EVERYTHING
Anthropic Negotiates a $6 Billion Acquisition of Decart AI
Artificial intelligence developer Anthropic has entered negotiations to acquire Decart AI, an Israeli startup specializing in hardware optimization and world models, for $6 billion. The report, initially disclosed by Bloomberg, details the largest prospective buyout in Anthropic's history as the firm prepares for an initial public offering in late 2026.
The negotiations highlight a critical inflection point for the AI industry, where the bottleneck has shifted from model size to operational efficiency. By bringing Decart’s technology in-house, Anthropic aims to alleviate server pressure caused by millions of daily active users relying on its Claude ecosystem.
The Compute Breakthrough Behind the $6 Billion Valuation
Decart emerged from stealth in October 2024 with $21 million in seed funding before rapidly raising a $32 million Series A and a $300 million Series B round. The startup's rapid growth stems from its specialized software architecture designed to extract maximum throughput from existing silicon graphics processing units.
Rather than relying solely on raw hardware upgrades, Decart developed a software stack that reorganizes how neural networks process matrix math. The company's core technological portfolio includes:
- DOS Stack: An inference and training engine that enables autonomous agents to execute complex reasoning tasks with significantly lower latency.
- Lucy Model: A generative world model tailored for commercial applications, allowing users to virtually try on apparel and interact with 3D product previews.
- Oasis Engine: An advanced physical simulator capable of rendering real-time interactive digital environments without traditional physics code.
Decart Chief Executive Officer Dean Leitersdorf explained the vision behind the technology during the company's recent funding announcement, stating: «Once systems can understand and operate within the physical world, the range of what becomes possible expands dramatically — from robotics and autonomous systems to entirely new forms of commerce and live experience.»
Slashing Inference Costs to Unlock Physical AI Agents
The ultimate strategic payoff of this acquisition extends far beyond lowering electricity bills for data centers. For everyday users, the heavy computational cost of LLMs has remained the single biggest barrier to continuous, real-time AI assistance, often limiting agents to text boxes and delayed responses.
Integrating Decart’s optimization stack will allow Anthropic to deploy real-time voice, vision, and spatial reasoning directly onto client devices and edge servers. By drastically reducing per-token computing expenses, our newsroom foresees a shift where AI agents transition from passive chat tools into persistent digital companions that actively monitor, navigate, and shape complex real-world workflows without ballooning enterprise overhead.