Anthropic Abandons $6 Billion Acquisition of Israeli AI Startup Decart
Anthropic has decided to withdraw from talks to acquire Israeli AI startup Decart AI for approximately $6 billion. The companies may still explore other collaboration options.

Anthropic has decided not to proceed with the acquisition of Israeli artificial intelligence startup Decart AI, according to a Bloomberg report citing sources familiar with the matter. Anthropic had explored the potential transaction and conducted due diligence on Decart, but ultimately chose to withdraw from the deal. However, the companies may still explore other forms of collaboration, according to one of the sources. Representatives for both Anthropic and Decart declined to comment.
Anthropic, the developer of the Claude chatbot, had been in talks to acquire Decart in a deal valued at approximately $6 billion, though no final agreement was reached and there was always a possibility that negotiations would fall through.
Infrastructure and Efficiency
Decart AI develops software designed to reduce the training and operational costs of artificial intelligence systems by improving chip efficiency. The strategic rationale behind the potential acquisition was to help Anthropic's existing computing infrastructure handle larger demands. Anthropic, which rarely makes large acquisitions, is investing heavily in computing power to develop new products and serve customers ahead of its planned initial public offering (IPO). According to sources familiar with the preparations, the company aims to raise an amount in its IPO that would equal or exceed the sum SpaceX plans to raise in its own offering.
Israeli Founders and Technology
Decart was founded in 2023 by three Israeli engineers: brothers Dean and Orian Leitersdorf, and Moshe Shalev. The startup offers an optimization system designed to help AI developers "squeeze every drop of performance out of every chip," according to the company's website. The solution is designed for use both in training AI models and in running them after they have been trained, a process known as inference.





