Amazon in talks to buy Decart AI startup
People familiar with the matter say the talks could lead to a deal worth about $7 billion to buy Decart, the Israeli‑American company that makes software for switching between AI chips.
Amazon is in talks to acquire the startup Decart, which specializes in AI technologies, in a deal that could exceed about $7 billion, according to people familiar with the negotiations who were briefed on the talks on Monday, Oct. 12.
Decart, founded in 2023 by three Israeli engineers — Din, Orian Leitersdorf and Moshe Shalif — is based in San Francisco and develops software that allows developers to switch between different AI chips. It is also working on two models called Lucy and Oasis to generate and simulate realistic three‑dimensional environments.
The sources said the deal could be concluded soon unless the talks falter or another buyer emerges. Decart had more than $400 million in venture capital funding from investors including Sequoia Capital, Benchmark and Radical Ventures, and was valued at about $4 billion last May, according to specialized data.
Amazon is seeking to bolster its AI chip capabilities and has developed the Trainium3 chip to compete with firms that dominate the GPU market. Reports said Amazon’s cloud services granted Decart early access to Trainium3 in 2025 after a meeting between the parties and expressions of admiration for the company’s founders.
The current talks are part of Amazon’s intensive efforts to expand its AI stack from chips to cloud services. Among deals and investments the company has made or announced previously are the acquisition of MGM Studios for $6.5 billion and another announced acquisition valued at about $11 billion.
In additional moves to support the development of custom chips and processors, Amazon Web Services in September 2026 announced a multi‑year agreement to license chip‑design technologies worth more than $1 billion. In February a strategic partnership was announced that included a planned $50 billion investment and a commitment to deploy large amounts of Trainium computing capacity within its cloud infrastructure.
