On October 13, 2026, Calcalist reported that Decart shareholders were in talks with Amazon and put the proposed valuation at $5.5 billion. The Next Web separately relayed a Wall Street Journal report of talks at about $7 billion, noting that it had not independently verified the WSJ account. Calcalist reported the $5.5 billion figure; The Next Web relayed the WSJ’s roughly $7 billion report.
The figures belong to separate reports, not agreed transaction terms. Decart’s work with Amazon’s Trainium chips offers technical context for the reported interest.
Why the reports cite different valuations
Calcalist’s report put the proposed valuation at $5.5 billion. The roughly $7 billion figure came from the Wall Street Journal, as relayed by The Next Web, which said it had not independently verified that report. The two accounts do not describe a single agreed price.
For context, Calcalist reported that Decart raised $300 million in May 2026 at a $4 billion valuation. That was a funding-round valuation, separate from the reported Amazon discussions.
The talks followed earlier acquisition discussions involving other companies. The Next Web reported that Anthropic withdrew from discussions in September 2026 after due diligence. Calcalist said Nvidia declined to reopen negotiations before Decart shareholders pursued talks with Amazon.
Decart’s software and Amazon Trainium
Calcalist described Decart’s software as helping AI workloads run more efficiently across different chip architectures. That capability could be relevant to Amazon’s Trainium chips, although it is strategic context—not a stated reason from Amazon for pursuing a deal.
Amazon said in a December 2, 2025 announcement that Decart achieved four-times-faster frame generation at half the cost of GPUs for real-time generative video on Trainium3. The statement is Amazon’s account of Decart’s work.
At AWS re:Invent 2025, Decart’s CEO said the demonstrated setup reached 80% tensor-core utilization—the share of specialized AI computing cores being used for a workload. The CEO also said Decart worked with the AWS team to port its model stack and infrastructure to Trainium.
