Anthropic has ended its push to acquire Israeli AI infrastructure startup Decart AI after completing due diligence, walking away from a deal that had been valued at around $6 billion, according to multiple reports. The decision marks a significant shift in one of the most closely watched potential transactions in the artificial intelligence sector this year, even as both companies leave open the possibility of future collaboration.
What happened: Deal explored, then abandoned after due diligence
Anthropic, the maker of the Claude chatbot, had been in talks to acquire Decart AI in a transaction that could have valued the startup at about $6 billion, Bloomberg reported, citing people familiar with the matter. The San Francisco–based AI lab conducted due diligence on Decart but ultimately chose not to proceed with the acquisition, the reports said.
Neither Anthropic nor Decart immediately responded to requests for comment outside business hours, and Reuters said it could not independently verify the report at the time of publication. Despite the collapse of acquisition talks, the two companies could still explore other forms of partnership, according to the people briefed on the discussions.
Why Decart mattered: Chip-efficiency software at the heart of AI economics
Decart’s appeal to Anthropic centred on its core product: software that makes AI chips run more efficiently, thereby reducing the cost of training and operating large models. The startup’s optimization stack is designed to help developers “squeeze every drop of performance from every chip” during both training and inference, the phase when trained models generate responses.
For an AI company scaling products that require massive computing capacity, such efficiency gains directly affect unit economics and margins. Market commentary has suggested Decart’s technology can raise hardware utilization from an industry average of 40–50% to above 80%, effectively delivering a computational boost comparable to buying an additional $10–15 billion worth of hardware without new capex. That kind of leverage is especially valuable as AI labs race to control infrastructure costs ahead of public listings and further expansion.
Decart’s profile: Israeli-founded, Nvidia-backed, with “world models”
Founded in 2023 by three Israeli engineers—brothers Dean and Orian Leitersdorf and Moshe Shalev—Decart has quickly become a prominent name in AI infrastructure. The company is backed by Nvidia and has raised about $450 million to date, with its most recent round valuing it at nearly $4 billion, up from $3.1 billion in August 2025, according to press reports.
Beyond chip optimization, Decart has built its own AI models and platforms. These include:
Lucy, a live-video editing model that can take a real-time video feed of a person and generate a high-resolution overlay that makes it appear as if they are trying on clothing or accessories, a capability useful for fashion e‑commerce.
Oasis, a simulation or “world model” platform that creates three‑dimensional, physics-aware environments for applications such as robotics, logistics, and autonomous systems.
Decart says it trains its AI on text and millions of hours of video to learn real‑world physical properties, enabling more accurate simulations and real-time generative video. Its technology is already being used by influencers and platforms for live streaming, and by advertising and e‑commerce companies, including eBay, which is both an investor and a customer.
Anthropic’s context: IPO plans and a widening acquisition strategy
The talks with Decart unfolded as Anthropic prepared for a potential initial public offering. Earlier reports indicated the company was exploring acquisitions that could help it handle surging demand for its Claude models ahead of a planned listing, with marketing for an IPO expected as early as mid‑October and a listing targeted before the U.S. midterm elections in November.
Anthropic’s valuation has soared in recent funding rounds, with reports in mid‑2026 placing it at around $965 billion, overtaking OpenAI to become the world’s most valuable AI startup. In that context, a $6 billion acquisition would have been strategically meaningful but still relatively small compared with Anthropic’s overall market value.
The decision to walk away after due diligence suggests that, despite the strategic fit on paper, Anthropic identified issues—whether related to technology integration, financial terms, regulatory considerations, or operational risks—that made the deal unattractive at the proposed price. The company has not publicly detailed its reasons, and sources cited by Bloomberg and other outlets spoke on condition of anonymity.
Market implications: A signal for AI M&A and infrastructure valuations
The collapse of the Anthropic–Decart talks is likely to be read as a data point by investors and founders across the AI stack. On one hand, it underscores how intensely AI labs are hunting for infrastructure advantages, especially in chip efficiency and inference cost reduction. On the other, it highlights that sky‑high valuations for promising startups do not guarantee deals will clear the due‑diligence hurdle, even when the strategic rationale appears strong.
For Decart, continuing as an independent company means it can keep selling its optimization tools and world models to multiple AI labs and enterprises rather than becoming part of a single vertically integrated stack. That could be advantageous in a market where many players are wary of ceding critical infrastructure to a dominant competitor. At the same time, the episode may prompt questions about how high valuations in AI infrastructure will translate into exit outcomes if more acquirers adopt a similarly cautious stance post–due diligence.
For Anthropic, stepping back from the deal may reflect a disciplined approach ahead of an IPO, where investors will scrutinize both growth and capital efficiency. Passing on a $6 billion transaction after diligence could be framed as prudent risk management, especially if internal models suggested integration challenges or uncertain returns relative to the price.
What could come next: Collaboration, competition, and the road to IPO
Even without an acquisition, Anthropic and Decart may still pursue commercial or technical collaborations, such as licensing Decart’s optimization stack for specific workloads or co‑developing features for Claude’s inference pipeline. Such arrangements would allow Anthropic to capture some efficiency benefits without taking on the full balance‑sheet and integration burden of a $6 billion deal.
Meanwhile, Decart’s roadmap—spanning chip optimization, live video generation, and physics‑based simulation—positions it at the intersection of several high‑growth AI segments. As more enterprises adopt generative video, robotics, and autonomous systems, demand for tools that lower compute costs and improve real‑time performance is likely to rise, keeping Decart in contention for future partnerships or transactions.
For now, the key takeaway is clear: despite intense hype around AI M&A, large-ticket deals will still be subject to hard‑nosed diligence and strategic calculus, especially as leading players like Anthropic gear up for public markets.
