Anthropic PBC is reportedly in discussions to acquire Nvidia-backed artificial intelligence startup Decart AI for approximately $6 billion, Bloomberg reported, citing people familiar with the matter.
The potential transaction has not been finalised and negotiations could still collapse, according to a Reuters report.
The talks come as Anthropic prepares for a potential mega initial public offering and seeks to rapidly expand its computing capabilities. The AI company is looking to address capacity constraints as demand for its services continues to grow.
Decart develops AI infrastructure and optimisation technology designed to help developers improve the performance of AI applications across a broad range of chips. The startup also has a strong focus on generative video and has developed so-called world models that can make real-time changes to live video streams.
The acquisition could give Anthropic additional technology and infrastructure to handle increasing demand for its AI services, Reuters reported.
Among Decart's products is Lucy, an AI model capable of editing live video in real time, according to the company's website. The startup has also developed Oasis, a model designed to generate simulated environments for training and evaluating robotics and autonomous driving systems.
Decart announced in May that it had raised $300 million in a funding round led by Radical Ventures. NVIDIA participated in the financing as a new investor, strengthening the startup's ties to the semiconductor giant.
The reported acquisition discussions also come shortly after Anthropic signalled plans to deepen its work on AI hardware.
Last week, Anthropic said it was recruiting engineers with expertise spanning the hardware and software stack. The company said the hires would help it co-design custom chips and AI models to make its Claude artificial intelligence system faster and more efficient.
A deal for Decart could therefore provide Anthropic with additional expertise in AI infrastructure and optimisation as it scales its computing resources and meets growing demand for Claude.
Anthropic said Wednesday that it will introduce an invisible watermarking system for text processed by its Claude artificial intelligence chatbot as the company moves to comply with European Union regulations.
However, the company cautioned that the watermark should not be interpreted as proof that Claude generated the content. Instead, the marker indicates only that the text was processed by Claude at some point.
The feature, unveiled Monday in San Francisco, is designed to remain invisible to users and persist even when text is copied and pasted into another document or platform.
Anthropic said the technology nevertheless has limitations. The presence of a Claude watermark indicates that the content may have been processed by the chatbot, but it does not establish who or what originally created the text, the company said.
The company also warned that the absence of a watermark does not necessarily mean Claude was not involved. Very short passages or content that has undergone extensive editing may not retain the marker, Anthropic said.
The watermarking technology is part of Anthropic's broader effort to provide greater transparency around AI-generated and AI-processed content while meeting regulatory requirements in Europe.
For supported file formats, Anthropic said Claude will use a different approach. When the chatbot generates a file such as a JPG image, it will include digitally signed provenance metadata.
That information will follow an open standard designed to document the origin and history of digital content, potentially making it easier for users and systems to determine how a file was created or modified.
Catch all the Business News, Market News, Breaking News Events and Latest News Updates on Live Mint. Download The Mint News App to get Daily Market Updates.
Oops! Looks like you have exceeded the limit to bookmark the image. Remove some to bookmark this image.
Facts Only
* Anthropic PBC is reportedly discussing the acquisition of Decart AI for approximately $6 billion.
* The potential transaction has not been finalized and negotiations could collapse.
* Decart develops AI infrastructure and optimization technology for improving AI application performance across chips.
* Decart focuses on generative video and has developed world models for real-time changes to live video streams.
* Decart's products include Lucy, an AI model for real-time video editing, and Oasis, a model for generating simulated environments.
* Decart raised $300 million in a funding round led by Radical Ventures; NVIDIA participated as an investor.
* Anthropic is recruiting engineers to co-design custom chips and AI models for Claude efficiency.
* Anthropic plans to introduce an invisible watermarking system for Claude text to comply with EU regulations.
* The watermark indicates processing by Claude but does not prove authorship.
* Claude will use digitally signed provenance metadata for supported file formats like JPG images.
Executive Summary
Full Take
The situation reveals a dynamic interplay between strategic corporate consolidation and the technical demands of scaling frontier AI systems. Anthropic's pursuit of Decart AI reflects an acknowledgment that performance gains in large language models are increasingly bottlenecked by underlying infrastructure optimization. This move suggests a shift from solely focusing on model architecture to owning or tightly integrating the physical and software layers necessary for operational scale. The context of Anthropic simultaneously deepening work on custom hardware and implementing content provenance features demonstrates a strategy aimed at achieving both competitive advantage and regulatory compliance concurrently.
The development of Decart's technology, particularly real-time video world models, positions it at the cutting edge where abstract AI theory meets tangible, high-throughput computational demands. When Anthropic seeks this acquisition, the pattern suggests a desire to internalize specialized optimization expertise rather than relying solely on external partnerships for capacity expansion. The watermarking initiative, while framed as transparency, introduces an underlying tension between visibility and proprietary control—the desire to signal processing history versus ensuring user trust regarding content ownership. The potential transaction forces consideration of what constitutes essential infrastructure: is it owned by the model developer (Anthropic) or specialized third-party infrastructure providers (Decart)?
Bridge Questions: If Anthropic acquires Decart, how might this reshape the development and open-sourcing strategy for future foundational models? What are the long-term implications of embedding provenance metadata across all multimodal outputs versus relying on user perception alone to build trust? How does the pursuit of hardware control affect the broader distribution and accessibility of advanced AI capabilities?
Sentinel — Human
The content reads like standard business journalism synthesizing reported facts and company statements regarding an industry's competitive landscape and regulatory response to AI development.