As the global artificial intelligence race shifts from pure model parameter scaling to infrastructure cost optimization, the industry's major players are making massive bets on efficiency. In what would mark the largest acquisition in its corporate history, AI research lab Anthropic, the developer behind the Claude chatbot ecosystem, is in advanced talks to acquire Israeli AI startup Decart AI for approximately $6 billion, according to financial reports confirmed by Bloomberg and Reuters.
While negotiations remain ongoing and a final agreement has not been executed, the potential deal underscores how critical inference speed, chip efficiency, and hardware optimization have become as leading labs prepare for public market debuts. For Anthropic, the acquisition represents not just a technology purchase but a strategic bet that the company winning the next phase of AI will be the one that can deliver intelligence at the lowest cost per token.
Checking the Ledger | The Decart AI Profile
Founded in 2023 by former Israeli military intelligence Unit 8200 engineers Dean Leitersdorf, Orian Leitersdorf, and Moshe Shalev, Decart has experienced one of the fastest valuation ascents in modern technology. The startup was valued at $4 billion following a $300 million Series B led by Radical Ventures in May 2026, meaning the $6 billion acquisition price would represent a 50 percent premium in just three months.
Decart initially drew public attention for its flashy generative video capabilities, including its Oasis demo, which generates interactive, simulated environments in real time, and Lucy, a real-time live video editing tool. However, its core appeal to Anthropic lies in the unglamorous layer of chip optimization. Decart develops specialized software designed to maximize throughput and reduce latency across a broad spectrum of AI hardware, including Nvidia GPUs and custom cloud accelerators.
This optimization layer is invisible to end users but critical to the economics of running a large-scale AI service. As demand for Claude across enterprise software and developer APIs continues to surge, serving queries at scale represents the majority of an AI lab's operational expenditure. Integrating Decart's optimization stack allows Anthropic to serve significantly more user queries using its existing compute footprint, directly lowering electricity and hardware overhead.
Why Decart | Squeezing Performance Out of Expensive Compute
The $6 billion price tag reflects intense strategic demand for Decart's infrastructure talent. Industry insiders note that semiconductor giant Nvidia, which participated in Decart's $300 million funding round in May, held preliminary acquisition discussions with the startup before higher offers emerged from rival tech firms and cloud providers. The fact that Nvidia itself, the dominant supplier of AI hardware, considered acquiring Decart underscores the strategic value of the startup's optimization technology.
The compute efficiency equation driving the deal is straightforward. Growing demand for Claude requires expanding data center capacity, which drives up power and chip costs. Decart's inference optimization directly reduces the cost per token, allowing Anthropic to serve more queries at lower operational expense. Every percentage point improvement in inference efficiency translates into millions of dollars in annual savings at Anthropic's scale, and the proprietary nature of the optimization stack means those savings are not available to competitors.
This is particularly important given the growing tension between AI labs and their enterprise customers over token pricing. Anthropic CEO Dario Amodei has faced criticism from enterprise buyers who argue that frontier model pricing remains too high relative to inference costs. A proprietary optimization stack that demonstrably lowers cost per token strengthens Anthropic's position in those negotiations while improving the unit economics of its API business, a dynamic that mirrors the broader infrastructure efficiency race across the AI industry.
Bidding War and the Path to an IPO
For Anthropic, the transaction arrives at a pivotal structural moment. Coming on the heels of major cloud capacity expansions and confidential S-1 filings ahead of an anticipated initial public offering, securing proprietary chip optimization technology provides a key financial defense. By demonstrating to Wall Street that it can aggressively control its long-term inference margins, Anthropic is positioning itself not just as a creator of frontier intelligence, but as a sustainable, cost-disciplined enterprise.
The contrast with Anthropic's recent export control battles is instructive. While the company has been fighting regulatory battles over the global availability of its Claude Fable 5 and Mythos 5 models, it has simultaneously been building the infrastructure to deliver those models at significantly lower cost. The Decart acquisition, if completed, would accelerate that cost reduction trajectory dramatically, putting Anthropic in a stronger position against cost-competitive open-weight models emerging from China.
The $6 billion price also signals a broader market dynamic. AI infrastructure startups are becoming the most sought-after acquisition targets in technology, as the major labs recognize that the marginal value of a 10 percent improvement in inference efficiency far exceeds the marginal value of comparable improvement in model capabilities. When the market is expanding as fast as AI, the company that can serve demand at the lowest cost wins, and Decart's optimization stack is a direct path to that cost advantage.