TL;DR
Anthropic is reportedly discussing an acquisition of Israeli AI startup Decart for about $6 billion. Decart’s software could lower computing costs and expand Anthropic beyond language models, but neither company has confirmed a deal.
Anthropic is reportedly in talks to acquire Israeli artificial intelligence startup Decart for about $6 billion, a potential deal centered on software that improves chip efficiency and models that simulate physical environments. The proposed acquisition has not been announced by either company, and the negotiations may not produce an agreement.
Decart develops the Decart Optimization Stack, or DOS, which is designed to make AI training and inference workloads run more efficiently across computing hardware. People familiar with the discussions cited in reports said the technology could help reduce Anthropic’s computing costs and allow its existing infrastructure to serve more demand.
The startup also builds real-time world models, AI systems trained to generate or modify interactive visual environments. Its products include Lucy, which processes live video, and Oasis, which creates simulated environments intended for uses such as robotics and physical AI. Those capabilities would extend Anthropic beyond the language models behind Claude.
The reported price represents a large premium to Decart’s latest disclosed private-market value. Decart said in May 2026 that it raised $300 million in a round led by Radical Ventures. That financing valued the company at nearly $4 billion, according to The Wall Street Journal. Decart has raised more than $450 million from investors including Nvidia, Sequoia Capital, Benchmark, Zeev Ventures and Adobe Ventures.
Reported acquisition talks · August 13, 2026
Why Anthropic may pay $6 billion for Decart
The strategic prize is not simply another AI model. Decart’s optimization software could make every chip work harder, lower the cost of serving Claude, and open a route into interactive video, simulation, robotics, and physical AI.
01 · Strategic logic
What Anthropic could be buying
Decart combines infrastructure optimization with real-time world models. The first capability addresses a clear and immediate cost problem; the second could widen Anthropic’s product frontier beyond the language models behind Claude.
More output from each chip
The Decart Optimization Stack, or DOS, is designed to improve training and inference efficiency across AI hardware.
Serve more Claude demand
Greater hardware utilization could increase request capacity without requiring an equivalent expansion in chips and data centers.
Move beyond language
Lucy processes live video, while Oasis generates interactive environments for simulation, robotics, and physical AI applications.
02 · The compute flywheel
How efficiency can compound
Advanced AI systems consume costly accelerator time at both ends of the model lifecycle. Even a modest efficiency improvement can propagate through Anthropic’s infrastructure and unit economics.
Optimize execution
Software coordinates model workloads more efficiently across available processors.
Lower unit cost
Each training run or inference request consumes less effective compute capacity.
Unlock capacity
Existing infrastructure can potentially support more users, agents, and reasoning workloads.
Reinvest the gain
Saved capacity can support larger models, lower prices, or faster product expansion.
03 · Valuation trajectory
From $500M to a reported $6B
Founded in Israel in 2023, Decart moved through a rapid sequence of private-market valuations. The reported acquisition figure would represent another sharp step upward—and an unusually large purchase for Anthropic.
Decart’s reported valuation climb
Relative bar length indexed to the reported $6 billion acquisition discussion.
04 · Capability comparison
Immediate value versus long-range upside
The infrastructure case is easier to connect to Anthropic’s existing business. World models offer broader upside, but their commercial value and fit remain less established.
| Capability | Strategic fit | Potential benefit | Evidence status | Primary uncertainty |
|---|---|---|---|---|
| DOS optimization stack | ✓Direct | Lower training and inference costs | ~Company claims | Savings at Anthropic scale are undisclosed |
| Infrastructure researchers | ✓Direct | Specialized expertise across hardware and model execution | ✓Clear asset | Retention terms are unknown |
| Lucy live-video system | ~Adjacent | Real-time visual understanding and generation | ~Early market | Product integration and demand |
| Oasis world models | ~Optionality | Simulation, robotics, entertainment, and physical AI | ✗Unproven value | Commercial model remains uncertain |
05 · Confirmation test
The deal is still a hypothesis
Neither company has announced a signed agreement. The reported price, transaction structure, regulatory conditions, staff treatment, and technology performance remain unresolved.
Has Anthropic agreed to buy Decart?
No public agreement has been confirmed. The discussions could end without a transaction.
What does the $6 billion include?
It is unknown whether the figure is cash, shares, or includes employee-retention packages.
Does DOS deliver at hyperscale?
No public evidence quantifies Decart’s savings inside Anthropic’s production environment.
What could derail the talks?
Price, structure, staff retention, regulatory review, competing bids, or performance concerns.
The strategic chain
From software layer to competitive leverage
Bottom line
The $6 billion question is whether software can substitute for infrastructure.
If DOS materially increases the productivity of Anthropic’s massive compute base, the acquisition premium may be rational. If the gains do not hold at scale, Anthropic would be paying heavily for talent, optionality, and an unverified efficiency thesis.
Not confirmed
Compute Economics Drive the Price
For Anthropic, the strongest strategic case is control over computing efficiency. Training and operating advanced AI models requires large quantities of costly chips and data-center capacity. Software that extracts more performance from each processor could lower the cost per AI request, increase capacity without an equivalent hardware expansion and reduce pressure created by growing Claude usage.
A purchase would also give Anthropic specialized infrastructure talent and technology spanning chips, model execution and real-time visual systems. That could narrow its dependence on outside suppliers while placing it closer to parts of the computing stack associated with companies such as Nvidia. The world-model business offers a separate route into video, simulation and embodied AI, though its commercial value remains less established than the infrastructure case.
As an affiliate, we earn on qualifying purchases.
Decart’s Valuation Rose Rapidly
Decart was founded in Israel in 2023 and has grown rapidly through successive financing rounds. Its estimated valuation rose from about $500 million in late 2024 to $3.1 billion in 2025 and nearly $4 billion in May 2026. A $6 billion sale would mark another sharp increase within a short period.
Anthropic, the company behind Claude, has raised tens of billions of dollars while expanding its computing capacity through major cloud and infrastructure agreements. The reported Decart talks suggest it may also use acquisitions to obtain proprietary efficiency technology, rather than relying only on greater purchases of chips and data-center services. A transaction at the reported price would be an unusually large acquisition for the company.
“An ultra-optimized inference and training stack that enables agents and reasoning models to run smarter and faster.”
— Decart, describing DOS in its May 2026 funding announcement
real-time environment simulation software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Deal Terms Remain Unconfirmed
No final agreement has been announced, and it is not yet clear whether the discussions will continue, whether the reported $6 billion figure would be paid in cash or shares, or whether that amount includes employee-retention packages. Regulatory conditions and the planned treatment of Decart’s investors and staff are also unknown.
There is also no public evidence showing how much Decart’s optimization software could save Anthropic at scale. Decart’s performance descriptions are company claims rather than independently verified benchmarks. The relative weight Anthropic places on DOS, Decart’s researchers or its world models has not been disclosed.
Distributed AI Systems: A practical guide to building scalable training, inference, and serving systems for production AI
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Negotiations Face Confirmation Test
The next milestone is public confirmation from either company or a formal transaction announcement. Any agreement would be expected to disclose the purchase structure, approval requirements and whether Decart would continue operating independently or be integrated into Anthropic.
Until then, the reported $6 billion price remains provisional. A breakdown in talks, a revised valuation or interest from another bidder could change the outcome.
As an affiliate, we earn on qualifying purchases.
Key Questions
Has Anthropic agreed to buy Decart?
No. Anthropic is reported to be in discussions, but neither company has confirmed a signed agreement. The talks could end without a transaction.
Why might Decart be worth $6 billion to Anthropic?
Decart’s software is designed to increase chip efficiency, potentially lowering the cost of training and operating Claude. Its researchers and real-time world models could also expand Anthropic’s technical capabilities.
What are world models?
World models are AI systems intended to simulate environments and physical behavior. Decart uses them for interactive video and simulated settings that may support entertainment, robotics and other physical-AI applications.
How was Decart valued before the reported talks?
Its May 2026 funding round reportedly valued it at nearly $4 billion. The proposed acquisition price would place it about 50% above that valuation.
What could stop the acquisition?
The companies may fail to agree on price, structure or retention terms. Regulatory review, competing bids and questions about the technology’s performance at Anthropic’s scale could also affect the negotiations.
Source: Anthropic
Source: Anthropic
