Tempus AI, Inc. (NASDAQ: TEM) — The Data Business Is Real; The Roll-Up, the Related Parties, and the Cash Burn Are Too

Date: June 26, 2026 | Independent equity research — initiation Price at analysis: ~$54.87 (close 2026-06-25) | Market cap: ~$9.8B | Enterprise value: ~$10.5B Shares: 173.7M Class A (1 vote) + 5.0M Class B (30 votes, all Lefkofsky) | FY-end: December

⚡ Claude’s Take

This block is the author’s own independent opinion and general information only — not investment advice. The analysis in the sections below takes no position and carries no price target; it discusses valuation only as embedded expectations.

Verdict: HOLD / AVOID-at-this-price for a quality-seeking fundamental investor; not a short. A genuinely interesting business wrapped in a governance package and a balance sheet I don’t want to underwrite at ~6.6x forward sales. Accumulation zone for the patient growth investor: low-$40s to high-$40s (≈5x forward sales), where the spring-2026 panic priced it. Conviction: medium.

Tempus is two companies stapled together. One — the Data and applications business — is rare and possibly special: a 500-petabyte, clinically-linked, de-identified oncology dataset that the world’s largest drug makers (AstraZeneca, GSK, BMS, and now Merck and Gilead) are signing $100M+ multi-year deals to license and build models on. That business grew ~31–44% in 2025–26, carries software-like margins, and shows early signs of a real data-scale/switching-cost moat. The other — Diagnostics — is a fast-growing but reimbursement-dependent, capital-and-labor-intensive lab business competing head-on with Foundation Medicine, Guardant, Caris and Natera, where the 2025 revenue explosion (+111%) was mostly the Ambry Genetics acquisition, not organic genius. The bull pays one blended ~6.6x-forward-sales multiple for both; the bear notes you’re paying a data-company multiple for a business that is ~75% lower-quality diagnostics.

What keeps me on the sidelines rather than long: (1) it still burns cash — ~$220M operating cash outflow and ~$240M negative FCF in 2025, funded by a $750M convertible and an at-the-market equity program, with GAAP profitability still unproven; (2) governance I have to hold my nose at — 30:1 supervoting control by a serial-monetizer founder, a “controlled company” exemption, and a dense web of related-party arrangements (Pathos, where AstraZeneca pays Tempus $35M and Tempus pays Lefkofsky-affiliated Pathos $35M, which can be settled in Pathos’s own preferred stock; plus the SoftBank “SB Tempus” Japan JV) that make the highest-margin revenue line harder to take at face value; and (3) valuation with no margin of safety — at ~6.6x forward sales on a company that is GAAP-lossmaking, you are underwriting flawless execution and durable ~25% growth for years. The factor tape agrees this is speculative, not value: beta 2.7, a SmallSize loading north of 2, a -59% drawdown over the past year and a round-trip from $40 IPO → $104 → $42 → $55. This is abandoned speculative growth bouncing off a floor, not a compounder on sale.

I’d own it cheaper, smaller, and only if I believed the data moat. The single piece of evidence that flips me bullish: a clean, organic, GAAP-cash-generative quarter with the AstraZeneca MSA renewed/expanded on arm’s-length terms and Data bookings durable through 2027 — proof the moat funds itself. The single piece that flips me bearish: an AstraZeneca/large-pharma non-renewal, a Data-revenue or related-party restatement, or the convert/ATM treadmill forcing dilution into a down tape. Tag: “Great dataset, dear price, founder’s casino.”

📈 Stock Price Action — Five-Year Event Map

Tempus has only traded publicly since June 2024, so this is a two-year, not five-year, price history — but it is a textbook IPO-mania round-trip. The stock priced at $37, opened at ~$40, sank to an all-time low of $22.89 within two weeks (June 2024), then ran ~4.5x to an all-time high of $104.32 (Oct-9-2025) on AI-data enthusiasm and the Ambry-fueled revenue surge, collapsed ~60% to a 52-week low of $41.73 (Mar-30-2026) in the early-2026 growth-stock drawdown, and has since recovered ~31% to ~$54.87, leaving it ~47% below its peak. It sits below its 200-day EMA (~$57) but above its 50-day (~$50). The moves are Facts; the attributed drivers are Interpretation.

Event 1 — the IPO immediately broke issue and bottomed at $22.89, a reminder that the market’s first vote was skeptical. Event 2–3 — the run to $104 conflated two things: real Data-business traction and an acquisition (Ambry) that mechanically doubled Diagnostics revenue, which many extrapolated as organic. Event 4 — the -60% unwind coincided with the broad 2025–26 unprofitable-growth de-rating and pointed short reports questioning data-revenue quality and related-party disclosures. Event 5 — the recent bounce is fundamentally earned (a genuine beat-and-raise with two marquee new pharma logos), but leaves the stock still ~half off its high and squarely in speculative-growth territory.

1. Executive Summary

Tempus AI is a Chicago-based precision-medicine company founded in 2015 by serial entrepreneur Eric Lefkofsky (Groupon co-founder). It operates two product lines. Diagnostics (FY2025 revenue $955.4M, 75% of total) is next-generation-sequencing genomic testing — oncology therapy-selection (solid-tumor xT, liquid-biopsy xF, RNA xR), hereditary/germline testing (xG plus the acquired Ambry Genetics), minimal-residual-disease (MRD) testing, and a suite of AI “algos.” Data and applications (FY2025 $316.4M, 25%) licenses de-identified, clinically-linked multimodal data and AI-modeling services to pharmaceutical and biotech customers, plus clinical-trial matching (Trials) and analytics (Insights, Lens, Next).

The investment debate is sharp. Bulls see a one-of-a-kind data flywheel: clinical testing generates proprietary linked clinical-molecular-imaging data, which is licensed to pharma at software margins, which funds more testing — a two-sided network with a widening data-scale moat, ~25% guided revenue growth for three years, gross margins that have climbed from 32% (2021) to 63% (2025), and a path to meaningful adjusted-EBITDA inflection (FY2026 guide ~+$65M). Bears see a cash-burning roll-up: ~$220M of negative operating cash flow in 2025, a 2025 revenue “surge” that was overwhelmingly the Ambry acquisition rather than organic growth, a founder-controlled company with 30:1 supervoting and a dense related-party web (Pathos, SB Tempus), heavy reliance on a handful of large pharma data contracts (AstraZeneca’s minimum commitment runs only through 2026), and a balance sheet leaning on a $750M convertible and an ATM equity program.

Both readings contain truth. The Data business is the asset that could justify a premium; the Diagnostics business is a competitive, reimbursement-dependent grind that happens to throw off the raw material the Data business monetizes. At ~$54.87, Tempus trades at ~6.6x forward (FY2026E) revenue and ~8x trailing — rich for a GAAP-lossmaking company, though near the cheapest end of its own short trading history (near the cheapest end of its short post-IPO price-to-sales history). The market is underwriting durable ~25% growth, sustained margin expansion, an intact AstraZeneca relationship, and clean conversion of bookings to cash. This analysis takes no position and sets no price target; it lays out what must be true for each side. The competitive-advantage verdict: a credible, still-unproven moat in Data; a weak-to-average competitive position in Diagnostics; economics that are improving but not yet self-funding.

2. Business Overview

Tempus describes itself as a “healthcare technology company” applying AI to precision medicine. The unifying asset is its Platform — the infrastructure that ingests, structures, and harmonizes multimodal patient data (clinical/EHR, molecular/genomic, imaging, and outcomes) and layers analytics and machine-learning models on top. Two commercial product lines sit on that platform.

Diagnostics (FY2025 $955.4M; FY2024 $451.7M; FY2023 $363.0M). This is the lab-testing business and the larger revenue line. Sub-businesses:

• Oncology therapy selection — comprehensive genomic profiling to guide cancer treatment. xT (solid-tumor DNA panel, FDA-approved), xF (liquid biopsy / circulating-tumor DNA, FDA submission pending), and xR (RNA/transcriptomic). Oncology unit volumes grew ~28% year-over-year in Q1-26; average selling price (ASP) sits around $1,720–$1,740, with management guiding to roughly +$500 of incremental ASP over the next one-to-two years as additional FDA approvals and ADLT (Advanced Diagnostic Laboratory Test) pricing land.
• Hereditary / germline testing — xG plus Ambry Genetics, acquired in February 2025, which dramatically expanded Tempus’s hereditary-cancer and broader germline volumes. Ambry is the principal reason Diagnostics revenue more than doubled in 2025.
• Minimal residual disease (MRD) — tumor-informed cancer-recurrence monitoring, ~97% delivered via a partnership with Personalis; growing off a tiny base (~6,500 tests in Q1-26, +~500% YoY) and deliberately metered until reimbursement improves.
• Algos — algorithmic/AI tests (e.g., Homologous Recombination Deficiency, Tumor Origin, immune-profile scores) with a ~40% attach rate on solid-tumor orders, a differentiator versus pure-sequencing competitors.

Data and applications (FY2025 $316.4M; FY2024 $241.6M; FY2023 $168.8M). Higher-margin, software-like:

• Insights — licensing of de-identified, clinically-linked multimodal data (500+ petabytes) and AI-modeling services to pharma/biotech. Anchor customers include AstraZeneca, GSK, Bristol-Myers Squibb, with Merck (a “very large strategic” deal) and an expanded Gilead relationship added in Q1-26. The business now has “almost half a dozen” customers at the $100M+ total-contract-value tier.
• Trials — clinical-trial matching for pharma sponsors.
• Next / Lens / Hub — analytics, data-access, and ordering platforms.

Business model / revenue recognition. Diagnostics is largely transactional (revenue on delivery of test results, reimbursed by Medicare, commercial payers, and self-pay). Data is contractual/subscription-like, often multi-year, frequently with prepayments and deferred revenue that convert to cash over time. The strategic logic — explicitly stated in the 10-K — is network effects: each product line “is designed to enable and enhance the other,” with testing generating the data that powers the licensing business, which in turn funds and informs better tests.

Verdict: A genuinely differentiated platform with two very different economic engines bolted to it — a competitive clinical-lab business (Diagnostics) and a scarce, high-margin data-licensing business (Data). The bull case lives almost entirely in the smaller, faster-compounding Data line.

3. Industry Dynamics

Tempus straddles two industries with opposite structural characteristics.

Precision-oncology diagnostics is a large, growing, but structurally challenging market. Comprehensive genomic profiling (CGP) of cancer patients remains underpenetrated — management argues we are “early-to-middle” innings, with many cancer patients still not receiving CGP, supporting industry unit growth of plausibly low-double-digits for 3–5 years. The tailwinds are real: expanding clinical guidelines, more targeted therapies requiring biomarker testing, and a secular shift toward liquid biopsy and MRD monitoring. But the profit pool is contested. Reimbursement is administered-price-driven (Medicare MolDx/MAC rates, ADLT designations, commercial coverage), capping pricing power; the business is reagent-, instrument-, and labor-intensive; and the competitive set is deep-pocketed: Foundation Medicine (Roche-owned), Guardant Health (liquid-biopsy leader), Caris Life Sciences (IPO’d 2025), Natera (MRD/Signatera leader), Exact Sciences, Neogenomics, plus the reference-lab giants Labcorp and Quest. In hereditary testing, competitors include Myriad Genetics and GeneDx. This is a Marathon “capital-attracting” industry — high growth pulling in capital and competition, which historically compresses returns. Barriers to entry are moderate (clinical validation, payer relationships, sales infrastructure, menu breadth) but not insurmountable, and switching costs for an ordering oncologist are low.

Healthcare data / real-world-evidence licensing is structurally far better. The buyers — large pharma R&D organizations — are price-insensitive relative to drug-development economics, contracts are large and multi-year, and the differentiating asset (a clinically-linked, longitudinal, multimodal, de-identified dataset at scale) is genuinely hard to replicate. The competitive set here — Flatiron Health (Roche), Komodo Health, IQVIA, Veeva (more software than data), Truveta, nference — is real but narrower, and Tempus’s combination of generating its own molecular data through testing and linking it to clinical outcomes is differentiated. This is the part of the business with the most attractive industry economics and the clearest barrier to entry.

Regulation cuts both ways: FDA oversight of diagnostics (LDT/IVD framework, companion-diagnostic rules), CMS/MolDx reimbursement, HIPAA/de-identification and data-privacy law (critical to the Data business’s license to operate), and evolving AI-in-healthcare regulation. A de-identification or privacy challenge to the data business would be a tail risk to the highest-quality revenue line.

Verdict: A structurally mixed picture — Diagnostics sits in a good-growth-but-contested-returns industry typical of medical diagnostics; Data sits in a structurally attractive, high-barrier niche. The blended industry quality is better than a pure diagnostics lab but worse than the Data multiple alone would imply.

4. Competitive Position

The moat question is the whole investment case, so it deserves precision in Greenwald’s taxonomy.

Data and applications — a candidate intangible-asset + data-scale + switching-cost moat (real but unproven durability). The 500+ petabyte clinically-linked dataset is a genuine intangible asset: it is the product of years of testing volume, EHR integration, and data harmonization that a new entrant cannot buy off the shelf. The moat mechanism is economies of scale in data plus customer captivity: the more testing Tempus runs, the deeper and more longitudinal the dataset, the more valuable it is to pharma, and the more pharma builds proprietary models on Tempus’s platform (AstraZeneca’s foundation model; smaller customer-specific models on Tempus GPUs). Once a pharma R&D team has built models against Tempus data and embedded it in compound-selection and trial-design workflows, switching costs rise. The clinching evidence the bulls cite is the migration of customers from one-off projects to $100M+ strategic, multi-year relationships, and management’s claim of renewing “at or above” historical levels. The skeptic’s caveat: “switching costs” in data licensing are asserted, not yet proven across a renewal cycle in a downturn; the largest contract (AstraZeneca) has a minimum commitment that runs only through December 2026 (extendable to 2028 at AZ’s election); and the related-party Pathos/SB Tempus arrangements make a portion of the highest-margin revenue harder to take at arm’s-length face value. If a moat claim cannot be tied to a financial outcome that deteriorates without it, it isn’t a moat — here the outcome (durable $100M+ renewals at rising TCV) is emerging but not yet established through a full cycle.

Diagnostics — weak-to-average differentiation in a crowded market. Tempus is growing oncology units faster than most peers (~28% in Q1-26), and management attributes this to platform breadth (algos, integrated ordering via Hub, RNA capability). That is a real but modest and shared advantage — Foundation Medicine, Guardant, and Caris all offer comprehensive profiling, and the ordering physician faces low switching costs. There is no pricing power (reimbursement is administered), no meaningful scale-cost advantage versus Roche-backed Foundation or the reference labs, and the Ambry-acquired hereditary business competes directly with Myriad and GeneDx. The algos attach rate and the “embedded-in-workflow” stickiness are the best differentiators, but they are early.

ROIC test. Greenwald’s and Marathon’s acid test — does the business earn returns above its cost of capital? — currently fails decisively: ROIC and ROE are deeply negative (return on assets -15% in 2025), because the company is still lossmaking. The moat is therefore prospective: it would be validated only by sustained positive returns that a competitor cannot compete away. Today it is a hypothesis backed by suggestive evidence (margin expansion, $100M+ data renewals, network-effect logic), not a proven economic fact.

Verdict: A credible but unproven moat concentrated in the Data business; a competitive, undifferentiated-to-modestly-differentiated position in Diagnostics. The honest characterization is “emerging network-effect/data-scale advantage, not yet visible in returns on capital.”

5. Growth History and Forward Opportunities

History. Revenue compounded from $257.9M (2021) to $1,271.8M (2025) — a ~49% CAGR — but the path matters: 2022 +24%, 2023 +66%, 2024 +30%, 2025 +83.5%. The 2025 surge is misleading on an organic basis: Diagnostics revenue jumped from $451.7M to $955.4M (+111%), and the dominant driver was the Ambry Genetics acquisition (closed February 2025), not underlying volume. Stripping the acquisition, organic growth in 2025 was far lower — plausibly in the ~25–30% range, consistent with the Q1-26 organic figures (total +36%, of which Diagnostics +35% and Data +40.5%, with oncology units +28%). Investors who extrapolated the headline +83% as organic mis-modeled the business.

Data business growth has been the cleaner, higher-quality story: $168.8M (2023) → $241.6M (2024, +43%) → $316.4M (2025, +31%), with Q1-26 +40.5% and three consecutive quarters of >$100M bookings and rising total-contract-value. This is the line where the durable, high-margin compounding lives.

Forward opportunities.

• Oncology volume + ASP. Continued unit growth (~20%+) plus a guided ~+$500 ASP uplift from xF/xR FDA approvals and ADLT pricing — a rare lever where both price and volume can rise.
• MRD. Deliberately throttled today (~6,500 tests/quarter) pending Personalis-driven reimbursement; management explicitly signals it could be “a very formidable MRD player” if it unleashed its full sales force — a large optionality if MRD reimbursement approaches Natera-Signatera levels.
• Data expansion. Management sees the U.S. data-and-modeling business reaching “multi-billion dollars,” extending from oncology (essentially all of today’s data revenue) into neurology (an Alzheimer’s multimodal-model project underway), cardiovascular, and rare disease — plus international.
• Guided trajectory. ~25% revenue growth for three years (~$1.6B → ~$2.5–3.0B), with FY2026 guidance $1.59–1.6B and adjusted EBITDA ~+$65M, implying meaningful operating leverage.

Verdict: High-quality forward growth in Data; solid but acquisition-flattered, reimbursement-dependent growth in Diagnostics. The blended story is attractive if organic momentum and Data renewals hold; the historical headline growth rate overstates the organic engine and should not be naively extrapolated.

6. Financial Quality

Revenue quality. Two-thirds-plus of revenue (Diagnostics) is transactional and reimbursement-exposed; the rest (Data) is contractual but partly related-party and partly concentrated in a few large pharma names. Gross margin has expanded impressively — 32.4% (2021) → 40.6% → 53.8% → 55.0% → 62.7% (2025) — driven by mix shift toward Data, rising ASPs, and lab-efficiency scale. That trajectory is the single best evidence that economics improve with scale.

Profitability. Still negative on a GAAP basis. FY2025: operating loss -$252.9M (op margin -19.9%), net loss -$245.0M, diluted EPS -$1.41, GAAP EBITDA -$150.5M. The 2024 figures look far worse (op loss -$691M, EBITDA -$654M) but were distorted by $534M of IPO-triggered stock-based compensation; normalizing for that, the underlying loss has been narrowing. Management’s preferred metric, adjusted EBITDA, was roughly breakeven in FY2025 and is guided to ~+$65M in FY2026, with Q1-26 at -$3M (a ~$13M YoY improvement). The inflection is real but is a non-GAAP inflection that adds back ~$125M of annual SBC.

Cash flow — the crux. This is where the bear case bites. FY2025 operating cash flow was -$218M and free cash flow -$239M; capex is light (~$21M), so the burn is operating, not capital. The company funded itself with a $750M 0.75% convertible note (due 2030), an Ares term loan, and an active at-the-market equity program, plus $377M of cash spent on acquisitions (Ambry, Deep 6 AI, Paige.AI). Management argues cash flow inflects sharply positive from Q2-26 as prepaid Data contracts convert to quarterly billings and adjusted EBITDA scales, and asserts it “does not need more cash.” Net income is diverging favorably from a still-negative cash-from-operations — the gap is SBC and working-capital timing, not a quality red flag per se, but the absolute cash burn is the fact that matters: until OCF turns durably positive, the equity story depends on the convert/ATM treadmill.

Balance sheet. Q1-26: cash + short-term investments ~$639M (plus ~$82M long-term investments) against ~$1.32B of debt (the convert + Ares term loan), for net debt of ~$712M. Stockholders’ equity is ~$416M but is negative on a tangible basis — goodwill ($470M) plus intangibles ($340M) from the acquisition spree exceed book equity. The current ratio (~3.3x) is healthy and the convert is cheap (0.75%) and not due until 2030, so there is no near-term liquidity cliff — but there is also limited cushion if growth or Data renewals disappoint and capital markets close.

ROIC/ROE. Not meaningful (negative). Return on assets -15.3% (2025). These will only become useful metrics if/when the company reaches sustained profitability.

Verdict: Economics are clearly improving with scale (gross margin +30 points in four years, adjusted-EBITDA inflection), but the business is not yet self-funding. Quality is “improving-but-unproven,” with the cash burn and goodwill-heavy, negative-tangible-equity balance sheet as the principal demerits.

7. Capital Allocation

M&A — an aggressive, externally-funded roll-up. 2025 was a major acquisition year: Ambry Genetics (~$600M, hereditary testing — the largest deal and the engine of the 2025 Diagnostics surge), Deep 6 AI (clinical-trial matching), and Paige.AI (digital/AI pathology). These created ~$470M of goodwill and ~$340M of intangibles. The strategic logic is coherent (Ambry adds hereditary volume and data; Deep 6 strengthens Trials; Paige adds pathology AI), but the cash was funded by debt and equity issuance, not internal generation, and integration/return on these deals is unproven. This is textbook Marathon “capital-cycle” behavior — deploying external capital aggressively into a high-growth space — which the framework warns often destroys value when returns are competed away.

Dilution and issuance. Share count rose from 161.9M (2024) to 178.3M (2025) and continues via ~$125M/year of SBC and an active ATM program. SBC at ~10% of revenue is high (typical of venture-stage tech) and a real economic cost the adjusted-EBITDA metric papers over.

No buybacks or dividends (appropriate for a cash-burning growth company), and no near-term prospect of return of capital.

Related-party density — the governance flag. Several arrangements warrant scrutiny:

• Pathos (a Lefkofsky-affiliated company): under the AstraZeneca/Pathos foundation-model agreements, AstraZeneca pays Tempus $35M, Tempus pays Pathos $35M, and Pathos may settle up to 50% of the data-license fees it owes Tempus in its own Series D preferred stock. Tempus carries a ~$25M related-party asset for future Pathos services. This circularity makes a slice of the marquee AstraZeneca relationship harder to evaluate at arm’s length.
• SB Tempus Corp — the SoftBank Japan joint venture — with IP-license and data-license agreements that are related-party transactions.
• Recursion Pharmaceuticals — a data/services + equity arrangement.

None is necessarily improper, but the concentration of related-party deals around the founder and the highest-margin revenue line is exactly what skeptics (including 2024–25 short reports) have flagged, and it lowers the trust an outside investor can place in headline Data economics.

Incentive alignment / governance. Tempus is a controlled company: Lefkofsky’s 5.0M Class B shares carry 30 votes each, giving him ~58.2% of voting power despite a far smaller economic stake, and the company has elected the controlled-company exemption from certain governance rules. Lefkofsky’s track record is that of a serial monetizer (Groupon, Echo Global Logistics, InnerWorkings, Mediaocean) — a brilliant company-builder, but one whose history is about creating and selling equity, not compounding it for outside minority holders. Insider Form 4 activity (182 filings since the 2024 IPO) is dominated by RSU vesting and 10b5-1-planned sales; no discretionary open-market insider purchases were identified — there is no “insider bought the dip” signal here.

Verdict: Below-average capital-allocation and governance profile. The M&A is strategically logical but externally funded and unproven; the founder controls the company through supervoting and has a monetization (not minority-compounding) track record; and the related-party web around the Data business is a genuine, non-trivial trust discount.

8. Changes and Headwinds — Last Two Years

Strategic / corporate.

• IPO (June 2024) at $37/share.
• Ambry Genetics acquired (Feb-2025) — transformational for the hereditary business and the dominant driver of 2025 reported growth.
• Deep 6 AI (Mar-2025) and Paige.AI (Aug-2025) acquired — trial-matching and pathology-AI tuck-ins.
• AstraZeneca/Pathos foundation-model collaboration (Apr-2025) — large, strategic, and related-party-entangled.
• Merck (new $100M±tier strategic Data deal) and expanded Gilead relationship signed in Q1-26.
• $750M convertible notes (0.75%, due 2030) issued in 2025 to fund operations and M&A.

Operating.

• Q1-26 beat and raised FY2026 guidance to $1.59–1.6B revenue / ~+$65M adjusted EBITDA.
• Adjusted-EBITDA inflection underway (multiple consecutive quarters of YoY improvement).
• FDA progress: xT approved and being expanded (tumor-only amendment to accelerate ADLT migration); xF liquid-biopsy submission pending; RNA (xR) to follow — collectively the basis for the guided ~+$500 ASP uplift.
• Investor Day scheduled shortly after the May-2026 call.

Headwinds / overhangs.

• Cash burn and dilution — the central financial overhang until OCF turns durably positive.
• AstraZeneca renewal — the MSA minimum commitment ($220M, extendable to $320M) runs only through December 2026; a non-renewal or step-down would hit the highest-margin line and sentiment.
• Short-seller scrutiny — 2024–25 reports (e.g., Spruce Point) questioning data-revenue quality, Ambry accounting, and related-party disclosures kept a bid under the bear case and contributed to the -60% drawdown.
• Reimbursement risk — MolDx/MAC and ADLT pricing decisions can move ASPs either way; MRD economics remain unfavorable until reimbursement broadens.
• Stock-price round-trip — the -47%-off-peak tape is itself an overhang on employee retention (SBC strike prices) and capital-raising flexibility.

Verdict: The last two years strengthened the operating story (Data wins, guidance raises, margin inflection) while adding balance-sheet and governance complexity (debt, goodwill, related-party deals). On net, the thesis is more proven on growth and more exposed on quality/financing than at IPO.

9. Risk Analysis

Verdict: The dominant risks are financial (cash burn/dilution) and quality/governance (concentration + related parties), not operational. A catastrophic loss is unlikely near-term given liquidity and a cheap, long-dated convert, but a 40–60% drawdown on a growth scare or a Data-renewal miss is entirely plausible — and has already happened once.

10. Valuation Discussion (Embedded Expectations)

Tempus has no earnings and negative tangible book, so conventional P/E and P/B are uninformative (book-multiple metrics are noise here). The relevant lenses are EV/revenue, EV/gross profit, and a forward path to FCF.

Where it trades. At ~$54.87, market cap is ~$9.8B and enterprise value ~$10.5B (adding ~$0.7B net debt). On FY2025 revenue of $1.27B, that is ~8.3x trailing EV/sales; on FY2026E revenue of ~$1.6B, ~6.6x forward EV/sales. Gross profit of ~$800M (2025) implies ~13x EV/gross-profit. the price-to-sales multiple sits near the low end of its (short, ~2-year) post-IPO trading range — i.e., cheap relative to itself, having de-rated from a peak EV/sales of ~15x at the October-2025 high to ~6.6x today. Against peers, ~6.6x forward sales is a premium to most diagnostics labs (Guardant, Exact, Natera typically trade single-digit EV/sales; Foundation is inside Roche) but a discount to high-growth software/data names — appropriate only if the market weights the Data business heavily.

A sum-of-the-parts frames the debate better than a blended multiple:

• Data and applications (~$316M 2025 revenue, growing ~30–40%, software-like margins): if valued at ~10–15x revenue (a defensible range for scarce, high-margin, high-growth healthcare data), that is ~$3.2–4.7B of enterprise value on its own.
• Diagnostics (~$955M 2025 revenue, reimbursement-dependent, lower-margin, competitive): at ~3–5x revenue (in line with growing diagnostics labs), that is ~$2.9–4.8B.
• Summed: ~$6.1–9.5B, versus the ~$10.5B EV. The current price therefore embeds either an above-range multiple on one or both pieces, or confidence that organic growth and margins will substantially exceed today’s run-rate.

Scenario analysis (illustrative, not a target):

• Bear (~$30–42). Organic Diagnostics growth decelerates toward the mid-teens, an AstraZeneca step-down or a Data-revenue/related-party disclosure problem compresses the Data premium, cash burn forces dilutive raises, and the multiple re-rates to ~4x forward sales. This is roughly the spring-2026 low.
• Base (~$45–65). ~25% revenue growth holds for 2–3 years, adjusted EBITDA scales to the guided ~$65M and beyond, OCF inflects positive in 2026–27, AstraZeneca renews, and the stock holds ~5–7x forward sales. The current price sits inside this band.
• Bull (~$80–110+). The Data business compounds 35–40% toward “multi-billion,” MRD reimbursement unlocks a large new volume engine, ASPs rise ~$500, GAAP profitability arrives, and the market re-rates toward a data-company multiple. This is roughly the October-2025 high.

Embedded-expectations summary. At ~6.6x forward sales on a GAAP-lossmaking company, the market is underwriting: durable ~25% growth, continued gross-margin expansion, a sustained adjusted-EBITDA-to-FCF inflection, an intact AstraZeneca relationship, and clean conversion of Data bookings to cash — with essentially no margin of safety if any leg slips. The market is plausibly correct on the Data business’s scarcity and too sanguine on the organic durability of Diagnostics and on the related-party/financing risks. No price target; no recommendation.

11. Variant Perception

Consensus belief. Sell-side is broadly constructive: Tempus is a fast-growing, category-defining AI-healthcare platform with a unique data asset, an inflecting margin profile, and a long runway; the recent beat-and-raise and marquee Data wins (Merck, Gilead) validate the model. The de-rating from $104 is seen as a multiple reset, not a thesis break.

Strongest bull case. The Data business is a structurally scarce, high-margin, network-effect asset that the world’s largest drug makers are paying $100M+ to access and build models on — and it is still early (oncology only; neurology/cardio/rare-disease and international ahead). Bundle that with a Diagnostics business growing units ~28% with ~$500 of ASP upside and an MRD optionality that management is deliberately throttling, and you have a company that can compound revenue ~25%+ for years while inflecting to positive cash flow. At ~6.6x forward sales — half its peak multiple — you are buying a future data-and-AI healthcare leader at a reset price.

Strongest bear case. This is a cash-burning, founder-controlled roll-up whose headline 2025 growth was bought (Ambry), not earned; whose highest-margin revenue is concentrated in a few large pharma contracts (AstraZeneca’s runs out in 2026) and entangled in related-party arrangements (Pathos, SB Tempus) that reduce its credibility; which funds itself with converts and ATM equity while diluting holders ~$125M/year in SBC; and which trades at a premium-to-diagnostics multiple with negative tangible equity and no GAAP profits. The factor tape (beta 2.7, SmallSize loading >2, -59% drawdown) confirms this is speculative growth, not a compounder — and speculative growth re-rates violently when sentiment turns.

The 3–5 assumptions that matter most:

1. Organic growth durability — can Diagnostics grow units ~20%+ and Data ~30%+ organically (ex-acquisitions) for multiple years?
2. Data-contract durability — does AstraZeneca renew/expand on arm’s-length terms, and do the $100M+ relationships prove sticky through a renewal cycle?
3. Cash-flow inflection — does OCF turn durably positive in 2026–27 without further dilutive financing?
4. Related-party integrity — are Pathos/SB Tempus economics genuinely arm’s-length, or do they overstate Data quality/margins?
5. Margin trajectory — does gross margin hold/expand and adjusted EBITDA scale toward and beyond the ~$65M guide?

Factor-positioning read (from the tape). Tempus is not a value or momentum name in any factor sense — its loadings are dominated by Market (~2.4) and SmallSize (~2.25) with negative LowVolatility, i.e., a high-beta, high-vol, small-cap speculative growth profile. It round-tripped 4.5x then -60%, and is bouncing ~31% off the March low. The tape says: abandoned speculative growth finding a floor, with consensus offsides in both directions at the extremes — too euphoric at $104, arguably too despairing at $42.

Where consensus may be wrong. The most underappreciated risk is the gap between reported and organic growth (Ambry) combined with related-party opacity in the Data line — together they mean the market may be paying a clean-data-compounder multiple for a business whose cleanest numbers deserve a haircut. The most underappreciated upside is MRD, a genuinely large, deliberately-throttled option the market gives little credit for today.

12. Fact vs. Interpretation

13. Open Questions

1. What is the true organic growth rate? Ex-Ambry/Deep6/Paige, what did Diagnostics and total revenue grow in 2025 and 1H-26? (Q1-26 figures suggest ~mid-30s total, but a precise organic bridge is not cleanly disclosed.)
2. Will AstraZeneca renew/expand the MSA past December 2026, and on what terms? This is the single most important Data-business datapoint.
3. How arm’s-length are the Pathos and SB Tempus economics, and how much Data revenue/margin do they represent?
4. When does operating cash flow turn durably positive, and is the FY2026 guide achievable without further ATM issuance?
5. What is the real economic return on the Ambry/Deep6/Paige acquisitions — are they accretive to organic growth and margins, or goodwill parked on the balance sheet?
6. How large and how soon is the MRD opportunity once Personalis reimbursement broadens — is it a 2027 or a 2029 story?
7. What is Lefkofsky’s selling cadence and does the dual-class structure sunset?
8. Do the short-seller concerns (data-revenue recognition, Ambry accounting) have any substance that surfaces in future filings?

14. What Must Be True

Bull case — what must be true (and its falsification test).

• The Data business compounds ~30–40% organically toward a multi-billion-dollar run-rate, with $100M+ pharma relationships renewing at or above prior levels and expanding into new disease areas. Falsification: a flat or declining Data-revenue quarter, or an AstraZeneca/large-pharma non-renewal or step-down, in 2026–27.
• Organic total growth holds ~25% and cash flow inflects durably positive in 2026–27 without dilutive raises. Falsification: organic growth decelerates below ~15%, or OCF stays negative into 2027 forcing a meaningful equity raise.
• Margins continue to expand and adjusted EBITDA scales past the ~$65M guide toward GAAP profitability. Falsification: gross margin stalls below ~60% or adjusted EBITDA misses the FY2026 guide.

Bear case — what must be true (and its falsification test).

• Reported growth is acquisition-flattered and organic momentum fades, leaving a competitive diagnostics business growing mid-teens dressed as a hyper-growth platform. Falsification: a clean, disclosed organic bridge showing sustained ~25%+ organic growth.
• The Data premium is unjustified — concentration in a few pharma contracts plus related-party entanglement means the high-margin line is less durable/clean than it appears. Falsification: multi-year renewals at rising TCV with arm’s-length, related-party-free economics, confirmed in filings.
• The cash burn forces dilution into a weak tape, and the multiple re-rates toward diagnostics-peer levels (~4x sales). Falsification: positive OCF and self-funding by 2027 with no ATM usage.

15. Source Appendix

See Appendix B below for the full, dated, primary-source citation list.

No buy/sell recommendation and no price target appears in the analysis sections above. The single exception is the “Claude’s Take” block, which is the author’s own independent opinion and general information only — not investment advice.

APPENDIX A — Standard Diligence Questionnaire

Tempus AI, Inc. (NASDAQ: TEM) — as of 2026-06-26

Supplemental to the analysis. Fact / Interpretation / Assumption labels applied where material.

General

What thoughtful questions have other investors asked? The dominant questions: (1) how much of 2025’s +83% growth was organic versus the Ambry acquisition (Interpretation: most of the Diagnostics +111% was Ambry); (2) durability and renewal of the large pharma Data contracts (AstraZeneca’s minimum runs through Dec-2026); (3) the related-party arrangements (Pathos, SB Tempus) and whether Data economics are arm’s-length; (4) when cash flow turns positive without dilution; (5) the credibility of short-seller (e.g., Spruce Point) claims on data-revenue recognition and Ambry accounting.

Cyclicality & Earnings Nature

Earnings at a cyclical high or low? Neither — the company is pre-profitability (GAAP net loss -$245M FY2025). Diagnostics volumes are secularly growing, not cyclical; Data is contract-driven and lumpy quarter-to-quarter (back-half weighted). Internal or external drivers? Predominantly internal (volume scaling, new Data deals, ASP/FDA progress) rather than macro. Revenue stability? Diagnostics is recurring-transactional and reimbursement-exposed; Data is multi-year contractual but concentrated. Market size/outlook? Large and growing — precision-oncology CGP is underpenetrated (“early-to-middle innings”); the U.S. data-and-modeling TAM is, per management, “multi-billion dollars,” extending beyond oncology into neurology, cardio, and rare disease; international ahead. (Interpretation: TAM is real but management-sized.)

Business Quality & Competitive Moat

Industry getting more/less competitive? More — diagnostics is a capital-attracting, crowded field (Foundation Medicine/Roche, Guardant, Caris, Natera, Myriad, GeneDx, Labcorp, Quest). Data licensing is a narrower, higher-barrier niche (Flatiron/Roche, Komodo, IQVIA, Truveta). How profitable (ROIC/ROE)? Negative today (ROA -15%, no positive ROIC); a moat must still prove itself in returns. Industry profitability / barriers? Diagnostics: moderate barriers (clinical validation, payer relationships, menu, sales force), low physician switching costs, administered pricing → contested returns. Data: high barriers (scarce clinically-linked multimodal dataset), emerging switching costs. Easily understood? Moderately — two distinct engines on one platform; the related-party web adds opacity. Undermined by low-cost foreign labor? No — regulated, U.S.-centric clinical lab + proprietary data. Do brands matter? Less brand than data-scale, FDA approvals, payer coverage, and physician workflow embedding. Nature of competition? Menu breadth, turnaround time, algos/AI differentiation, reimbursement, and (in Data) dataset depth. Switching costs? Low for ordering physicians; potentially high (but unproven) for pharma that build models on Tempus’s platform.

Financial Condition & Balance Sheet

Assets not on the balance sheet? The 500+ petabyte dataset is internally generated and not capitalized at fair value — the core intangible is largely off-balance-sheet (Interpretation: a hidden asset if the moat is real). Off-balance-sheet liabilities? Operating leases (capitalized); commitments under Personalis/Pathos/SB Tempus arrangements; ATM dilution overhang. How conservative is the accounting? Average-to-aggressive — heavy non-GAAP emphasis (adjusted EBITDA adds back ~$125M SBC), acquisition purchase-accounting (goodwill/intangibles), and related-party revenue all warrant scrutiny; short-sellers have alleged data-revenue and Ambry-accounting concerns (unproven). CapEx-hungry? No — capex ~$21M (light, asset-light lab + cloud); the cash drain is operating, not capital.

Capital Allocation & Management

FCF generation / use / philosophy? Negative FCF (-$239M FY2025); capital is consumed, not allocated — funded by a $750M 0.75% convert (due 2030), an Ares term loan, and an ATM. Philosophy: aggressive growth investment + M&A, externally financed. Significant acquisitions? Yes — Ambry (~$600M, Feb-2025), Deep 6 AI (Mar-2025), Paige.AI (Aug-2025); ~$470M goodwill created. Buying back shares? No. Issuing shares to insiders? Yes — ~$125M/yr SBC; share count 161.9M→178.3M. Director/management compensation? Equity-heavy; controlled-company governance. Management motivations? (Interpretation) Founder Lefkofsky controls 58.2% of votes via 30:1 Class B and has a serial-monetization history (Groupon, Echo, InnerWorkings, Mediaocean) — a company-builder/seller, not a demonstrated minority-compounder; related-party deals concentrate around him.

Valuation & Market Data

ADR / MLP / K-1? No — U.S. C-corp, Class A common (NASDAQ: TEM); no K-1. Dividend policy? None (appropriate for a cash-burning growth company). How profitable? Not yet (GAAP loss); gross margin 62.7% and rising; adjusted EBITDA inflecting to ~+$65M guided FY2026. Net income vs. cash from operations diverging? Yes — net loss -$245M vs. OCF -$218M; the gap is SBC + working-capital timing. Watch for OCF turning positive (the key tell). Valuation: ~6.6x forward EV/sales, ~8.3x trailing; price-to-sales near the low end of its short post-IPO history (cheap vs. itself, premium vs. diagnostics peers).

Risks & Downside

What would cause the stock to decline? A growth/organic-deceleration scare, an AstraZeneca/large-pharma non-renewal, a related-party/data-revenue disclosure problem, a dilutive capital raise, a reimbursement cut, or a broad unprofitable-growth de-rating (it already fell -60% once). Catastrophic-loss risk? Low near-term — ~$720M liquidity and a cheap, long-dated convert; not a going concern. Total-loss risk? Low but non-zero on a multi-year horizon if the moat fails to translate to profits and capital markets close (Assumption).

Recent News & Events

Has the environment changed recently? Positively at the margin — Q1-26 beat + raised FY2026 guidance ($1.59–1.6B / ~+$65M adj EBITDA); new $100M±tier Merck and expanded Gilead Data deals; software-validation publication (Jun-2026); Angiosarcoma Awareness research collaboration (Jun-2026). Significant acquisitions / accounting changes? Three 2025 acquisitions (above); purchase accounting added substantial goodwill/intangibles; $750M convert issued. Other recent changes? Continued FDA progress (xT expansion, xF submission, planned xR), Investor Day held shortly after the May-2026 call.

APPENDIX B — Source Appendix

Tempus AI, Inc. (NASDAQ: TEM) — as of 2026-06-26

Primary sources first. All figures reconciled to SEC filings where possible; third-party aggregated market data is labeled as such and used for cross-check/positioning, not as primary authority.

Primary — SEC filings (EDGAR, CIK 0001717115)

1. Form 10-K, FY2025 (filed 2026-02-24), tem-20251231.htm — business description, two-segment revenue (Diagnostics $955.4M / Data $316.4M / total $1,271.8M), gross margin, operating/net loss, SBC, debt ($750M 0.75% Convertible Senior Notes due 2030; Ares term loan), acquisitions (Ambry 2025-02-03, Deep 6 AI 2025-03-11, Paige.AI 2025-08-22), related-party arrangements (Pathos, SB Tempus, Recursion), AstraZeneca MSA ($220M min, extendable to $320M, through Dec-2026), GSK collaboration, dual-class structure (173.7M Class A / 5.0M Class B; Lefkofsky 58.2% voting; controlled company), risk factors. https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0001717115
2. Form 10-K, FY2024 (filed 2025-02-24), tem-20241231.htm — FY2024 financials ($693.4M revenue, IPO-year SBC $534M).
3. Form 10-Q, Q1-2026 (filed 2026-05-05), tem-20260331.htm — Q1-26 balance sheet (cash+ST inv ~$639M; debt ~$1.32B; net debt ~$712M; equity $416M; goodwill $470M + intangibles $340M), Q1 results.
4. Form 10-Q corpus (2024-08-06 through 2026-05-05) — quarterly trend data.
5. 8-K, 2026-05-05 — Q1-2026 results and raised FY2026 guidance ($1.59–1.6B revenue; ~+$65M adjusted EBITDA).
6. DEF 14A, 2026-04-07 — proxy; executive compensation, governance, controlled-company disclosures, annual meeting May-21-2026.
7. Form S-1 / S-1/A (2024) — IPO prospectus; use of proceeds, pre-IPO capital structure, AstraZeneca/GSK agreement terms.
8. Form 4 corpus (182 filings since IPO) — insider transactions; dominated by RSU vesting / 10b5-1 sales; no discretionary open-market purchases identified.

Primary — Company materials

9. Q1-2026 earnings call transcript (2026-05-05). Management commentary: revenue $348.1M (+36%); Diagnostics $261.1M (+35%, oncology units +28%); Data $87M (+40.5%); Merck (new $100M+ strategic deal) and expanded Gilead; ASP ~$1,720–1,740 with ~+$500 guided uplift; MRD ~6,500 tests (+500%) deliberately throttled (Personalis reimbursement); FY2026 guide raised; ~25% growth guided over 3 years; FDA xT amendment / xF submission. (Management commentary treated as hypothesis, validated against filings.)
10. Tempus IR site / investor presentation — segment KPIs, TCV/bookings, Investor Day materials (referenced in transcript).

Third-party — quantitative (cross-check, not primary)

11. Public financial-data aggregators — multi-period income statement, balance sheet, cash flow, enterprise value, and valuation multiples (FY2021–2025 + Q1-26). EV ~$8.8–10.5B; EV/TTM sales ~6.4–8.5x. Reconciled to SEC filings.
12. Public market-price history — IPO 2024-06-14 ($40.25 close); all-time low $22.89 (2024-06-25); all-time high $104.32 (2025-10-09); 52-week low $41.73 (2026-03-30); last $54.87 (2026-06-25). Price-to-sales near the low end of its ~2-year post-IPO range; P/E n/m (losses).
13. Public factor/risk-model data — beta ~2.7; loadings dominated by Market and SmallSize, with negative low-volatility loading and no value/quality/momentum loading; trailing 1-year return ~-16% with a ~-59% max drawdown.

Industry / peer context

14. Competitor public materials — Foundation Medicine (Roche), Guardant Health (GH), Caris Life Sciences, Natera (NTRA), Exact Sciences (EXAS), Myriad Genetics (MYGN), GeneDx, Flatiron/Komodo/IQVIA (data) — competitive-set framing.

Notes on conflicts / caveats

• Third-party aggregated market data are estimates; SEC filings are primary and authoritative where they differ.
• Post-IPO valuation percentiles use only ~2 years of history and are own-history (not cross-sectional) context.
• 2024–25 short-seller reports (e.g., Spruce Point) referenced as bear-case framing only; their specific allegations are not independently verified here and are flagged as Open Questions.