Anthropic has walked away from its $6bn Decart deal, Bloomberg reports
Anthropic has decided against buying Decart AI, Bloomberg reported, citing people familiar with the matter who asked not to be identified.
The company had explored a deal and completed due diligence before walking away, and one of those people said the two may still find other ways to work together. Representatives for both companies declined to comment to Bloomberg.
The talks were first reported in August at around $6bn, with the same caveat attached even then: nothing had been finalised, and the negotiations could fall apart. They have.
What Anthropic wanted is the detail that has been widely misread since the talks surfaced. Decart is known publicly for world models, but the acquisition rationale was efficiency.
Its optimisation stack is designed to make chips work harder across both training and inference, and the idea, per Bloomberg, was to let Anthropic’s existing compute absorb more demand.
The company’s own website promises to “squeeze every ounce of performance from every chip”. This was a deal about the cost of serving customers, not about entering a new product category.
That matters because of what Anthropic is doing next. The company rarely makes large acquisitions, and its most notable to date was $400m for a team of fewer than ten people.
Six billion would have been an order of magnitude larger, and it comes while the company is spending heavily on compute ahead of an initial public offering it expects to match or exceed SpaceX’s record listing, according to people who have described the preparations to Bloomberg.
Buying a way to make existing infrastructure go further is a rational thing to want in that position. Paying a substantial premium for it, weeks before a listing, is a harder line to defend to incoming shareholders.
Nobody involved has said which consideration ended the talks, and the reporting does not establish whether Anthropic balked at the price or at something diligence turned up.
A company preparing to list has every reason to want its cost base looking efficient and its acquisition history looking disciplined, and those two impulses point in opposite directions here. Buying Decart would have improved the first and complicated the second.
Decart raised $300m in May in a round led by Radical Ventures, with NVIDIA, Adobe Ventures, Valor Equity Partners and Atreides Management joining, alongside earlier backers Sequoia Capital, Benchmark and Zeev Ventures.
That round valued the company at close to $4bn, per the Wall Street Journal, up from $3.1bn the previous August. A $6bn purchase would have been roughly a 50% mark-up four months later.
Founded in 2023 by the brothers Dean and Orian Leitersdorf with Moshe Shalev, Decart runs two businesses that sit unusually far apart. One is the chip-efficiency layer Anthropic was interested in.
The other is world models, systems trained on text and millions of hours of video to internalise how physical objects behave, aimed at applications from autonomous driving to online retail.
Its Lucy model takes live video of a person and generates high-resolution video of them wearing something they are not, which is the virtual try-on problem that has defeated fashion e-commerce for a decade. eBay is both an investor and a customer.
Chief executive Dean Leitersdorf told Bloomberg at a conference in Paris in July that the technology is also used for live streaming on Twitch, TikTok and YouTube.
Whether the collaboration the two sides are said to be still considering means a commercial agreement for that efficiency stack, rather than ownership of it, is the thing to watch. It would be the cheaper version of the same idea, and it would not need explaining in a prospectus.
Get the TNW newsletter
Get the most important tech news in your inbox each week.
How it works
Once you click Generate, Ollama reads this article and crafts 5 comprehension questions. Your answers are graded against the article content — general knowledge won't be enough. Score 70+ to count toward your certificate.
Questions are cached — you'll always get the same 5 for this article.