Vertical giants build proprietary models as private data value surges

pratyushmaini · x · 2026-09-01

Trend: As open-source base models improve and post-training infra matures, companies with massive vertical datasets are moving beyond data licensing to training their own models.
Advantage: While general-purpose frontier models lead in breadth, vertical models offer superior leverage on specific proprietary data, avoiding the high R&D costs of competing with top labs.
Example: The partnership between @datologyai and @thomsonreuters is cited as the first public example of this shift in action.

More from Companies & People