Amazon in Talks to Buy AI Startup Decart for $7 Billion

Amazon is discussing the purchase of AI startup Decart, and the deal could value the company at about $7 billion. The Wall Street Journal reported this on Sunday. The story was written by Keith Clark and Lauren Thomas, and its sources are people familiar with the negotiations. According to them, an agreement could come soon if the talks don't collapse and another buyer doesn't step into the process. The available text contains no official comments from either party, so for now these are talks, not a closed deal.

What we know about the deal

There are few facts so far, and it's worth separating what the outlet confirmed from background context.

• Buyer: Amazon
• Target: Decart, an AI startup
• Valuation benchmark: about $7 billion
• Source: The Wall Street Journal, anonymous sources
• Status: talks; no signed agreement
• Condition stated in the report: a deal is possible if the talks don't fall apart

The report does not disclose payment terms, deal structure or closing timeline. It is unclear whether this would be a full acquisition or a more complex arrangement. Regulatory risk is also hard to assess for now: Amazon has long been under close scrutiny from antitrust agencies, and large AI acquisitions get separate review.

What Decart does

According to earlier reports, Decart was founded by Israeli entrepreneurs Dean Leitersdorf and Moshe Shalev. The company is known for generative models that render images in real time. Its most visible project is Oasis, a Minecraft-style game world in which a neural network, not a game engine, draws every frame. The team later showed models that process a video stream on the fly. These details come from past coverage, not from the WSJ story, and should be checked separately against current sources.

The same public data says Decart raised about $100 million in 2025 at a valuation of roughly $3.1 billion. If the $7 billion figure is confirmed, the company's value would more than double in about a year. For a startup whose products are still closer to technology demos than to a mass-market business, that is a high price. The buyer, by all appearances, is paying for the team and its knack for squeezing speed out of GPUs.

Why Amazon wants it

Amazon already has its own Nova model family, including video generation, and its own Trainium chips for training and inference. It has also put billions of dollars into Anthropic, but it evidently doesn't want to depend on a single model partner. Buying Decart would fill a different gap: low-latency generation of video and interactive environments.

Where this could be useful remains hypothesis for now:

• Video and advertising. Generating and editing clips on the fly for Amazon's ad products
• Games and streaming. Twitch and Prime Video work with live video streams, and Amazon Games remains a notable player
• Cloud. Real-time models give AWS a reason to sell more inference on Trainium
• Retail. Virtual try-on and interactive product listings

The WSJ did not name any of these scenarios. This is reasoning, not fact.

What to watch next

For anyone following the market or building products on generative video, there are a few practical markers.

First: wait for official confirmation and terms. The phrase "in talks" often ends in a deal, but not always. Second: if Decart goes to Amazon, check whether open access to its API and models will remain, and if you depend on them, set up a fallback. Third: watch how competitors react. Google, Meta and OpenAI are also investing in video and "world models," and the acquisition of a startup with this profile will force them to assess what they lack. Fourth: for investors, the most important signal is the multiple. If the $7 billion figure is confirmed, the market gets a fresh benchmark for valuing small teams with narrow but strong technology.