Artificial intelligence leader Anthropic PBC is accelerating efforts to strengthen its technological moat through a major acquisition just weeks ahead of its highly anticipated initial public offering. According to sources familiar with the matter, the AI standout is in advanced negotiations to acquire Israeli AI infrastructure startup Decart AI in a deal valued at approximately $6 billion. If completed, this would mark Anthropic's largest-ever acquisition and its most critical technology enhancement move before going public. Notably, chip giant Nvidia was originally the potential buyer in these negotiations, but Anthropic swooped in to snatch the deal just as talks were nearing completion.

Decart AI, founded in 2023 in Israel by brothers Dean Leitersdorf and Orian Leitersdorf along with Moshe Shalev, focuses on maximizing the operational efficiency of various chips. Its software optimizes the training and inference processes of AI models, helping developers extract more computing power from existing hardware. The startup's technology spans a chip performance optimization layer that enables AI developers to run models efficiently on everything from Nvidia to AMD chips, providing strategic value amid the current scarcity of AI computing power. It also develops generative video and "world models" that can modify live video streams in real time, showcasing its pool of high-quality infrastructure talent.

In May, Decart completed a $300 million funding round led by Radical Ventures, with participation from Nvidia, Atreides Management, Valor Equity Partners, and Adobe Ventures, achieving a post-money valuation of nearly $4 billion. This means that in just three months, Anthropic's offer has bumped Decart's valuation by 50%. Adding to the drama, Decart was close to finalizing a sale agreement with Nvidia, but when a "larger" buyer stepped in near the end of talks, the founders decided to switch directions. Industry observers widely believe that the "interceptor" is Anthropic.

For Anthropic, the potential $6 billion acquisition carries three layers of strategic significance. First, it creates a leverage effect for computing efficiency. Anthropic has committed hundreds of billions of dollars to building data centers equipped with expensive chips. Decart's optimization technology allows existing infrastructure to handle greater demand, effectively adding leverage to Anthropic's computing investments in the context of ongoing chip supply constraints. The Decart team is expected to be integrated into Anthropic's inference and performance organization. Second, it serves as a valuation safeguard ahead of the IPO. Anthropic plans to complete its listing in September or early October, targeting a valuation between $965 billion and $1 trillion. Completing a high-profile strategic acquisition just before going public helps craft a more complete "full-stack AI" narrative for public market investors. Third, it creates a differentiation against OpenAI. While OpenAI continues to compete on scale, Anthropic is building a moat on efficiency. Decart's ability to deliver higher performance on the same chips complements Anthropic's "safe and controllable" technology philosophy.

Anthropic's IPO is entering its final sprint. The company has already confidentially filed a draft S-1 registration statement with the Securities and Exchange Commission. Prediction market platform Kalshi shows that traders now see an 85% probability of Anthropic announcing its listing within 2026. However, cracks have appeared in the trillion-dollar valuation story during pre-IPO investor meetings, with investors focusing on three major concerns: the competitive threat from low-cost Chinese AI systems, tensions between Anthropic and the Trump administration, and the cautionary example of SpaceX, whose stock price plunged from $225 to $108 after listing.

Anthropic's pursuit of Decart reflects a deeper trend in the AI industry: foundation model companies are racing to acquire inference and efficiency capabilities rather than relying entirely on in-house development. As AI model sizes continue to balloon and computing costs remain high, "efficiency" is becoming a competitive dimension as important as "performance." Meanwhile, both OpenAI and Anthropic have committed hundreds of billions or even trillions of dollars to building data centers. Under the pressure of enormous capital expenditures, improving the utilization efficiency of existing infrastructure through acquisitions is emerging as a more cost-effective path. Nvidia, originally an investor and potential buyer of Decart, lost the deal to Anthropic in the final stages. This reveals a deeper trend: AI model developers are proactively building their own technology stacks rather than relying solely on chip suppliers. When competition over raw computing power becomes homogeneous, "how to use computing power more efficiently" is becoming the core competitive advantage of AI companies. The Decart acquisition marks the official start of the AI efficiency war. Of course, the deal still faces uncertainty—sources emphasize that negotiations are not finalized and could still fall through. But for Anthropic, aiming for a trillion-dollar IPO, acquiring Decart AI is no longer just a business decision—it is a declaration of technological ambition to the market.