RESEARCH MEMO · AI Biotech
Tempus × Recursion: Data Moat Becomes Contracted — Not Yet a Cash Moat
AI Biotech Research Memo | Contract visibility improves; pricing power and cash conversion remain unproven | 24 September 2026
Companies referenced: Tempus AI, Inc. (Nasdaq: TEM) | Recursion Pharmaceuticals, Inc. (Nasdaq: RXRX)
The amendment moves Tempus from a high but terminable ceiling to a lower, locked-in commitment. Yet the data scope is smaller, most consideration may be paid in RXRX shares, and Tempus separately owes Recursion $12 million for TxFM. Data utility and relationship durability are better evidenced; pricing power, net cash benefit and free-cash-flow conversion are not.
Three committed $14 million anniversary payments
$4 million minimum cash each year; the balance may be cash or RXRX shares
Longer relationship with no convenience-termination right
Tempus' non-refundable fee for a worldwide, non-exclusive, oncology-only TxFM license
Contracted monetization partly validated | Pricing power, net cash and FCF remain unproven
01 · RESEARCH BRIEF
The one-minute brief
The parties signed the amendment on 15 September and disclosed it on 21 September. They extended the relationship through November 2029 and Recursion surrendered its convenience-termination right. Up to $84 million of partly discretionary fees over the next two years became $42 million of committed payments—$14 million at each of three future anniversaries. Only $4 million per year must be cash; Recursion may settle the balance in cash or shares, and the accessible de-identified record count falls. In a reciprocal agreement, Tempus will pay a non-refundable $12 million fee and provide linked pathology and clinical data for a two-year, oncology-only, worldwide non-exclusive TxFM license. The renewal and locked-in term strengthen evidence of utility, not proof of unit pricing, revenue recognition, cash collection or free cash flow. The posture remains an evidence-validation watchlist.
Audio transcript
The Tempus–Recursion amendment is easy to misread as forty-two million dollars of locked-in cash revenue for Tempus. The better interpretation is that up to eighty-four million dollars of partly discretionary fees over two years became forty-two million dollars of committed payments over three years. Recursion surrendered its convenience-termination right, which is meaningful evidence that it values the data relationship. But only four million dollars per year must be cash; the balance may be cash or RXRX shares, and the accessible record count is lower. Tempus also owes Recursion twelve million dollars for a two-year, oncology-only, worldwide non-exclusive TxFM license and will provide linked pathology data. This is a bilateral asset exchange, not one-way rent extraction. Our updated view is that contracted monetization is partly validated, while pricing power, net cash benefit and free-cash-flow conversion remain unproven. The next tests are the filed agreements, actual settlement, revenue recognition and a second independent customer.
Known facts and open questions
- Confirmed
- The relationship runs through November 2029; Recursion gave up convenience termination; three $14 million anniversary payments are contractually committed.
- Must be separated
- $42 million of committed consideration is not current-period recognized revenue, cash received or free cash flow; aggregate minimum cash is only $12 million.
- Still unproven
- Comparable unit pricing, actual cash/share settlement, recognition timing, net cash benefit and a second comparable customer.
- Research stance
- Evidence-validation watchlist: wait for the filed agreements, cash conversion, repeatability and FCF proof.
Data utility
A named customer renews and exchanges model and data assets
Contract commitment
The exit right disappears and three payment dates become firm
Pricing and repeatability
Smaller scope, reciprocal value and one public example leave both unknown
Cash loop
Settlement, recognition, operating cash flow and FCF do not yet connect
| Term | Original structure | Amended structure | ACIS read |
|---|---|---|---|
| Term and exit | Five years; convenience termination after year three | Six years through November 2029; convenience termination removed | Relationship durability improves |
| Future fees | Up to $42 million at each of two future anniversaries; up to $84 million, partly discretionary | $14 million at each of three future anniversaries; $42 million committed | Potential ceiling and annual scale fall as certainty rises |
| Payment quality | Relevant fees could be settled in shares at Recursion's election | At least $4 million cash per year; balance in cash or shares | Only $12 million aggregate minimum cash |
| Data scope | Original agreement scope | Lower total accessible unique-record count | Unit pricing cannot be inferred without comparable counts |
02 · THESIS → EVIDENCE → UPDATE
What changed in the thesis?
Data utility
- Prior thesis
- Tempus has broad clinical and molecular data, but scale or a buildout plan alone cannot prove that customers will keep paying.
- New evidence
- Recursion renewed, extended the term, surrendered convenience termination and continued exchanging data and model assets.
- Updated view
- Evidence that one named customer finds the data useful improves; that is not proof of an irreplaceable data moat.
Contract economics
- Prior thesis
- The original structure offered a higher potential ceiling, but future fees could disappear if Recursion exited.
- New evidence
- Up to $84 million of partly discretionary fees over two years became $42 million committed over three years, while record scope fell.
- Updated view
- Total-consideration visibility improves, but face value, annual scale and scope decline; higher pricing power cannot be claimed.
Cash conversion
- Prior thesis
- Contract value, recognized revenue, cash received, operating cash flow and free cash flow require separate tests.
- New evidence
- Only $4 million per year must be cash, the rest may be shares, and Tempus separately owes a $12 million cash license fee.
- Updated view
- The flows cannot be mechanically netted, but the disclosed floors do not establish better net cash generation; cash and FCF remain unproven.
Repeatability
- Prior thesis
- One bilateral relationship cannot show that more customers will renew at comparable scope and pricing.
- New evidence
- This structure combines cross-licensed data, models, pathology and clinical information, making its economics unusually specific.
- Updated view
- A second independent multi-year customer without a cross-license or equity settlement is needed before the thesis becomes repeatable data pricing.
03 · EVIDENCE & ANALYSIS
Evidence and analysis
01｜A bilateral asset exchange, not one-way rent extraction
Tempus receives Recursion's $42 million commitment, but it also pays a non-refundable $12 million fee for a two-year, oncology-only, worldwide non-exclusive TxFM license and provides linked de-identified pathology and clinical data. The minimum cash inflow happens to equal the license fee's face value, but payment timing, performance obligations and accounting differ, so they cannot simply be netted. The prudent reading is a restructuring of reciprocal data and model assets—not $42 million of locked-in cash revenue for Tempus.
02｜Tempus already monetizes data, but the cash loop is incomplete
Tempus reported second-quarter 2026 revenue of $382.5 million, up 22%, and Data & Applications revenue of $93.2 million, up 28%, with Insights up 36%. Segment GAAP gross margin was 70.2%, down from 72.7%. Multi-year RPO was $395.3 million, with about 51% expected to be recognized in the next 12 months. Q2 adjusted EBITDA was $8.0 million, yet first-half operating cash flow was negative $80.8 million, and GAAP net income included a $98.5 million unrealized gain. Data monetization is real, but adjusted profit or RPO does not prove FCF.
03｜Recursion gains a model-license example while remaining cash consumptive
Recursion reported Q2 revenue of $7.7 million and operating cash flow of negative $105.9 million for the quarter and negative $187.0 million for the first half. Cash and restricted cash were $556.8 million. Purchases of Tempus records fell to $3.1 million from $22.7 million a year earlier. TxFM gains an external-license example, but it is one two-year, non-exclusive, oncology-only deal supported by reciprocal data value; it does not replace validation from Recursion's own pipeline and partner milestones.
04｜Three scenarios for the next four to six quarters
Upside: settlement arrives in cash or highly liquid value, Tempus Data & Applications grows at least 25% with roughly 70%+ gross margin, a second independent multi-year customer appears and Recursion milestones stay on schedule—connecting the data moat to repeatable contracts and cash conversion. Base: settlement includes shares, purchase cadence varies, segment growth stays in the low-to-mid-20s with 68%–72% margin, and operating cash flow improves unevenly—utility strengthens, FCF does not. Downside: settlement is equity-heavy, record purchases stay low, no comparable customer appears, segment growth falls below 20% with margin below 68%, and Recursion milestones slip—the amendment is better read as a lower amount over a longer term, not pricing-power proof.
04 · INVESTMENT IMPLICATIONS
Industry and asset implications
For Tempus
Total-consideration visibility and relationship durability improve, but the $14 million average annual face value is only about 0.9% of the midpoint of 2026 revenue guidance. It is a quality signal, not a stand-alone earnings inflection.
For Recursion
It preserves longer data access at lower future fees and externalizes TxFM for the first time; model licensing still cannot substitute for proprietary clinical progress, partner milestones or runway.
For AI biotech
Data assets are moving from scale narratives toward contractual pricing. Renewals, comparable unit economics, cash collection and clinical output—not headline contract value—will determine which platforms earn a premium.
For research decisions
Update the view to “total contracted consideration is more visible; cash realization is not.” Keep watching without treating committed payments as revenue, cash or FCF, or extrapolating one customer into sector-wide pricing power.
The decision hinge is not whether a contract exists. It is payment quality, revenue and margin delivery, and whether a second independent customer repeats a multi-year commitment.
05 · VALIDATION & RISKS
What to verify next
Next 30 days | Term transparency
The next 10-Q files the full amendment and TxFM agreement, clarifying accessible records, settlement form, performance obligations and recognition policy.
Failure signal: The exhibits remain absent, or the full terms reveal weaker scope, exit protection or payment quality than the 8-K summary.
Next 90 days | Cash and use
Tempus discloses cash/share settlement, healthy RPO conversion and receivable/deferred-revenue quality as operating cash flow improves; Recursion record purchases and data use recover.
Failure signal: Settlement is equity-heavy, purchases stay low, and Tempus operating cash flow fails to improve in the seasonally stronger second half.
Next 2–4 quarters | Operating repeatability
Data & Applications sustains growth and roughly 70% margin, cash collection exceeds the floor, and a second independent multi-year customer appears.
Failure signal: Growth stays below 20% for two quarters with margin below 68%, while new contracts depend on discounts, equity or cross-licenses.
Next 6–18 months | Model and pipeline
TxFM gains a second customer or measurable deployment output, and Recursion's partner and clinical milestones remain on schedule.
Failure signal: The license is one-off, deployment produces no measurable output, or key pipeline and partner milestones slip.
What would change our view?
Key risks are settlement quality—most consideration may be paid in RXRX shares, exposing Tempus to price and liquidity risk; mismatches among payment, data delivery, access periods and revenue recognition; a lower record count that prevents unit-pricing comparison; customer concentration and a one-off bilateral structure; a non-exclusive, oncology-only, two-year TxFM license; Tempus' integration and cash-conversion burden; and Recursion's high spending, low revenue and potential financing need. We would weaken the “partly validated contracted monetization” view if the filed terms are worse, settlement is equity-heavy, record use keeps falling, Data & Applications growth and margin deteriorate together, or no second comparable customer appears by mid-2027.
06 · FAQ
Key questions
Does this prove that Tempus has an irreplaceable data moat?
No. It shows that one named customer will lock in a relationship and committed consideration, supporting utility. Smaller scope, a worldwide non-exclusive model license and reciprocal data exchange leave irreplaceability to be tested through more customers, unit economics and operating output.
Why is $42 million of committed payments not cash-flow visibility?
Only $12 million must be cash over three years; the rest may be RXRX shares. Tempus separately owes a $12 million TxFM fee. Commitment, revenue recognition, cash settlement and FCF may also occur at different times.
Why can evidence improve when the amount falls from up to $84 million to $42 million?
Part of the original structure was discretionary and could be avoided through convenience termination. The amendment removes that exit and fixes three payment dates. Willingness and duration are more observable; economic value is not necessarily higher.
What datapoint would most change the view?
First, the full terms and actual settlement in the next 10-Q. Then two consecutive quarters of Data & Applications margin and operating-cash-flow conversion, plus an independent multi-year customer without cross-licensing or equity settlement.
07 · TERMS & SOURCES
Terms, sources and related research
Key terms
- Committed payment
- Contractually due consideration; not necessarily current revenue, cash collected or free cash flow.
- Convenience termination
- A contractual right to exit on notice without showing that the counterparty breached the agreement.
- Share settlement
- Paying in equity instead of cash, leaving the recipient with price, liquidity and monetization risk.
- RPO
- Remaining performance obligations: contracted consideration not yet recognized as revenue; recognition still depends on performance.
- TxFM
- Recursion's RNA foundation model; this license is worldwide, non-exclusive, oncology-only and two years long.
- Free cash flow
- Typically operating cash flow less capital expenditure; not the same as adjusted EBITDA or accounting net income.
[1] U.S. SEC | Recursion Form 8-K | Amendment terms and TxFM consideration | 21 September 2026 ↗
[2] Tempus | Collaboration extension and TxFM license announcement | 21 September 2026 ↗
[3] Tempus | Second-quarter 2026 results ↗
[4] U.S. SEC | Tempus 2026 Q2 Form 10-Q ↗
[5] U.S. SEC | Recursion 2026 Q2 Form 10-Q ↗
[6] Recursion | Second-quarter 2026 results ↗
[7] Tempus | 100,000-sample multimodal whole-genome initiative | 11 September 2026 ↗
This report uses company releases and SEC filings available through 24 September 2026. The full amendment and TxFM agreement remain pending with the signing-quarter 10-Q. Comparable record counts, unit pricing, actual cash/share settlement and recognition timing are not yet disclosed. Company statements about data scale, model efficiency, clinical potential and runway remain forward-looking. This report provides no target price or position recommendation. For research and education only; not investment or medical advice.