Tempus AI, one of the most hotly contested stocks in the U.S. stock market right now.

On one side, Cathie Wood has it as the top holding in the ARKG fund (9.49% weight), aggressively adding shares at the $43 price level.
On the other side, there are short reports from Spruce Point, collective securities lawsuits, the CEO selling over $34 million, and an 18% short interest.
One side is the world's largest AI medical data platform. The other side is questioning "funds flowing back" and swirling related-party transactions.
I did one thing: ignoring stock price and sentiment, just asking one question—if we rebuilt Tempus from scratch today, how much would it cost?
The answer surprised me.
1. What is Tempus AI?
It is the healthcare version of Palantir.
Palantir aggregates fragmented government and enterprise data to create intelligence value with AI. Tempus does the same—data sources are hospitals, clients are pharmaceutical companies and doctors.
Core asset: a multimodal database from over 45 million patients across 4,500+ U.S. hospitals—genomic sequencing + electronic health records + imaging + pathology slides, all structured and linked.
Over 350PB of data, the largest in the world.
Two revenue streams: providing AI-driven precise treatment matching to doctors, and licensing de-identified data to pharma companies for drug R&D.
Nineteen of the top 20 pharma companies are clients.
FY2025 revenue: $300k (+83%), first EBITDA positive in Q3 2025. But the stock price has fallen from $104 to $43, a 59% drop.
2. Reset Cost Method: How much does it cost to rebuild from zero?
I broke down Tempus into 7 asset categories and estimated the rebuilding costs item by item.
Asset 1: Multimodal clinical database (accounts for 55-65% of total value)
This is the core of the core. 40 million+ patient records, 4 million+ genomic samples, 7 million+ pathology slides, 1.9 billion+ radiology images.
How to price? Look at comparable transactions—
In 2018, Roche acquired Flatiron Health for $1.9 billion, at that time Flatiron had only 2.2 million patient EHR records. Implied per patient record: about $860–950.
The same year, Roche bought Foundation Medicine for $5.3 billion, which had about 300k genomic profiles. Implied per genome: roughly $17,667.
4M genomic samples × $1,500–$3,000 per sample (net sequencing cost after insurance reimbursement) = $60–12B.
Just sequencing costs alone are close to or exceed the current market cap.
Conservative valuation: $4 billion, neutral: $8 billion, optimistic: $10 billion.
Asset 2: CLIA-certified laboratory network
Four labs (Chicago/Atlanta/RTP/California Ambry). Ambry Genetics alone cost $600 million to acquire.
Subtotal for labs: $665–745 million.
Asset 3: FDA-approved product portfolio
xT CDx (648-gene companion diagnostics, PMA approved), xR RNA (510k), ECG-AF (the first FDA-approved AI for atrial fibrillation), Paige (the first FDA AI pathology).
Average development cost for PMA: $94 million, first-pass success rate only 30–40%.
Including regulatory risk premium: $170–450 million.
Asset 4: Data network from 4,500+ hospitals
Cost per hospital data pipeline: $115,000–485k (tech integration + legal compliance + BD).
Plus centralized platform and cloud infrastructure: $510k–3.25 billion.
Reference: Roche valued Flatiron’s 265 oncology clinics at $1.9 billion. Tempus’s network of 4,500+ is 17 times larger.
Asset 5: AI models + 271 patents
R&D investment over three years: $664 million.
PRISM2 foundational model with AZ partnership valued at over $200 million.
Over 2,000 top-tier journal papers.
Estimated value: $485k–2.06 billion.
Asset 6: Pharma client relationships
19/20 top pharma companies, TCV over $1.1 billion, NRR 126%.
Estimated value: $300–500 million.
Asset 7: 2,800+ talent team
AI engineers + bioinformatics scientists + clinical staff.
Estimated value: $200–300 million.
3. Summary: $10.9–21.7 billion vs. market cap of $21.7B
Scenario | Original Rebuilding Cost | Adjusted Rebuilding Cost
Conservative | $7.15 billion | $10.9 billion
Neutral | $12.75 billion | $21.7 billion
Optimistic | $17.3 billion | $37 billion
Adjustment factors include: time-opportunity cost premium (10 years cannot be compressed), regulatory risk premium, network effect flywheel premium, functional depreciation.
Even the most conservative estimate of $10.9 billion exceeds the current $7.6B market cap by 35%.
In the neutral scenario, the market only values Tempus at 37% of its rebuild cost.
Another perspective: in March 2026, Abbott acquired Exact Sciences for $23 billion (EV/Revenue 7–8x).
Exact’s data assets are far inferior in multimodal coverage and AI capability compared to Tempus.
4. Why does the market only assign 37% of the rebuild cost?
Because good assets ≠ good stocks. The reasons for market discounting are clear:
1. Securities collective lawsuit (Shouse v. Tempus, filed June 2025)
Five allegations: inflated contract values, SoftBank joint venture "funds flowing back" (SB Tempus FY2025 revenue $65.3 million vs. FY2024 only $4.5 million), aggressive billing by Ambry, AstraZeneca paying $200 million via CEO-related company Pathos as a "bridge," and overly optimistic core business prospects.
The company says "without basis, will vigorously defend." The case is still ongoing.
2. CEO control + insider selling
Lefkofsky controls 64.3% voting rights via Class B shares. Over the past six months, 118 insider sales with zero buys. The CEO himself sold over $34 million.
He is co-founder of Groupon—after IPO, the company’s market cap fell from $16 billion to less than $1 billion now.
3. Growth rate dropped from 83% to 25%
Of FY2025’s 83%, about 50% came from Ambry acquisition. Organic growth: 28–33%.
Guidance for 2026 at 25% reflects high base effects from acquisitions, not business deterioration. But the market is waking up from "AI super stock" expectations.
5. Cross-validation with 10 valuation methods
Method | Implied Target Price
Peer P/S comparison | $71

EV/Revenue growth premium | $60

Abbott/EXAS acquisition reference | $66

DCF | $42

Segment valuation | $53

PEG adjustment | $63

Forward P/E discounting | $37

Rebuild cost method (conservative) | $62 ($1.09 billion ÷ 176 million shares)
Rebuild cost method (neutral) | $123 ($10.9B ÷ 176 million shares)
ARK implied long-term | $150+
Including the median of 11 methods with rebuild valuation, about $62, current $43 has roughly 44% upside.
14 analysts: 8 buy / 6 hold / 0 sell. Consensus target: $76–79.
6. Conclusion
The rebuild cost method tells us one thing: assets built over 10 years are only valued at a 40% discount by the market.
Is this discount reasonable? Partly—lawsuits, governance, insider selling are real issues.
But is the 35–63% discount excessive? If you believe in two things, the answer is yes—
First, AI-driven precision medicine is one of the biggest healthcare investment themes for the next decade.
Second, once the 40 million+ multimodal patient database is built, time is the deepest moat—money can accelerate but cannot eliminate the 10-year accumulation.
Time is non-compressible, data is non-replicable—that’s the core value revealed by the rebuild cost method.
Key catalysts: Q1 2026 earnings report, FDA approval of xF liquid biopsy, new drug collaborations, lawsuit developments.
Buy on dips, stop-loss at $33–35.
Good data, complex people, let time give the answer.
⚠️: This article is solely personal investment research and does not constitute any investment advice! #危机投资模型