Anthropic Scales AI Compute
Anthropic Nears $6 Billion Acquisition of AI Startup Decart, Bolstering Compute Ahead of IPO
Key Points
Decart's Decart Optimization Stack (DOS) is designed to optimize AI training and inference across Nvidia GPUs, Amazon Trainium, and Google TPUs, enabling more work to run on existing chips and lowering costs.
Decart's video-focused technology includes the Lucy model for real-time live video editing and Oasis (with Oasis 3) to generate simulated environments for robotics and autonomous driving training.
Decart has recently raised about $300 million in a May funding round led by Radical Ventures, with Nvidia joining as a new investor, highlighting strong investor interest ahead of the deal.
Anthropic has been scaling compute capacity through major investments, including a $30 billion purchase of Microsoft Azure compute capacity and a $50 billion data-center investment with Fluidstack, along with expanding its Amazon partnership to secure up to 5 gigawatts of capacity.
Anthropic is in talks to acquire AI startup Decart for roughly $6 billion, according to Bloomberg and Yahoo Finance. The deal would be the Claude maker's largest known acquisition and comes as Anthropic races to expand computing power ahead of a potential IPO.
Negotiations are still early and could fall apart, sources say. But the price tag signals how urgently AI companies are scrambling to squeeze more performance out of existing chips — and how much the right technology to do that is now worth.
What Decart Does and Why Anthropic Wants It
Decart built a system called the Decart Optimization Stack, or DOS. It makes AI chips work more efficiently — so the same hardware can handle more tasks at lower cost. DOS works across Nvidia GPUs, Amazon Trainium chips, and Google TPUs. For Anthropic, which is spending billions on compute, that kind of efficiency gain is extremely valuable.
Decart also has real-time video technology. Its Lucy model edits live video using AI. Its Oasis system — now in its third version — creates simulated environments for training robots and self-driving cars. Bloomberg reports the deal is partly about adding these multimodal and video capabilities to Anthropic's lineup.
Decart Just Raised $300 Million Before Acquisition Talks
Decart closed a funding round of about $300 million in May, led by Radical Ventures. Nvidia joined as a new investor in that round — a notable sign of confidence from the world's dominant chip maker. The fresh funding makes the reported $6 billion acquisition price a steep premium over where the startup was valued just weeks ago.
That gap between the funding round valuation and the $6 billion offer reflects the intense demand for compute optimization tools right now. AI companies are hitting real limits on how much hardware they can buy or access. Technology that stretches existing chips further has suddenly become worth far more.
Anthropic Is Already Spending Big on Compute Infrastructure
This deal fits a broader spending spree. Anthropic has committed to a $30 billion purchase of Microsoft Azure computing capacity. It also struck a $50 billion data-center deal with Fluidstack. And it expanded its Amazon partnership to secure up to 5 gigawatts of capacity. That is an enormous amount of computing power — enough to run major AI services at scale.
Acquiring Decart would let Anthropic do more with that hardware. Instead of only buying more chips, it could run existing chips harder and cheaper. The deal also targets Decart's engineering talent, specifically for Anthropic's inference and performance team, according to Yahoo Finance.
A Strategic Move as Anthropic Eyes a Public Listing
Anthropic has not set a date for an IPO, but the company has been building toward one. Acquisitions like this one signal it wants to own more of its core technology stack before going public. Having in-house tools to lower AI costs could be a strong selling point to investors.
The Decart talks show how the AI industry is changing. Building a great model is no longer enough. Companies now compete on who can run those models faster, cheaper, and at greater scale. At $6 billion for a startup focused on chip efficiency and video AI, Anthropic is betting that infrastructure is just as important as the model itself.
Sentiment
Mostly PositivePublishers
10
Articles
146
Reach
156
