Anthropic, the San Francisco-based artificial intelligence company behind the Claude family of large language models, is in advanced discussions to acquire Decart, an Israeli AI startup specializing in real-time model inference technology, according to a report by Globes, Israel’s leading business daily. The deal, if completed, would mark one of the more strategically significant acquisitions in the competitive AI sector this year, potentially giving Anthropic a meaningful edge in the race to deploy faster, more responsive AI systems at scale.

The move reflects growing investor and corporate appetite for Israeli AI talent, a theme explored in depth in coverage of Israel’s AI strategy, which has positioned the country as a global hub for deep-tech research and development. Decart has drawn attention for its work on making AI inference — the process by which trained models generate outputs — dramatically faster and more compute-efficient, a capability that sits at the heart of commercial AI deployment challenges facing every major player in the sector.

What Decart Brings to the Table

Founded by veterans of elite Israeli military intelligence and academic research units, Decart has developed proprietary techniques that allow large language models to operate with substantially lower latency than competing approaches. The startup gained wider recognition after demonstrating a real-time interactive version of Minecraft powered by its AI inference engine, a showcase that circulated widely in AI research circles and signaled the company’s ability to handle computationally demanding, continuous-generation tasks in near-instant timeframes.

That capability is not merely a technical curiosity. For Anthropic, whose Claude models are increasingly deployed in enterprise settings where response speed and throughput directly affect customer satisfaction and pricing competitiveness, acquiring Decart’s inference optimization stack could translate into measurable improvements in service quality. Industry analysts estimate that inference costs can account for upward of 60 to 70 percent of total operational expenditure for AI model providers at scale, making efficiency gains in this area commercially valuable well beyond their research novelty.

Decart’s team, though relatively small by the standards of its potential acquirer, is understood to include some of the more sought-after researchers working on real-time generative systems. Anthropic, which has raised several billion dollars from investors including Google and Spark Capital and was most recently valued at approximately 18 billion dollars, has the financial resources to complete a transaction of meaningful scale without materially altering its capital position.

Strategic Implications for the Broader AI Race

The reported talks arrive at a moment when competition among frontier AI developers has intensified sharply. OpenAI, Google DeepMind, Meta, and Anthropic are all investing heavily in both model capability and the infrastructure required to serve those models efficiently to millions of simultaneous users. Inference optimization has emerged as a distinct technical discipline within that broader competition, with startups and internal teams alike racing to reduce the cost and latency of running large models in production environments.

Acquisitions of this type have become a preferred mechanism for established AI companies to absorb specialized expertise quickly rather than attempting to replicate it internally over a longer development cycle. The pattern echoes broader consolidation dynamics visible across the sector, where well-capitalized incumbents are absorbing early-stage teams working on specific technical problems — from chip design to data pipeline efficiency — that directly affect commercial competitiveness.

For Israel’s technology ecosystem, a successful acquisition of Decart by Anthropic would reinforce the country’s reputation as a source of foundational AI engineering talent. Deals of this nature also typically involve retention packages and continued research operations in the country of origin, meaning the economic and intellectual benefit need not be entirely transferred abroad. As reported by Globes business news, neither Anthropic nor Decart has publicly confirmed the talks, and the terms and timeline of any potential agreement remain undisclosed. Investors and observers in both the Israeli startup ecosystem and the broader AI industry will be watching closely for any formal announcement in the weeks ahead. The agents funding wave shows no signs of abating, and acquisitions like this one underscore how quickly the sector’s center of gravity can shift when the right technical capability meets the right acquirer.