MidCurveCapital
Equity Investor, Biotech & Software Focus
Diagnostics companies such as Tempus AI and Personalis control data AI labs cannot easily get, because it is a byproduct of running diagnostics businesses that the labs have neither the appetite nor the skill to build. Per David Friedberg, via @dnapway, nearly every big pharma company refused Anthropic's request to share proprietary data for a new life sciences model. That leaves @lefkofsky's Tempus holding the winning hand. x.comÂ
Tempus AI is a data company that happens to sell diagnostics, positioning it for the next re-rate in AI-enabled healthcare and drug discovery. It holds 45M+ de-identified records and 500+ petabytes, growing every time a clinician runs a test. Diagnostics funds the moat, and the data business is the moat, with 41% growth and 73% margins, a layer nobody can rebuild from scratch. x.comÂ
SiTime, Tempus AI and Lemonade look like AI beneficiaries that need not rely on rising frontier-model spending, unlike a crowded trade vulnerable to concentrated capital and circular financing. SiTime can expand into robotics even if data-center growth slows; clinical decisions still need Tempus; Lemonade's models may improve loss ratios and retention while lower compute costs could help it. x.comÂ
Tempus AI is a second- or third-tier genomic lab whose AI story is marketing mumbo jumbo, and its management team are promoters. x.comÂ
Tempus AI has a data moat no one has replicated: a multimodal, patient-linked genomic, clinical and imaging dataset at scale. It holds over 8 million imaging files, 1.5 million matched clinical and molecular profiles and 4.5 million sequencing samples, which drug companies leaning on AI increasingly need. It is expanding from oncology into cardiology, neuropsychiatry and radiology. Revenue grows 25% and gross margin rose from 32% in 2021 to 64%, though last quarter's profit was a one-time item. … Show more
Tempus AI could be a picks-and-shovels play on AI pharma, since its data was reportedly used in Moderna's cancer cure development process. It is a speculative holding: whether the company is legitimate is hard to tell. x.comÂ
Tempus AI is a short idea because the stock is being pumped on recent FDA clearances for its EKG-AI tools, while the revenue from them is far off and may never materialize. x.comÂ
Tempus AI is applying Palantir's playbook to US healthcare. Proprietary data from more than 5,000 hospitals and two-thirds of US academic centers let it train a healthcare AI no one else can. At 4x sales it is priced like a lab, though it is an AI company. x.comÂ
Hims & Hers' compounding direct-to-consumer distribution makes it stupidly undervalued next to Tempus AI. Hims has $2.6B in trailing revenue at a $6.7B market cap, with 66.5% annual revenue growth from Q3 2021 to Q2 2026; Tempus has $1.4B in revenue, a $14.9B market cap and 40.4% annual growth. x.comÂ
Tempus AI may go lower near term because of valuation, leadership reputation and insider selling, despite an amazing business. With about 25% growth expected and profitability uncertain, it is not obviously cheap. Glassdoor reviews matter in medicine's high-trust setting, and ongoing insider sales echo founder patterns that ended badly. The data piece, about 5% of market cap, can earn a high multiple, so the stock looks reasonably priced if adoption continues. x.comÂ
Tempus AI's diagnostics-to-data-to-AI flywheel makes its roughly 7.2x forward sales interesting, about half Natera's and Guardant's 16x, though both grow faster. Q2 revenue rose 22% to $382.5M, Data Licensing & Modeling grew 36%, and 2026 guidance is about $1.6B of revenue and $65M of adjusted EBITDA. As a ~25% growth diagnostics company, ~7x may be reasonable; if growth reaccelerates toward 30%+, Data + Applications grows much larger and margins expand, the valuation could look very different. … Show more
