Anthropic is in advanced talks to acquire Decart, a startup that has built a name for itself by generating interactive, real-time simulations using AI, at a staggering valuation of $6 billion. The deal, reported by Calcalist, would rank among the largest AI acquisitions on record and signals just how aggressively the frontier-model race is pushing companies to buy capabilities they cannot build fast enough on their own. For Anthropic — already one of the best-funded AI labs in the world — this would be an emphatic statement about where it sees the next frontier of intelligence: not just language, but dynamic, interactive worlds. Decart fits squarely into the broader surge in Israeli startup funding that has defined the early part of this decade.
Decart’s most visible product to date is Oasis, a real-time AI-generated Minecraft-style world that the company released in late 2024. Oasis demonstrated that a sufficiently trained model could render a playable, interactive environment frame by frame without a traditional game engine — a proof of concept that caught the attention of the broader AI research community and apparently of Anthropic’s leadership team. The startup had previously raised funding at a far more modest valuation, making the leap to $6 billion a testament to how quickly the market has re-rated companies with credible world-model capabilities.
Why World Models Are the New Battleground
The strategic logic here is hard to miss. Every major AI lab is quietly racing toward what researchers call world models — systems that don’t just predict the next word in a sequence but can simulate plausible physical and interactive environments. OpenAI has Sora for video generation. Google DeepMind has a suite of simulation research. Anthropic, whose public identity has been built almost entirely around Claude and safety-focused language modeling, has been conspicuously absent from that conversation. Acquiring Decart would change that overnight.
World models matter beyond gaming demos. They are considered a potential pathway to more robust reasoning, planning, and agent behavior — capabilities that every frontier lab is chasing as the low-hanging fruit of pure language scaling starts to thin out. If an AI system can simulate consequences, it can plan better, act more reliably, and operate as a genuine agent rather than a sophisticated autocomplete engine. That is the capability gap Anthropic would be purchasing for $6 billion, and by that measure, the price tag starts to look less like a luxury and more like a strategic necessity. The competitive dynamics here echo the kind of large-scale venture bets increasingly defining the cybersecurity and AI sectors alike.
What the Deal Would Mean for Anthropic’s Roadmap
Anthropic has raised billions from Amazon and other backers and is widely understood to be burning cash at a rate commensurate with frontier model training. Adding a $6 billion acquisition to that equation is not trivial, and it raises immediate questions about structure: whether this is a cash deal, an all-stock arrangement, or some hybrid. Calcalist did not specify deal terms beyond the valuation figure, and neither Anthropic nor Decart has made a public statement confirming the talks.
Still, the implications for Anthropic’s product roadmap are significant regardless of how the financing shakes out. Decart’s team has demonstrated genuine technical chops in an area that complements rather than duplicates what Anthropic already has. Folding that talent and those models into an organization with Anthropic’s compute resources and research infrastructure could accelerate timelines that would otherwise take years to close. For users and enterprise customers who have come to rely on Claude for text-based reasoning, a Decart acquisition could eventually translate into agents that understand and navigate simulated environments — a capability that would open entirely new categories of automation and software interaction. The AI industry has been consolidating fast, and this deal, if it closes, would be one of its most consequential chapters yet.
