Anthropic in Talks to Buy AI Startup Decart for $6 Billion Ahead of IPO

Anthropic's Reported $6 Billion Acquisition Talks With Decart

AI developer Anthropic is in talks to acquire Israel-based startup Decart for about $6 billion, according to Bloomberg sources cited by Forbes. The negotiations have not reached a final stage, and people familiar with the matter caution that the deal could still collapse.

The attraction is not a consumer chatbot. Decart is developing so-called world models, which aim to simulate physical reality, as well as software intended to reduce the cost of training AI models and improve the efficiency of chips used in that training. A source told Bloomberg that the technology would help Anthropic use its existing infrastructure more efficiently, build new products and meet rising demand.

If completed, the acquisition would be the largest in Anthropic's history. The Decart team would reportedly join the unit working on AI model performance and operations. The move arrives just as Anthropic prepares to go public in October; it has already filed with the US Securities and Exchange Commission. The Financial Times reports that investors expect a record listing that could value the company at around $2 trillion, surpassing SpaceX's $1.75 trillion mark.

Decart was founded in 2023 by three Israeli engineers, brothers Dean and Or Leitersdorf and Moshe Shalev. In May the company announced a $300 million funding round led by Radical Ventures, giving it a valuation of almost $4 billion. That was up from $3.1 billion in August 2025, underlining how quickly AI infrastructure startups have gained value.

What the Decart Deal Means for Anthropic's Infrastructure and IPO

The Deal Logic: Buying Training and Chip Efficiency, Not a New Model Brand

The reported rationale is operational rather than content-facing. Anthropic would gain software aimed at reducing model-training costs and improving chip efficiency. That matters because large AI labs' biggest constraint is often compute supply and cost, not raw model quality. Decart's world-model work is more experimental, but it points toward physical-world simulation that could eventually support products beyond text and code. The key link: Anthropic expects the technology to improve how it uses infrastructure it already has.

A Significant Premium Before a Landmark IPO

The reported $6 billion price would be roughly 50 percent above the almost $4 billion valuation Decart received in May, and nearly double its $3.1 billion valuation from August 2025. That premium is not just a sign of Decart's momentum; it also reveals urgency. With an October IPO planned and a $2 trillion valuation expectation reported by the Financial Times, Anthropic may be trying to present a stronger infrastructure story to potential public investors. The cost is high relative to Decart's known scale, but it is modest next to the IPO valuation target.

The Competitive Signal for AI Infrastructure

The deal would not affect every rival in the same way. Labs that already own their training infrastructure may be pushed to prove their per-chip efficiency as costs rise. Cloud providers and enterprises buying AI services could benefit if Anthropic converts efficiency gains into expanded capacity or more competitive pricing. But no such integration or pricing change is confirmed; the talks are still ongoing, and deal terms are not public.

What Anthropic Investors, Rivals and AI Customers Should Track

Because the acquisition is not final, none of the following should be treated as a completed shift. They are the specific points tied to this deal that investors, competitors and AI customers can follow.

• For potential Anthropic IPO investors: The reported $6 billion price compares with Decart's $4 billion May valuation and its $3.1 billion August 2025 mark. If a deal is signed, check SEC filings for how the purchase is paid and whether the integration is weighted toward cash or equity dilution.
• For rival AI labs: Decart's software targets lower training costs and better chip efficiency. If Anthropic integrates it successfully, competitors with less efficient internal infrastructure may face pressure to match per-unit compute economics.
• For enterprise AI customers: Anthropic's stated rationale includes meeting rising demand and developing new products. A completed deal would make capacity expansion and accelerated product releases the main things to assess in the next two to three quarters.
• For Decart employees and backers: The team is expected to join Anthropic's model performance group. Retention, earn-out and deal-collapse risk are not yet disclosed, so final terms should be treated as open until an agreement is signed.