Tempus holds more than 8.6 million de-identified patient records (about 1.4 billion documents), over 8 million of them with imaging data, which it licenses to pharmaceutical and AI customers.
more than 8.6 million de-identified patient records; over 8 million with imaging data
Dated 24 Feb 2026Source checked 27 Oct 2026Basis: stated
In the source's own words
“Tempus has amassed approximately 1.4 billion documents, across more than 8.6 million de-identified patient records, including over 1.6 billion pages of rich clinical text that we use to train our large language models. The database also includes over 8 million records with imaging data”
Where it was read
Also read from the same source
- An initial batch of de-identified patient records licensed by Tempus AI to SB Tempus (its 50:50 joint venture with SoftBank) for specified uses in Japan, non-exclusive, transferable, with a limited right to sublicense, term ending 31 March 2026
- GSK's minimum spend commitment to Tempus AI for licensed de-identified data and related services (sequencing, trial matching, organoid modelling, algorithm development), non-exclusive, August 2022 to December 2027
- AstraZeneca's minimum spend commitment to Tempus AI for licensed de-identified data and related services under their master services agreement, non-exclusive, November 2021 to December 2026
- Tempus's Data and applications line (mainly licences of de-identified records, plus trial services) earned $316.4 million in 2025, up 31%, with $70.9 million of the increase from its Insights data licences; signed contracts carried a remaining total contract value of more than $1.1 billion at the end of 2025, and net revenue retention was about 126%.
- Price exclusivity by what it covers: sole distribution earns a little more, sole use by one buyer earns a multiple
- An announced health-data contract total is a ceiling paid over deliveries, not money received
What it is about
Figures it counts in
It is shown on its own, not counted into a figure.