// news · industry · multimodal2026-08-14source: Fortune / reporting

Anthropic is in talks to buy Decart for about $6bn — its largest deal, and it is about inference cost

Reported on 13 August: Anthropic in discussions to acquire the Israeli startup Decart for roughly $6bn, a premium of about 50% over its ~$4bn May valuation. Decart's team would join Anthropic's inference and performance organisation. It would be Anthropic's fifth acquisition of the year and by far its largest.

Read the destination rather than the headline. If Decart's team lands in the inference and performance organisation, this is not a video-generation acquisition wearing a strategy. It is a purchase of chip-utilisation engineering, and the video products are the demonstration that the engineering works.

Decart, founded in 2023 by Dean and Orian Leitersdorf and Moshe Shalev, builds world models plus software that lowers training cost by improving how efficiently chips are used. Its Lucy model transforms live video in real time; its Oasis model generates simulated environments for training and testing robotics and autonomous-driving systems. Real-time video generation is an unforgiving efficiency benchmark — you either hit frame rate or you do not.

The economics explain the premium. Anthropic has committed to compute on a scale where a percentage improvement in utilisation is worth a great deal against fixed capacity, and buying a team that has demonstrably squeezed hardware is a faster route than hiring one.

Caveat where it belongs: this is reported as talks, valuations are unconfirmed, and deals fall through. A ~50% premium over a May valuation of nearly $4bn — itself up from $3.1bn in August 2025 — is what a strategic buyer pays when the alternative is building the capability itself.

See our analysis →

Fortune — Anthropic said in talks to buy startup Decart for $6 billion → · Quartz — Anthropic in talks to acquire Israeli AI startup Decart for $6 billion → · TechRepublic — Anthropic Reportedly Eyes $6 Billion Decart Acquisition to Boost AI Efficiency →