
Brand advantage Has a degree of industry recognition, but is still far from a "strong brand premium"
Cost advantage Shows signs of scale effects, but has not proven low enough to suppress peers
Scale advantage Yes, especially in multimodal data accumulation and pharma partnerships
Network effect In embryonic form; Diagnostics and Data reinforce each other
Switching costs Pharma customers' existing projects, historical data structures, and integration processes create some stickiness
Channel advantage Workflow embedding into hospitals and physicians is strengthening, but is still not an absolute advantage
Patent/regulatory barriers There are barriers in medical-testing and algorithm-related settings, but they are not insurmountable
Data advantage This is the most critical and most genuine source of advantage
Culture/operating capability Execution looks strong from M&A and platform expansion, but is still being validated
Capital allocation capability Medium; there is forward-looking positioning, but also dilution and financing-cost issues

My moat-strength score: 3/5. More precisely, it is a "moat in the process of forming," not one already proven to be wide for 10 years.

Is the moat widening, stable, or narrowing. My judgment is that it currently leans toward widening, but the evidence is still not enough to conclude with high confidence. The 2025 acquisitions of Ambry, Paige, Deep 6, and OneOme did broaden Tempus's database and capability boundaries in genetics, digital pathology, clinical-trial matching, and pharmacogenomics. On the other hand, the acquisitions also made the quality of growth harder to read and sharply raised complexity. The moat may be widening, or it may just be piling up more assets together without forming a high enough unit-economic return.

Management credibility and shareholder alignment. Eric Lefkofsky is the founder, chairman, and CEO. The 2026 proxy shows he holds about 21.4% of the economic interest and all of the Class B shares, for about 57.9% of combined voting power, and that directors and officers as a group hold about 22.1% of the economic interest and 58.3% of the voting power. The upside is strong interest alignment, and management will not be easily driven by short-term market sentiment; the downside is equally clear: minority-shareholder governance rights are weak, and the dual-class structure significantly reduces external constraints.

Is capital allocation rational. I give it a medium score. On the positive side, management uses the money mainly on reinvestment and M&A, rather than paying dividends for show; in May 2026 the company issued $460 million of 0% convertible notes, explicitly to optimize its capital structure and repay the Ares credit facility, a step that is rational from an interest-cost standpoint. On the negative side, over the past two years the company's capital allocation has had a clear "build the platform big first, prove the returns later" style: the Ambry deal consideration was $600 million, of which $375 million was cash and $225 million stock; in 2025 it also acquired Paige and Deep 6. For a company that has not yet formed stable free cash flow, this pace of M&A amplifies the uncertainty of integration and capital returns.

Dilution is a very real problem. At the end of 2025, the company had 9.061 million granted but unsettled awards outstanding under its equity-incentive plan, plus 5.895 million shares still available to grant; its 2024 Equity Incentive Plan also carries an evergreen mechanism that automatically replenishes at 5% of total shares outstanding, and the ESPP has an arrangement that automatically increases at 2% of total shares. This means that even if the company's operations improve, per-share value growth could be dragged down by ongoing dilution. Add to that the $726 million of convertible notes already in place in 2025 and the new $460 million of convertible notes in 2026, and the potential expansion of the share count cannot be ignored.

My management and capital-allocation score: 3/5. There is vision, execution capability, and a highly aligned founder, but governance is not "shareholder-friendly," and capital allocation is on the aggressive side.

Financial Quality

Let me state the most important conclusion first: Tempus's financial statements look great on "growth" but are still far from passing the bar on "free cash flow." Revenue is growing fast and gross margin is improving, but operating cash flow and owner earnings are still negative. More importantly, the 2024 statements were strongly distorted by IPO-related stock-based compensation, and the 2025 statements were affected by the Ambry and other acquisitions, so the comparability of the statements is not ideal. Therefore, one cannot look only at "revenue doubling" and "Adjusted EBITDA improving"; one must watch cash flow, working capital, dilution, and per-share value.

Metric 2022 2023 2024 2025 Q1 2026
Revenue ($ million) 320.7 531.8 693.4 1,271.8 348.1
Revenue YoY 65.8% 30.4% 83.4% needs a comparable annualized figure
Gross margin 40.6% 53.8% 55.0% 62.7% needs a comparable figure
Net margin -90.4% -40.3% -101.8% -19.3% -36.2%
Operating cash flow ($ million) -168.2 -214.3 -189.0 -218.1 about -73.3
Estimated free cash flow (CFO − capex − capitalized software, $ million) -186.6 -248.9 -211.1 -245.3 clearly negative
Period-end cash & cash equivalents + marketable equity securities ($ million) needs supplementing needs supplementing needs supplementing 759.7 639.1
Rough period-end interest-bearing debt ($ million) needs supplementing needs supplementing needs supplementing around 1,239.5 933.9
Weighted-average shares (million) 63.0 63.3 119.8 174.3 178.9

In the table, the 2022–2023 data come mainly from Tempus's 2024 S-1 summary financial and cash-flow statements; the 2024–2025 data come mainly from the 2024 full-year results disclosure and the 2025 10-K; the Q1 2026 data come from the Q1 2026 10-Q and public earnings disclosures. Gross margin, net margin, and free cash flow are calculated from the figures in the public statements; where the company's basis differs from my calculation basis, the company's statements prevail.

Are the profits real cash profits, or accounting profits. At present it is closer to "accounting profit and non-GAAP metrics are improving, but real cash profit has not yet appeared." In 2024 the net loss was as high as $706 million, of which $548 million came from stock-based compensation and related payroll taxes, which significantly distorted GAAP net income; but even excluding such items, 2025 operating cash flow was still a net outflow of $218 million, showing that the problem cannot simply be blamed on accounting noise. In other words, Tempus's profit quality is improving, but it is not yet enough to prove that it is already a machine that steadily spits out cash.

Does growth require heavy capital investment. Looking only at fixed-asset capex, 2025 capex was $21 million and capitalized software was $6.216 million, whose share of revenue is actually not high; but if you include M&A, working-capital occupation, cloud resources, and R&D and sales investment together, this company's growth is very "capital-hungry." It is not heavy industry, but it is by no means a SaaS that can effortlessly turn revenue into free cash flow. In 2025 the company's net cash outflow for M&A reached $377 million.

Is the balance sheet sound. As of March 31, 2026, the company had $521 million in cash and cash equivalents, $118 million in marketable equity securities, and $2.135 billion in total assets; at the end of 2025 it had $728 million in convertible notes, $203 million in long-term debt, $209 million in convertible notes, and $100 million in revolving credit. Roughly estimated on the Q1 2026 basis, net debt is still around $300 million, and EBITDA has not yet steadily turned positive, so net debt/EBITDA and interest coverage do not look good, and strictly speaking are not very meaningful, because the denominator is far from "mature and stable." In May 2026 the company again issued $460 million of 0% convertible notes and repaid the Ares debt, which helps reduce the interest burden but also continues to build the valuation on top of future share price and future earnings.

Receivables, payables, inventory, and accounting risk. I have not pulled out the accounts receivable for every year in full from the materials I can currently verify, so I will not fabricate. What can be confirmed is that in both the 2024 and 2025 cash-flow discussions the company emphasized that changes in working capital clearly consumed cash; at the end of 2025 inventory was about $49.9 million, with inventory/work-in-progress up from the end of 2024; in Q1 2026 accounts payable fell from $81.99 million at the end of 2025 to $41.02 million, which is unfavorable for cash flow. Another point worth watching is that in the 2025 internal-control assessment by management and auditors, Ambry and Paige were excluded, together accounting for about 11% of total assets and 29% of total revenue in 2025. This does not indicate financial fraud, but it shows that the accounting and internal-control complexity of the post-acquisition integration period is rising.

My judgment on financial quality: If "financial quality" is understood as "growth speed + gross-margin improvement," it is making progress; if it is understood as "how much free cash flow each dollar of revenue turns into," it has not yet reached the standard a value investor wants.

Owner Earnings and Valuation

How to estimate owner earnings. The basic idea of Buffett-style Owner Earnings is:

Owner Earnings = net income + non-cash charges such as depreciation and amortization − maintenance capital expenditures − the increase in working capital needed to sustain operations

For a company like Tempus, the difficulty lies in two things. The first is whether stock-based compensation should be added back. My answer is: it cannot be mechanically added back, because for shareholders, ongoing stock-based compensation is essentially a cost, just paid in the form of dilution. The second is that the boundary between maintenance capex and growth capex is very blurry. Therefore, I can only give a conservative estimate, not a seemingly precise pseudo-number.

A conservative estimate of 2025 Owner Earnings. The 2025 net loss was -$245 million; operating cash flow was a net outflow of -$218 million; that year's capex was $21 million and capitalized software $6.216 million. Even granting that depreciation and amortization are relatively high, and that 2025 was affected by acquisition integration and some non-cash items, under a conservative framework 2025 owner earnings were still clearly negative, on the order of roughly -$250 million to -$350 million. The difference within this range mainly depends on whether you treat certain non-recurring items and part of the growth-related working-capital occupation as negligible. For a Buffett-style investor, what matters more is not the range's midpoint but the fact itself: it is not yet positive.

How many times Owner Earnings is the current valuation. The answer is: meaningless, or N.M. Because under the conservative basis owner earnings are negative. At the recent market cap of $7.86 billion, TEM is currently valued not as a "cash-flow machine" but as a "future platform-type, high-margin medical-AI asset."

At the most recent available close, TEM's share price was $43.93 and market cap about $7.86 billion. Roughly estimating from the Q1 2026 cash, marketable equity securities, and debt, enterprise value is about $8.16 billion, corresponding to a 2025 EV/Sales of about 6.4x; if we also factor in the $460 million of 0% convertible notes added in May 2026 and the Ares repayment, enterprise value is still roughly in the $8.4–8.5 billion range, corresponding to a 2025 EV/Sales of about 6.6–6.7x.

How to view relative valuation. The most comparable set is not pure software companies but the group of "molecular diagnostics / clinical data / precision-medicine platforms." Roughly estimating from recent market cap and 2025 revenue, Tempus's price-to-sales ratio is broadly below Natera and Guardant, but above the more mature, GAAP-profitable Illumina; if Exact Sciences is included, EXAS has weaker comparability due to M&A events and a different business structure, and can only be a reference. The key here is not "peers are all expensive, so it is not expensive," but rather: compared with equally high-valuation companies that are also growing, Tempus is not extremely cheap; and compared with companies that have already proven their earning power, it is not cheap enough to compensate for execution risk.

Company Recent market cap 2025 revenue Rough P/S Notes
Tempus AI $7.86 billion $1.272 billion about 6.2x Still has not proven stable FCF
Natera $26.37 billion $2.306 billion about 11.4x More mature growth, but still high valuation
Guardant Health $12.46 billion $982 million about 12.7x Screening expectations lift the valuation
Illumina $21.95 billion $4.34 billion about 5.1x More mature, more solid profitability

In the table, market caps come from recent market quotes; revenue comes from each company's official 2025 full-year disclosure. Because of differences in business structure, accounting basis, and growth stage, this is suitable only as a "valuation-framework reference," not for a one-size-fits-all conclusion.

Intrinsic-value estimate. Here I give three methods, but I must stress first: for a company like Tempus, still crossing the earnings inflection point, the valuation result is extremely sensitive to assumptions, and precision to the single dollar is false precision.

Method one: Owner Earnings discounting. Since current Owner Earnings are negative, I can only run scenario analysis for the case of "turning positive and scaling at some future point." My three scenarios are as follows:

Scenario Core assumptions Estimated intrinsic value
Conservative Revenue CAGR of about 10% over the next 10 years, Owner Earnings margin eventually reaching 10%, discount rate 12%, terminal growth 3% about $8–15/share
Neutral Revenue CAGR of about 12% over the next 10 years, Owner Earnings margin eventually reaching 15%, discount rate 10%, terminal growth 3% about $20–30/share
Optimistic Revenue CAGR of about 15% over the next 10 years, Owner Earnings margin eventually reaching 18%, discount rate 9%, terminal growth 3.5% about $45–60/share

These ranges are not "price forecasts"; they illustrate that at the rate of return a value investor requires, for the current price of $43.93 to hold, Tempus needs to be fairly close to the optimistic scenario. This is also the core reason I think the margin of safety is insufficient.

Method two: relative valuation. TEM's price-to-sales is not as high as Natera's or Guardant's, but it is not deeply cheap either. If over the next 3–5 years Tempus really thickens the margin and stickiness of Data and applications further, a price-to-sales of around 6x may not be outrageous; but if it ultimately looks more like a hybrid medical platform with "decent margins, mediocre cash flow, and ongoing reliance on financing and dilution," then this price is not cheap. In other words, relative valuation can only show that it is not at a bubble top; it cannot show that it has a margin of safety.

Method three: asset/liquidation value. This method is not friendly to Tempus. In Q1 2026 the company's total assets were $2.135 billion, including a large amount of goodwill and intangibles formed by acquisitions; at the same point in time, net intangible assets alone were about $331 million and goodwill about $470 million. If such assets are heavily discounted in a liquidation, Tempus's liquidation value provides almost no strong support to shareholders. Its value comes mainly from future cash flows, not from existing hard assets. This is also why I would not treat it as an asset-type stock where "you can mindlessly buy once it falls near book value."

Combined intrinsic-value range. In my view, a more reasonable expression is:

  • Conservative intrinsic-value range: $10–20/share

  • Fair intrinsic-value range: $20–35/share

  • Optimistic intrinsic-value range: $45–60/share

Viewed against the current $43.93:

  • Relative to the conservative range: clearly overvalued;

  • Relative to the fair range: on the high side or overvalued;

  • Relative to the optimistic range: close to the lower bound.

Therefore my price judgment is:

  • Ideal buy range: $18–28/share

  • Acceptable holding range: $28–38/share

  • Clearly overvalued range: above $50/share

  • Current price: $43.93, closer to an "optimistic-expectation price" than a "margin-of-safety price."

Margin of Safety, Risks, and the Bear Case

Is the margin of safety sufficient. My answer is clear: it is not sufficient. Not because Tempus has no prospects, but because realizing its value requires several fragile assumptions to hold at once: revenue keeps growing fast, the Data business maintains high NRR, Ambry/Paige/Deep 6 integrate smoothly, the reimbursement environment keeps improving, operating cash flow turns positive as soon as possible, and equity dilution stays controllable. If any one deviates, the central return at the current price shifts down.

The most fragile valuation assumption. Not revenue growth, but the terminal margin. The market can usually tolerate short-term losses at a growth company, provided the future margin can be good enough. Tempus's current bull case is essentially a bet that once the data business, AI applications, and diagnostic capabilities are combined, they can form a higher long-term margin than a traditional lab. If it turns out to be just a platform with a little more software than a traditional lab but still dragged down by reimbursement and sales investment, then a price-to-sales above 6x lacks support.

The most important risks. First, competition risk: the rivals Tempus faces include both diagnostics leaders and pharma-data and CRO leaders, so the track is not a blue ocean. Second, technology-substitution risk: the company itself admits in the 10-K that the deployment of foundation models could in the future weaken customer demand for its de-identified data products. Third, regulatory and reimbursement risk: Diagnostics relies heavily on third-party payer reimbursement, and pricing power is not fully in its own hands. Fourth, acquisition-integration risk: a large part of the 2025 growth came from Ambry consolidation, and if integration goes poorly, both revenue quality and cash flow will suffer. Fifth, financing and dilution risk: dual-class shares, ongoing stock-based compensation, and convertible notes could all make per-share value grow more slowly than corporate scale. Sixth, rising accounting and internal-control complexity: the 2025 internal-control assessment excluded Ambry and Paige.

The strongest bear case. I would sum up the bear logic in one sentence: "Tempus is not a cheap monopoly asset; it is a multi-business platform integration story, wrapped in a high valuation, that is still burning cash and diluting." This view is not absurd. Bears would say: the 2025 revenue surge was driven mainly by the Ambry acquisition; the 2024 GAAP loss was distorted by IPO-related equity expenses, masking the real unit-economic quality; 2025 operating cash flow was still -$218 million, showing the "earnings inflection point" has not truly been crossed; and although the data business looks attractive, its real long-term pricing power and terminal margin remain unproven.

What facts would overturn my Watch stance and turn me positive. If the following combination appears over the next two to three years, I would clearly raise the rating: first, operating cash flow steadily turns positive, and free cash flow turning positive no longer depends on one-off working-capital swings; second, the NRR of Data and applications stays stable at a high level while TCV keeps growing; third, the pace of share dilution slows markedly; fourth, after the acquired assets complete integration, gross-margin gains reach per-share cash flow rather than stopping at Adjusted EBITDA.

What facts would prove the judgment wrong, which I should acknowledge. Conversely, if the following occur, I would conclude that "this is not a business worth holding for the long term":

  • Data-business NRR clearly falls below 110% and stays there for several quarters.

  • TCV growth stalls, while revenue relies more on one-off projects.

  • Operating cash flow remains sharply negative after 2027.

  • Per-share dilution keeps clearly outpacing revenue and gross-margin growth.

  • The reimbursement environment reverses, especially the unit price and coverage scope of oncology testing or genetic screening being compressed.

  • The acquired assets expose clear internal-control or integration problems.

The largest permanent-capital-loss scenario. It is not a short-term halving of the share price, but this company ultimately proving that it can only sustain growth through M&A and financing, and struggles to turn platform scale into stable per-share cash flow. In that case, even if revenue keeps growing, shareholder returns may not be good, and could even produce years of low returns due to long-term dilution and multiple compression.

Comparison, Checklist, and Final Recommendation

Comparison with other opportunities. If you compare Tempus with a broad-based index, my view is that it is not the kind of "certain asset" clearly superior to the index, but a high-volatility, high-divergence, high-execution-risk single-company bet. Against the roughly 4.47% risk-free yield of the 10-year Treasury, Tempus currently has no positive owner-earnings yield, so buying it gets you not a cash return but a long-term option on future high growth and high margins. For some growth investors this is fine; for a long-term value investor, it usually means you must either demand a lower price or wait for cash flow to become clearer.

Checklist

Check item Assessment
Can I understand this business Pass
Does it have long-term stable demand Pass
Does it have a durable moat Uncertain
Does it have pricing power Uncertain
Can it generate stable free cash flow Fail
Is its return on capital excellent Fail
Is management trustworthy Uncertain
Is capital allocation rational Uncertain
Is the balance sheet sound Uncertain
Is the valuation below intrinsic value Fail
Is the margin of safety sufficient Fail
Does long-term holding let me rest easy Fail
Which key facts would make me sell See "signals triggering reassessment" below
Am I buying only because of the price or emotion Needs self-check; emotion-driven decisions not advised now

The above judgments are based comprehensively on the company's business structure, cash flow, financing and governance arrangements, and the degree to which the current valuation front-loads the future.

【Final Rating】 Watch

【One-Sentence Investment Thesis】 Tempus AI owns a rare and attractive prototype of a medical-data platform, but until it proves it can sustainably generate positive owner earnings, the current price lacks a sufficient margin of safety.

【Core Bull Reasons】

  • Diagnostics and Data and applications reinforce each other, and the business logic is genuinely real.

  • Revenue grew strongly in 2025, with 2025 NRR of about 126% and Remaining TCV above $1.1 billion.

  • After acquiring Ambry, Paige, Deep 6, and others, data breadth and product boundaries expanded clearly.

  • The 2026 0% convertible notes replacing high-cost debt is an improvement from a cost-of-capital standpoint.

  • The founder's shareholding and voting power are both high, so management will not easily make short-term decisions on a quarterly basis.

【Core Bear Reasons】

  • Under a conservative basis, owner earnings are still negative, and P/Owner Earnings does not hold.

  • 2025 operating cash flow was still a net outflow of $218 million, and free cash flow was clearly negative.

  • Part of the 2025 growth came from M&A, and the organic quality and integration effect still need observation.

  • Dual-class shares, evergreen incentives, and convertible notes all imply future per-share dilution risk.

  • The current valuation is closer to an "optimistic-expectation price" than a price that "can withstand mistakes."

【Key Assumptions】

  • The customer stickiness and order-expansion trend of Data and applications can be maintained.

  • The integration of Ambry/Paige/Deep 6/OneOme will not drag down cash flow and internal controls.

  • Medicare/commercial-insurance coverage and pricing for key testing products will not deteriorate.

  • The pace of dilution will ultimately be lower than the growth rate of per-share intrinsic value.

【Fair Buy Price】 I would treat $18–28/share as a more attractive entry range; $28–38/share is barely open to discussing a small long-term tracking position; at the current $43.93 I do not think there is a sufficient margin of safety. The basis is that, at the discount rate a value investor requires, only long-term assumptions close to the optimistic scenario can support the current price.

【Target Holding Period】 If bought in the future, it is suitable to view over 5–10 years or even longer; but only after it first crosses the "prove the cash flow" stage. Treating it now as a "buy it and close the account for 5 years" asset is not something I am comfortable with.

【Expected Annualized Return】 This is a subjective estimate based on the current price, not a return promise:

  • Conservative scenario: -5% to 2%

  • Neutral scenario: 4% to 8%

  • Optimistic scenario: 10% to 15% The reason the ranges are not high is that the current price has already front-loaded a considerable part of the success.

【Maximum Loss Risk】 If the market re-prices Tempus from a "future high-margin platform" to a "complex, diluting, high-investment hybrid medical-services company," the central valuation will shift down significantly. In an extreme case, if growth slows, cash flow keeps failing to turn positive, and acquisitions erode value, a drawdown of 50% or more in the share price is not hard to imagine; this does not necessarily mean the company goes bankrupt, but it would mean the investment logic was wrong.

【Tracking Metrics】

  • The NRR of Data and applications.

  • Remaining TCV and its growth rate.

  • Whether operating cash flow and free cash flow keep improving.

  • Diagnostics gross margin and Data gross margin.

  • Changes in share count and the scale of stock-based-compensation grants.

  • The integration progress of Ambry/Paige/Deep 6.

  • Changes in key reimbursement policies and unit prices.

  • Renewals and expansions of major pharma partnerships.

  • The gap between Adjusted EBITDA and GAAP cash flow.

  • Internal controls, audit opinions, and potential restatement risk.

【Signals Triggering Reassessment】

  • NRR clearly falls below 110% and stays there.

  • TCV stalls or retreats.

  • Operating-cash-flow improvement stalls.

  • Dilution significantly outpaces revenue growth.

  • Reimbursement for key tests is compressed.

  • Acquired assets expose internal-control or accounting problems.

  • The company again takes on large debt or issues shares, but per-share value does not improve in step.

【Open Questions and Limitations】 This analysis has three main limitations: first, Tempus has been public for a short time, with limited long-term public reporting history; second, 2024 was strongly distorted by IPO-related stock-based compensation, and 2025 was affected by M&A, lowering year-over-year comparability; third, the estimates of "maintenance capex" and "normalized owner earnings" cannot be as high-precision as for a mature company. These uncertainties themselves are an important reason I do not give a "Buy" rating.

【Final Recommendation】 Coolly put, Tempus AI is worth researching and worth putting on a watch list, but today it looks more like a growth stock that needs to prove itself than a value stock that already shows a clear margin of safety. If you hold to Buffett-style standards, I would advise: wait for a lower price, or wait for clearer cash flow, before acting.

This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.

Precision MedicineMolecular DiagnosticsHealthcare DataAIOncology TestingValue InvestingFree Cash Flow
Reader Q&A10

Baillie Framework · Ten Questions for Growth Investing

10

Hunting ten-year five-baggers among great growth stocks — pressing the upside question: "Can it get much bigger?"

Baillie Framework · Ten Questions for Growth Investing — score profile: 45/100 total Ceiling 5/10 · Revenue 2x 5/10 · Next engine 4/10 · Moat 5/10 · Reinvention 4/10 · Management 6/10 · Customer need 5/10 · Unit economics 5/10 · 5x path 3/10 · Blind spot 3/10 0510 How high is its market ceiling? Is it enlarging a slice of an existing pie, or creating an entirely new market? — 5/10 Ceiling 5 Can its revenue at least double over the next five years? Is growth driven mainly by volume, price, or new businesses? — 5/10 Revenue 2x 5 Five years from now, what will take over as the next growth engine? Does this "second curve" exist today? — 4/10 Next engine 4 What is its core competitive advantage? Will this moat widen or narrow over the next three to five years? — 5/10 Moat 5 If its core business is disrupted, does it have the DNA to reinvent itself? How does it treat mistakes and bad news? — 4/10 Reinvention 4 Does management (especially the founder) have a long-term vision, and interests deeply aligned with the company? Are they willing to sacrifice present profit for five to ten years out? — 6/10 Management 6 If it disappeared tomorrow, how much would customers miss it? Is its way of growing sustainable and free of harm to society and regulation? — 5/10 Customer need 5 What are this business's unit economics (gross margin, incremental returns)? Do they get better or worse as it scales? Where is the money it earns being spent? — 5/10 Unit economics 5 For it to rise five-fold in ten years, which conditions must hold simultaneously? Are they realistic? What expectations does today's price imply? — 3/10 5x path 3 Why has the market not yet realized all this? Is it that it cannot understand, looks down on it, or cannot see far enough? What would become the "narrative inflection point"? — 3/10 Blind spot 3
  • How high is its market ceiling? Is it enlarging a slice of an existing pie, or creating an entirely new market?5/10

    The ceiling genuinely exists, but Tempus is mainly "enlarging and reorganizing a slice of an existing pie" rather than creating an entirely new market from nothing. It stitches three existing pools of demand, diagnostics, clinical data, and AI applications, into one platform; the novelty lies in the "way of combining," while the demand pools themselves have long existed.

    First, how big is the pie and what type is it. Tempus straddles three mature areas of demand: oncology and genetic NGS testing (per the report, already contested by Roche/Foundation, Caris, Guardant, Natera, Quest, and Labcorp), and pharma real-world data and clinical-trial services (already occupied by Flatiron, IQVIA, ConcertAI, ICON, and others). None of these three were pioneered by Tempus; what it does is connect "testing accumulates data, data feeds back into testing" into a reinforcing loop, which the report calls the company's most credible source of a moat. So a more accurate description is: it does penetration and integration within the existing pools of precision medicine, molecular diagnostics, and healthcare data, rather than opening up a no-man's-land. (Note: the competitor list cited in the report's main text comes from its 10-K; here I do not attach possibly-stale search pages, since the competitors' identities can be independently verified.)

    What truly carries the scent of "creating a new market" is the emerging layer of selling multimodal clinical data to pharma companies to train foundation models. This segment is growing fast but the absolute size is still small: 2025 Data and applications revenue was about $316 million, up about 31% year over year, within which data licensing (Insights) grew about 38%; by Q1 2026 this segment reached $87 million in the single quarter, up about 41%, and it has already partnered with 19 of the top 19 pharma companies and 250-plus biotechs, with cumulative signings over $2 billion. This is indeed "building" a new healthcare-AI data-trading market, but today it accounts for only about a quarter of total revenue and cannot hold up the entire valuation.

    My qualitative judgment on the ceiling: the direction is a long, gentle slope, but the card Tempus holds is that of "an integrator in a big market," not that of "defining a new category and owning it exclusively." What Baillie Gifford's LTGG truly wants is the latter kind of market creator that can deliver five-fold returns over ten years; Tempus is closer to the former, with a high ceiling that must be realized by continually taking share in a crowded race and by turning the data layer into a new growth pole, a difficulty and uncertainty clearly higher than a clean new-market monopolist.

    Against the current price. Based on the June 10, 2026 close of $49.48 and market cap of about $8.88 billion (the report's snapshot price was $43.93 with a market cap of about $7.86 billion; the current price is already about 12.6% higher, partly catalyzed by the FDA approval of xT CDx in late May), the market is clearly already pricing in "this new data market will be enlarged by it." The high ceiling is real, but a "high ceiling" does not equal "today's price leaves a margin of safety"; these two things must be viewed separately.

    Jun 11, 2026
  • Can its revenue at least double over the next five years? Is growth driven mainly by volume, price, or new businesses?5/10

    Doubling revenue within five years is very likely achievable, and the bar is not even that high, but note that a good part of this "doubling" is propped up by M&A consolidation; the quality of organic growth is what really needs watching.

    First the arithmetic. 2025 revenue was $1.272 billion, up about 83.4% year over year, and Q1 2026 already reached $348 million, up about 36.4%. Even if the growth rate steps down from 36% all the way to the low double digits, achieving "at least a double" cumulatively over five years (i.e., annualized above about 15%) is a low-bar goal. The report actually acknowledges this; its disagreement was never about "whether it can keep growing," but about "whether this growth can turn into free cash flow." So on this Q2, my conclusion leans positive, but the reasoning has to land on the drivers.

    Breaking down the three drivers: volume is the main force, price is an add-on, and new businesses are the optionality.

    Volume: test volume on the diagnostics end is the hardest engine. In 2025 oncology test volume grew about 26% and genetic test volume about 29%, and Diagnostics full-year revenue was about $955 million, up about 111% year over year (including the Ambry genetic business consolidated in February 2025). Excluding M&A, double-digit growth in organic test volume is a sustainable demand base.

    Price: in late May 2026 xT CDx received FDA approval, covering 100% of the DNA test panel, which the market read as opening room for ASP (average selling price per test) uplift and reimbursement improvement. This is a rare "price-up" logic rather than pure volume. But the report also cautions that pricing power on the diagnostics end is not fully in its own hands and is constrained by third-party reimbursement.

    New businesses: data and AI applications are where the imagination lies, but the absolute size is still small. In 2025 Data and applications was about $316 million, up about 31% year over year, with data licensing up about 38%; in Q1 2026 this segment was $87 million in the single quarter, up about 41%. It has the fastest growth but accounts for only about a quarter of total revenue, so within five years it is more likely to be "an icing-on-the-cake accelerator" than something that doubles the whole pie on its own.

    In one sentence, back to the Baillie Gifford lens: a double can be achieved, but "getting big by M&A" and "getting strong organically" are two completely different qualities of doubling. The report repeatedly stresses that "a large part of the 2025 growth came from Ambry consolidation" and that the organic quality still needs observation; this is precisely the key to why the rating stops at "Watch" rather than crediting it as a growth stock. Based on the current $49.48 and market cap of about $8.88 billion, the market has already front-loaded a "high-quality double" into the price, leaving little buffer for surprises.

    Jun 11, 2026
  • Five years from now, what will take over as the next growth engine? Does this "second curve" exist today?4/10

    The second curve is already visible in embryonic form today: it is the layer of "selling multimodal clinical data to pharma companies to train foundation models + AI applications." But right now it looks more like an extension of the main curve than an independent new growth pole that can stand on its own.

    First, what this curve is and whether it exists today. It does exist and is accelerating: in Q1 2026 Data and applications was $87 million in the single quarter, up about 41% year over year, and it has already partnered with 19 of the top 20 pharma companies and 250-plus biotechs, with cumulative signings over $2 billion; the report also states that as of the end of 2025 Remaining Contract Value (TCV) exceeded $1.1 billion and Net Revenue Retention was about 126%, indicating existing customers keep expanding orders. A more specific inflection signal is that in 2025 Tempus partnered with AstraZeneca and Pathos to jointly build a multimodal foundation model in oncology, bringing an additional $200 million in data-licensing and model-development fees; this is precisely the concrete landing point of the "data asset → AI model → higher value-add" second curve.

    But I must honestly mark it down a notch. A "second curve" in the Baillie Gifford sense refers to a new engine that can take over and independently support the next decade; whereas Tempus's data layer currently: (1) accounts for only about a quarter of total revenue and is too small to take the baton; (2) essentially depends on the data continuously accumulated on the diagnostics end and is not a business that can run independently of the main business; (3) the report points to a fatal reverse risk, that the company admits in its 10-K that "the development of foundation models could in the future weaken customer demand for its de-identified data products." In other words, this second curve is both its greatest source of imagination and the carrier of a self-disruption fuse.

    Contrasting it with a true platform-type second curve makes it clearer: Schrodinger and Recursion are the type where "the platform capability itself can spill over into multiple product lines"; Tempus's data layer looks more like "a high-value-add monetization outlet for the main curve," with the right direction and high margins (the data business's margin is structurally higher than diagnostics), but its independence and certainty are not yet proven.

    Back to valuation and cadence. Anchored on the current $49.48 and market cap of about $8.88 billion and the report's snapshot of $43.93, the current price already implies the optimistic assumption that "the data/AI second curve will grow smoothly into a high-margin engine." The report's stance is restrained: constructive on direction, but unwilling to pay a premium for this not-yet-solidified second curve until it truly lands high margins into per-share cash flow rather than staying at the Adjusted EBITDA level. I agree; the second curve is visible, but "visible" is still several years of realization away from "able to take the baton."

    Jun 11, 2026
  • What is its core competitive advantage? Will this moat widen or narrow over the next three to five years?5/10

    The core competitive advantage is "multimodal medical-data scale + the ability to structure and connect diagnostic results with clinical data + the diagnostics-to-data two-way reinforcing loop," rather than brand or extreme low cost. Over the next three to five years, I judge the moat leans toward "widening," but I can only give medium confidence; it is a moat in the process of forming, not yet proven able to be wide for ten years. The report's moat-strength score is 3/5, and I concur.

    First, the real landing point of the advantage. In its 10-K Tempus describes its core selling point as making testing "get smarter the more it is done," and explicitly says a network effect exists between the diagnostics and data lines. This loop has quantifiable traces: in 2025 the Data business's NRR was about 126% and Remaining Contract Value (TCV) exceeded $1.1 billion, showing the data asset is not for show and existing customers keep expanding orders; by Q1 2026 it had covered 19 of the top 20 pharma companies and 250-plus biotechs, with cumulative signings over $2 billion. Add its long-term contract with GSK carrying a minimum committed purchase of about $180 million through 2027 (with a further option of up to about $120 million during 2028–2030), and switching costs and data scale, these two, do have an embryonic form.

    Why still only 3/5, and why I dare not say "unbreakable." The report gives three very hard counterpoints, and I agree with each: first, the diagnostics race itself is intensely competitive, and the rivals Tempus lists are heavyweight (Roche/Foundation, Caris, Guardant, Natera, Quest, Labcorp); second, the pharma-data business is not its exclusive domain, with Flatiron, IQVIA, ConcertAI, and others in the same profit pool; third, and most fatal, the company admits in its risk factors that "the development of foundation models could in the future weaken customer demand for its de-identified data products," meaning the economic rent of the moat could be self-eroded by technology. A data moat that "even the company itself admits could be hollowed out by upstream models" clearly falls short of the "wide enough for five-fold returns over ten years" purity that Baillie Gifford wants.

    The widening thrust is also real. The 2025 acquisitions of Ambry (genetics), Paige (digital pathology), Deep 6 (trial matching), and OneOme (pharmacogenomics) broadened the database and capability boundaries; in late May 2026 xT CDx received FDA approval, covering 100% of the DNA test panel, thickening the regulatory-qualification barrier too. But the report makes a key remark: acquisitions may just be "piling more assets together" without necessarily forming a higher unit-economic return; a widening moat and a bigger balance sheet are not the same thing.

    Conclusion and valuation implication. Over three to five years the moat will most likely widen at the margin, but "widening" and "wide enough to support the current valuation" are two different thresholds. Anchored on the current $49.48 and market cap of about $8.88 billion and the report's snapshot of $43.93, the market has already priced in "the data moat will keep widening and translate into high margins"; and whether the moat can truly translate into profit rather than mere scale is exactly the item the report leaves as the biggest uncertainty.

    Jun 11, 2026
  • If its core business is disrupted, does it have the DNA to reinvent itself? How does it treat mistakes and bad news?4/10

    The "DNA" for self-reinvention is partly present but has not yet been tested by a true life-or-death crisis; on treating mistakes and bad news, the company is candid in its disclosures, but the governance structure means external self-correction is weak. Putting these two together, my conviction on "whether it can save itself when the core business is disrupted" can only be neutral-to-cautious.

    First, fill in the implicit premise of this chained question, what counts as "the DNA to reinvent itself." It refers to whether, when the main business (here, diagnostic testing) is disrupted by new technology or a new reimbursement paradigm, the company has the underlying ability to "change the engine without changing its life." Tempus's favorable evidence is that it is a "platform + data" type rather than a single-product type: diagnostics accumulate data, data feed back into algorithms, and this loop in theory allows it, when testing is priced down, to shift its center of gravity toward monetizing data and AI applications. Its moves over the past two years do carry the flavor of "actively reconstructing capability boundaries," with successive acquisitions of Ambry, Paige, Deep 6, and OneOme and its partnership with AstraZeneca and Pathos to jointly build an oncology multimodal foundation model (an additional $200 million in data-licensing and model-development fees). This shows it does not cling to a single testing product and has the willingness and resources to migrate toward the data/AI layer.

    But the quality of this "DNA" must be discounted, for two reasons. First, its most likely disruption comes from the fuse it itself admits: the company states in its 10-K that "the development of foundation models could in the future weaken customer demand for its de-identified data products." That is, if disruption really comes from AI foundation models, what it needs to reinvent is exactly the second curve it pins high hopes on, and its room for self-rescue would be squeezed by the same force. Second, so far all of its "reinvention" has been done in a tailwind environment of expansion via financing and M&A, without a real stress test of "a cliff-like contraction of the main business forcing it to cut off an arm to survive," so whether the DNA is truly tough still lacks a real battle.

    Now how it treats mistakes and bad news, which bears on management honesty. There are quite a few positive signals: the report shows the company proactively discloses fragilities in its 10-K (reliance on a single supplier for lab materials/equipment/services, rising acquisition-integration complexity) and candidly admits that in 2025 management and auditors excluded Ambry and Paige from the internal-control assessment, together accounting for about 11% of total assets and 29% of total revenue. Saying openly "we have not yet fully taken over internal control of these two" is candor, not concealment.

    But the external constraints of the error-correction mechanism are weak. Founder Eric Lefkofsky holds all Class B shares, for about 57.9% combined voting power, with directors and officers as a group at about 58.3% voting power (economic interest only about 22.1%). Strong interest alignment is a plus, but it means that even if decisions go wrong and integration falters, external shareholders have almost no leverage to force a correction; "admitting mistakes" relies more on the founder's own conscience than on governance enforcement.

    Back to the Baillie Gifford lens and the price. Baillie Gifford favors teams that "generously admit errors and can regenerate"; Tempus passes on "admitting," has potential but is battle-unproven on "regeneration ability," and is weak on "external correctability." Based on the current $49.48 and market cap of about $8.88 billion (report snapshot $43.93), the market assumes it has enough resilience to cross a disruption; this trust currently rests more on narrative than on a track record.

    Jun 11, 2026
  • Does management (especially the founder) have a long-term vision, and interests deeply aligned with the company? Are they willing to sacrifice present profit for five to ten years out?6/10

    The founder's long-term vision is strong, interest alignment is deep, and he is clearly willing to sacrifice present profit for five to ten years out; on this item Tempus is actually a relative plus. But "deep alignment" also brings governance minuses (dual-class shares + strong dilution), and the two must be stated together. The report's management and capital-allocation score is 3/5, and I agree with this neutral-to-positive positioning.

    First, the depth of alignment, which is a solid strength. Eric Lefkofsky is the founder, chairman, and CEO. Per the 2026 proxy, he holds about 21.4% of the economic interest and all Class B shares, for about 57.9% combined voting power; directors and officers as a group hold about 22.1% economic interest and about 58.3% voting power. Baillie Gifford especially values founders who "put their net worth on the boat and do not trade on a quarterly basis"; Tempus is a good in-paradigm sample here: a founder holding two-tenths of the economic interest is almost incapable of sacrificing a decade of positioning for one quarter's statements.

    Next, "whether willing to sacrifice present profit for the long term," and the evidence is very direct. The main business is still loss-making: in 2025 operating cash flow was a net outflow of about $218 million, and in Q1 2026 there was again a net outflow of about $73.3 million, with a single-quarter net loss of $125.9 million. This money is spent mainly on reinvestment, M&A (Ambry, Paige, Deep 6, OneOme), and the data/AI platform, rather than paying dividends for show. In other words, management's "trading continuous losses for platform scale" is itself clear proof of sacrificing present profit for five to ten years out, exactly the trade-off Baillie Gifford likes to see.

    But the flip side of deep alignment is the governance cost, and this is where points must be deducted. First, dual-class shares make external-shareholder governance rights very weak: the founder controls nearly six-tenths of the vote with about two-tenths of the economic interest, minority shareholders have almost no restraining power, and the report bluntly says this "significantly reduces external constraints." Second, dilution is a very real problem: at the end of 2025 there were 9.061 million granted but unsettled awards, 5.895 million shares still available to grant, plus the 2024 Equity Incentive Plan's evergreen mechanism auto-replenishing at 5% of total shares and the ESPP auto-increasing at 2% of total shares. Weighted-average shares have already risen from about 119.8 million in 2024 to about 174.3 million in 2025; between "the company getting bigger" and "per-share value getting bigger," a gap has been opened by ongoing dilution.

    On capital allocation I give a neutral score. The rational side: in May 2026 issuing $460 million of 0% convertible notes to replace high-cost Ares debt is smart from an interest-cost standpoint; the aggressive side: high-frequency M&A while free cash flow has not yet turned positive (Ambry consideration of $600 million, of which $375 million cash + $225 million stock) amplifies the uncertainty of integration and returns.

    Conclusion: on the management dimension, Tempus is a clear plus on "long-term vision + interest alignment + willingness to sacrifice present profit," and a clear minus on "shareholder-friendliness + dilution discipline." Baillie Gifford would appreciate the former but also watch the latter closely; based on the current $49.48 and market cap of about $8.88 billion (report snapshot $43.93), whether per-share intrinsic value can outrun dilution is the metric on this company's governance side that most deserves ongoing monitoring.

    Jun 11, 2026
  • If it disappeared tomorrow, how much would customers miss it? Is its way of growing sustainable and free of harm to society and regulation?5/10

    If Tempus disappeared tomorrow, pharma data customers would miss it more than diagnostics customers, but neither side counts as "irreplaceable"; and its way of growing is on the whole constructive and does not rely on harming society, with the main sustainability risk coming from the data-privacy/reimbursement-compliance regulatory line rather than any original sin in the business model. This question must be split into "indispensability" and "social/regulatory sustainability."

    First, indispensability, viewed by customer type. Pharma/biotech is the stickier side: the de-identified multimodal clinical-molecular-imaging data that Tempus accumulates is an asset with historical structure and integration processes, and in 2025 the Data business's NRR was about 126% and Remaining Contract Value exceeded $1.1 billion, already embedded in the R&D processes of 19 of the top 20 pharma companies and 250-plus biotechs; switching suppliers means rebuilding data pipelines and models, so migration costs are not low and they would "miss it fairly much." The diagnostics side is far more replaceable: physicians want accurate, timely, reimbursable test reports, and Roche/Foundation, Caris, Guardant, Natera, Quest, and Labcorp can all take the orders; if Tempus disappeared, such demand would be quickly diverted to peers, and the clinical end's "missing it" would be limited.

    But "missing it" must take a key discount, since the reverse risk the report points to directly weakens indispensability: the company admits in its 10-K that "the development of foundation models could in the future weaken customer demand for its de-identified data products." If pharma companies can ultimately use general foundation models to replace buying data from Tempus, then even the stickiest side would loosen. So a more accurate judgment is: today it has upper-medium indispensability, but it is not sole-source (exclusively irreplaceable) and faces the long-term threat of being hollowed out by upstream technology.

    Now the social/regulatory sustainability layer. The business model itself is constructive: helping physicians make precise diagnoses, helping pharma companies accelerate drug development and clinical-trial matching, and using AI to assist oncology diagnosis and treatment; growth comes from "making medicine more precise and efficient," not the kind of business that profits by harming users or society. On this point it is cleaner than many "regulatory arbitrage" companies.

    The real constraints on sustainability lie on two regulatory lines. First, data privacy and compliance: the secondary commercial use of de-identified medical data is subject over the long term to privacy frameworks such as HIPAA and to public perception, and tightening regulation would directly compress the space for data monetization. Second, reimbursement sustainability: the diagnostics end "relies heavily on third-party payer reimbursement, and pricing power is not fully in its own hands" (the report's words); once the coverage scope and unit price of key oncology tests or genetic screening are compressed, the foundation of growth would be shaken. Conversely, the late-May 2026 FDA approval of xT CDx, covering 100% of the DNA test panel, is precisely a favorable development on this regulatory line, showing it is currently moving with compliance rather than against regulation.

    Conclusion (Baillie Gifford lens): Tempus's way of growing withstands the question of "does it harm society," qualifying as constructive growth; but its "indispensability" only reaches upper-medium and is non-exclusive, and the three lines of regulation/reimbursement/privacy are the real gates on its sustainability. Based on the current $49.48 and market cap of about $8.88 billion (report snapshot $43.93), the market is mostly pricing in "data stickiness = strong indispensability," which is exactly where the company itself has marked a fuse.

    Jun 11, 2026
  • What are this business's unit economics (gross margin, incremental returns)? Do they get better or worse as it scales? Where is the money it earns being spent?5/10

    Unit economics are improving and the gross-margin structure is not bad, but the test of "whether the money earned can turn into free cash flow" has so far not been passed; this is the dimension where Tempus most deserves a discount. Getting bigger has indeed brought gross-margin improvement, but cash flow is still negative, and the money is spent mainly on reinvestment, M&A, and ongoing stock-based compensation. The report judges both "can generate stable free cash flow" and "is return on capital excellent" as "Fail," and I agree.

    First, gross margin, which is truly getting better. In 2025 GAAP gross profit was about $797.9 million, up 109.4% year over year, against about $1.272 billion of revenue for a gross margin of about 62.7% (the report's basis), clearly up from about 55.0% in 2024. By segment it is healthier: diagnostics (genomics) gross margin is a bit above 60%, and the data/services business's margin is structurally higher (around 70%), which means the higher the share of Data and applications, the better the blended gross margin. So the chain of "getting bigger → gross margin improves" is currently supported by the evidence.

    But the incremental-returns step is stuck; the gross-margin improvement has not transmitted to cash. The key facts: for full-year 2025 Adjusted EBITDA was still -$7.4 million (an improvement of about $97.3 million year over year, with Q4 already turning positive to +$12.9 million), but the harder cash measure is still negative: 2025 operating cash flow was a net outflow of about $218 million, and the report estimates free cash flow at about -$245 million from capex + capitalized software; by Q1 2026 operating cash flow was again a net outflow of about $73.3 million. In other words, for every extra dollar of revenue earned, it has not yet steadily turned into a positive dollar of cash. The "excellent incremental returns" Baillie Gifford wants can here only count as "right direction, not yet solidified."

    Where the money goes, three places. First, reinvestment in the main business (labs, cloud, R&D, sales), the root of why its growth is "capital-hungry"; the report stresses it "is not heavy industry, but is by no means a SaaS that effortlessly turns revenue into cash." Second, M&A; in 2025 net cash outflow for M&A reached about $377 million (Ambry, Paige, Deep 6, OneOme). Third, stock-based compensation paid in the form of dilution; the 2025 net loss of $245.0 million includes about $136.3 million of stock-based compensation and related payroll taxes. From the Baillie Gifford lens, the first two are "spending for the future" and acceptable, but the third is essentially a cost shareholders bear through dilution and cannot be mechanically added back as profit; this is exactly why the report insists that "owner earnings are still clearly negative."

    Will it get better as it scales? Most likely yes, but with a question mark. The positives are that Q4 EBITDA has turned positive and management has given 2026 guidance of about $1.59 billion in revenue and Adjusted EBITDA of about +$65 million, pointing toward scale diluting fixed costs and gross margin continuing to climb. But there are two caveats: first, Adjusted EBITDA turning positive does not equal free cash flow turning positive, with stock-based compensation and working-capital occupation in between; second, piling up volume via M&A could dilute unit-economic returns, and the report cautions that "widening the moat" and "improving unit economics" are not the same thing.

    Conclusion (Baillie Gifford lens): decent gross margin, emerging scale effects, and clear 2026 guidance for an earnings inflection are the highlights of this dimension; but "the money earned cannot turn into free cash flow, and a large part is paid via dilution" is the hard flaw. Based on the current $49.48 and market cap of about $8.88 billion (report snapshot $43.93), the market is already pricing in "unit economics will keep getting better all the way to per-share cash flow," which is exactly the link the company has least proven so far.

    Jun 11, 2026
  • For it to rise five-fold in ten years, which conditions must hold simultaneously? Are they realistic? What expectations does today's price imply?3/10

    For Tempus to rise five-fold from the current price in ten years, five things must hold at once: "revenue keeps growing fast + terminal margin lifts significantly + free cash flow truly turns positive and stays sustainable + dilution is held down + the valuation multiple does not compress sharply." Each one alone is not outrageous, but requiring them to be realized continuously and simultaneously for ten years is very hard. The report's core conclusion is exactly that the current price is closer to an "optimistic-expectation price" than a "margin-of-safety price," and I agree.

    First, translate "five-fold" into a concrete target. Anchored on the June 10, 2026 close of $49.48 and market cap of about $8.88 billion (report snapshot $43.93, market cap about $7.86 billion, current price already about 12.6% higher), a five-fold gain in ten years means the market cap must reach about $44 billion and the share price about $247. Compared with today's sizes of peers, Natera at about $32 billion market cap and Guardant at about $12.8 billion, Tempus would have to grow within ten years into a precision-medicine-plus-data giant nearly 40% bigger than today's Natera, and by then already be a stable, profitable cash machine. This is not impossible, but it belongs to the "optimistic scenario."

    Weighing the reality of the five conditions that must hold simultaneously:

    First, revenue sustains high compound growth for ten years. 2025 revenue was about $1.272 billion, up 83.4% year over year, and 2026 management guidance is about $1.59 billion (about +25%). Growth will naturally step down; to support five-fold, the report's optimistic scenario assumes a ten-year revenue CAGR of about 15%, achievable but requiring the data/AI second curve to truly scale.

    Second, the terminal margin lifts significantly, which the report names as "the most fragile valuation assumption." The market is betting that "after combining data + AI + diagnostics, the long-term margin is clearly higher than a traditional lab"; if it ends up merely "a bit more software than a traditional lab, still dragged by reimbursement and sales investment," then a price-to-sales above 6x lacks support. This condition is the most subjective and the most likely to fall through.

    Third, free cash flow truly turns positive and stays sustainable. This is the least-proven link today: in 2025 operating cash flow was still a net outflow of about $218 million, Q4 Adjusted EBITDA had just turned positive to +$12.9 million, and full-year EBITDA was still -$7.4 million; 2026 guidance is Adjusted EBITDA of about +$65 million. But Adjusted EBITDA turning positive does not equal free cash flow turning positive, with stock-based compensation and working-capital occupation in between, and it must be solidified for consecutive years to count as passed.

    Fourth, dilution is held down. Dual-class shares + evergreen incentives (auto-replenishing at 5% of total shares) + ESPP (2%), and weighted-average shares have already risen from about 119.8 million in 2024 to about 174.3 million in 2025. If per-share value growth cannot outrun dilution, "the company rising five-fold" and "shareholders earning five-fold" will decouple.

    Fifth, the valuation multiple does not compress sharply in the process. The current price corresponds to a 2025 EV/Sales of a bit over 6x and about 5.8x on the 2026 guidance revenue, already growth-stock pricing, with almost no "multiple-expansion" dividend to borrow; a five-fold return basically has to be carried by fundamentals, and the multiple had best not shrink.

    What expectations does today's price imply? Putting the five conditions together, the current price already front-loads a "near-optimistic-scenario" set of assumptions; among the report's three Owner Earnings discounting scenarios, only the optimistic tier (revenue CAGR of about 15%, Owner Earnings margin eventually reaching 18%) gives an intrinsic value (about $45–60/share) that covers the current price; the conservative tier (about $8–15) and neutral tier (about $20–30) are both clearly below $49.48. In other words, the market assumes it will follow the optimistic script all the way, leaving an extremely thin buffer for "a little error."

    Conclusion: five-fold in ten years is not a fairy tale, but it requires five conditions to be realized simultaneously and continuously, of which "terminal margin" and "sustainable positive free cash flow" have not been proven so far. At the current price, the margin of safety is insufficient; this is exactly why the report stops the rating at "Watch" and places the more attractive entry range at $18–28.

    Jun 11, 2026
  • Why has the market not yet realized all this? Is it that it cannot understand, looks down on it, or cannot see far enough? What would become the "narrative inflection point"?3/10

    For Tempus this question must be answered in reverse: the market has not "failed to realize its merits." On the contrary, it is a fully-followed star stock, recently even chased up on good news. The real cognitive divergence is not that "the market underestimates its potential," but "the time gap between the optimistic narrative and the cash-flow proof." That is, today's more likely bias is that the market sees it "too near, too optimistically," rather than "cannot understand, looks down on it, or cannot see far enough."

    First, prove it is not a neglected obscure stock. Tempus has high coverage and good liquidity, with a June 10, 2026 close of $49.48 and market cap of about $8.88 billion, already up about 12.6% from the report's snapshot of $43.93, and this rally has a clear catalyst, the May 29 FDA approval of xT CDx, covering 100% of the DNA test panel, to which the market quickly gave positive feedback. A stock that gets bought up the moment news breaks clearly cannot be described as "looked down on" or "understood by no one." So the honest answer here to Baillie Gifford's classic "is the market failing to understand, looking down, or unable to see far" is: none of the above, the market basically understands it and is even a bit generous.

    So where exactly is the report's disagreement? In "the speed at which the narrative is believed" far exceeding "the speed at which cash flow is proven." The market is already pricing in "data + AI + diagnostics will combine into a high-margin platform" (current price corresponds to a 2025 EV/Sales of a bit over 6x), while the company so far still has full-year operating cash flow of a net outflow of about $218 million and full-year Adjusted EBITDA still -$7.4 million (only just turned positive in Q4). The report's stance is not "the market underestimates, I am bullish," but "the market has already front-loaded optimism into the price, the margin of safety is insufficient," a cautious "Watch" conclusion demanding a lower price or clearer cash flow, not a "wrongly-killed bargain."

    If one insists on finding "a direction the market may be wrong about," it is more likely a risk point of overestimation than underestimation: first, the market may underestimate the reverse risk the company itself admits, that the development of foundation models could in the future weaken customer demand for its de-identified data products (the second curve's built-in fuse); second, the market may underestimate the drag of acquisition integration and ongoing dilution on per-share value (the 2025 internal-control assessment excluded Ambry and Paige, together accounting for about 11% of total assets and 29% of total revenue). These two are where "the market may not see far enough," but the direction is both "risk underestimated," not "value underestimated."

    What would become the narrative inflection point (both ways)? Upward inflection: operating cash flow steadily turns positive for consecutive quarters, and free cash flow turning positive no longer depends on one-off working-capital swings; the Data business's NRR stays high (currently about 126%) while TCV keeps growing; acquisition gross-margin gains truly land in per-share cash flow rather than staying at Adjusted EBITDA; the pace of dilution slows markedly. Once these are solidified, the "cash-burning platform" narrative gets replaced by "a money-making healthcare-AI platform," and the valuation anchor is truly firmed up. Downward inflection: NRR falls below 110% and stays there, TCV stalls, cash-flow improvement stalls, or pharma companies start using general foundation models to replace buying its data; any one would knock the valuation from "future high-margin platform" back to "complex, diluting, high-investment hybrid medical-services company."

    Conclusion (Baillie Gifford lens): this is not a dust-covered stock whose "greatness the market has not yet realized," but a high-attention growth stock whose "narrative is already fully priced, just waiting for cash flow to deliver." The real inflection lies not in when the market "discovers" it, but in when the company produces sustainable positive owner earnings to deliver on the optimism already front-loaded; before that, the current price ($49.48) looks more like an optimistic-expectation price than a margin-of-safety price.

    Jun 11, 2026
Ask about this report

Members can ask about this report; once answered it appears under "Reader Q&A" on this page. You can also highlight a passage in the text to ask about it directly.