The proposed acquisition would be Anthropic’s largest known deal and could bring Decart AI’s infrastructure specialists into the Claude maker’s inference and performance operation.
Anthropic is in discussions to acquire artificial intelligence startup Decart AI for about $6 billion, in a deal that would significantly expand the AI company’s infrastructure capabilities at a time when computing capacity has become one of the industry’s biggest competitive pressures.
The negotiations remain ongoing and an agreement has not been finalised. People familiar with the private discussions cautioned that the talks could still fall apart. If completed, the transaction would become Anthropic’s largest known acquisition as the company moves toward a widely anticipated initial public offering.
The potential acquisition stands out not simply because of its size. Decart has developed technology aimed at lowering the cost of training AI models by helping computing chips operate more efficiently, a capability that could allow Anthropic to handle greater demand using its existing infrastructure.
Why Decart AI’s Infrastructure Technology Matters?
Building increasingly capable AI models requires enormous amounts of computing power, making the efficiency of chips, data centres and inference systems a central concern for leading AI companies.
Decart’s software is designed to improve how efficiently chips are used during AI workloads. For Anthropic, which has been spending heavily on computing capacity to develop products and support customers, the technology could provide another way to make more effective use of expensive infrastructure.
The discussions also appear to be centred on technical talent. If a deal is completed, Decart’s team is expected to join Anthropic’s inference and performance organisation, according to a person familiar with the matter.
That detail gives the potential transaction a broader strategic dimension.
Rather than simply acquiring another AI product, Anthropic would be bringing in engineers working directly on the performance and infrastructure problems that determine how quickly and economically AI systems can operate at scale.
Decart AI Also Builds Real-Time Generative Video Technology
Decart’s work extends beyond infrastructure optimisation.
The startup is developing generative video technology built around so-called world models, including systems capable of modifying live video feeds in real time. Its work in the area has also been viewed as evidence of the depth of the company’s infrastructure engineering capabilities.
That combination of chip efficiency and generative AI research gives Decart an unusual position in the market. Its technology touches both sides of the AI race: the increasingly expensive infrastructure required to run advanced models and the applications those models can ultimately power.
Representatives for Anthropic and Decart declined to comment on the acquisition discussions.
Decart Raised $300 Million Only Months Before the Talks
The reported $6 billion acquisition discussions come shortly after a major funding round for Decart.
In May, the startup said it had raised $300 million in a round led by Radical Ventures. Nvidia, Atreides Management, Valor Equity Partners and Adobe Ventures also participated, alongside earlier investors Sequoia Capital, Benchmark and Zeev Ventures.
That financing reportedly valued Decart at almost $4 billion, compared with a valuation of $3.1 billion in August 2025.
Decart itself is only a few years old. The company was founded in 2023 by Israeli engineers Dean Leitersdorf, Orian Leitersdorf and Moshe Shalev. Dean and Orian Leitersdorf are brothers.
The combination of its young age, recent funding round and reported acquisition price makes the talks particularly notable in an AI market where infrastructure companies are attracting growing strategic interest.
Anthropic’s Compute Requirements Are Growing
Anthropic has not traditionally relied on large acquisitions, making a potential $6 billion purchase unusually significant for the company.
Its interest in Decart comes as leading AI developers commit vast sums to the computing infrastructure needed to train and operate increasingly sophisticated models. Anthropic and OpenAI have each committed to spending tens of billions, and potentially hundreds of billions of dollars, on data centres equipped with costly AI chips. They are also increasingly relying on hardware from multiple providers to meet growing computing requirements.
Those costs have become one of the defining economic challenges of the current AI expansion. More efficient use of chips can therefore have consequences far beyond technical performance. It can influence how much demand an AI company can serve, how quickly it can deploy new models and how much additional infrastructure it needs to build.
That is what makes Decart’s chip-efficiency work particularly relevant to Anthropic.
Deal Is Still Far From Certain
Despite the reported price and strategic fit, there is no completed transaction.
The acquisition discussions are private, and people familiar with them have stressed that negotiations could still end without an agreement.
For now, the significance of the talks lies in what Anthropic appears willing to consider paying for infrastructure technology and engineering talent. A $6 billion purchase of a company founded in 2023 would represent a major change of scale for an AI developer better known for building its own technology than for making large acquisitions.
If the negotiations produce a deal, Decart would bring Anthropic technology focused on reducing AI training costs, engineers expected to work inside its inference and performance organisation, and additional expertise in real-time generative video.
In an industry where access to computing power is increasingly inseparable from the ability to compete, those capabilities could be the real prize behind the proposed acquisition.
