A three-year-old startup that most people had never heard of a month ago could end up anchoring one of the biggest tech deals of the fall. According to The Wall Street Journal, Amazon is negotiating to buy Israeli company Decart for roughly $7 billion, with talks reportedly far enough along that a deal could close within weeks.
Decart was founded in 2023 by brothers Dean and Orian Leitersdorf along with Moshe Shalev. The team splits its time between San Francisco and Tel Aviv, and the company's real asset is a pair of models rather than one. Lucy edits live video on the fly — shifting lighting, backgrounds and style mid-playback without re-encoding or lag. Oasis is the more ambitious project: it generates interactive simulated environments where robots and self-driving systems can be trained, effectively rendering a virtual world in real time like a video game with no pre-written script.
In May, Decart raised $300 million led by Radical Ventures, with Nvidia joining as a surprise new backer at a valuation near $4 billion. One detail stands out: the founders had earlier turned down a richer offer directly from Nvidia, choosing to stay independent instead. Total funding now tops $400 million, with Sequoia Capital and Benchmark also on the cap table.
What happened between May and October is the real twist. In August, Anthropic opened talks to buy Decart for about $6 billion in stock. Due diligence dragged through September, then collapsed — not over price, but because Anthropic wanted the Tel Aviv team relocated to San Francisco. The founders said no, and the deal died.
Amazon doesn't seem deterred by that failure — if anything, it's offering more money. The company has its own stake in the outcome: Lucy posted striking results on Trainium3 chips at AWS re:Invent back in December, generating frames four times faster at half the cost compared with GPUs. For Amazon, which has spent years trying to convince the market its own chips can rival Nvidia's, buying the team that squeezed those numbers out of Trainium looks less like a bolt-on acquisition and more like a strategic necessity.
Neither company has confirmed anything publicly, and the Journal notes the talks could still fall apart or a third bidder could step in — so $7 billion isn't locked in. But even as a rumor, the story says something bigger: demand for teams that can generate video and simulate the physical world in real time is intense enough that valuations are doubling within months, with two AI giants willing to fight over the same small group of engineers.
