Amazon in Talks to Buy Decart for $7 Billion as AWS Expands AI Chip Push

Key Takeaways

• Amazon is reportedly discussing an acquisition of Decart valued at around $7 billion, with no definitive agreement announced.
• Decart’s optimisation software and existing Trainium collaboration could help AWS attract more AI workloads to its proprietary chips.
• Anthropic previously explored a purchase before withdrawing, while valuation, integration and commercial returns remain key uncertainties.

Amazon Targets Software Behind More Efficient AI Computing

Amazon (NASDAQ: AMZN) is reportedly negotiating to acquire AI startup Decart in a transaction that could value the business at approximately $7 billion. The discussions bring renewed attention to the software needed to make alternative AI chips commercially competitive.

The potential acquisition emerged in reporting published on October 11. As of October 12, the transaction remains unconfirmed, with the price and completion timetable subject to negotiations.

For Amazon, the strategic opportunity extends beyond acquiring another AI model developer. Decart builds software designed to improve how AI workloads run across different computing platforms. That capability could support Amazon Web Services as it expands adoption of its Trainium chips.

The central question is whether ownership would help AWS turn technical performance improvements into stronger customer demand and profitable cloud revenue.

Why Decart Could Strengthen AWS Trainium

Amazon’s custom chips offer AWS another way to supply the computing power required to train AI models and operate them after deployment. However, hardware performance alone does not determine which platform developers choose.

Customers also need software that allows their models to run reliably and efficiently. Moving workloads between platforms can require engineering work, testing and adjustments to existing tools. Those costs can influence whether a customer adopts a different chip.

Decart’s optimisation technology addresses this part of the computing stack. The company describes its Decart Optimization Stack, or DOS, as supporting Nvidia hardware, AWS Trainium and Google TPUs.

The potential benefit for AWS is therefore practical: software that improves performance or simplifies deployment could make Trainium more attractive to developers evaluating computing costs.

This is an interpretation of the strategic fit, rather than a financial outcome confirmed by Amazon. Any benefit would depend on how widely the technology can be deployed and whether customers achieve meaningful savings on their own workloads.

Existing Collaboration Provides a Starting Point

Amazon and Decart already have a commercial and technical relationship.

In its December 2025 Trainium3 launch announcement, AWS identified Decart as a customer achieving faster inference for real-time generative video. AWS said Decart was delivering four times faster inference at half the cost of GPUs.

Inference is the process of running a trained model to produce an output. For live video applications, the speed and cost of that process are particularly relevant because generated content must keep pace with user interaction.

Those figures are company-reported results for Decart’s application. They should not be treated as a universal comparison covering every AI model, GPU configuration or cloud service.

Nevertheless, the collaboration offers a clearer basis for the acquisition rationale than a general ambition to expand in AI. Amazon has already worked with Decart’s technology, providing a foundation for evaluating how it might fit within AWS.

Decart Adds Live Video and World-Model Capabilities

Decart’s business combines computing optimisation with AI models for interactive visual applications.

Lucy focuses on transforming video in real time. Decart presents potential uses across advertising, virtual try-on, social platforms and gaming, where content can change as users interact with it.

Oasis is positioned as an interactive world model for physical AI. Its intended applications include simulated environments for robotics, autonomous vehicles and other systems that operate in the physical world.

These product lines have different functions. Lucy centres on editing visual experiences, while Oasis generates environments that respond to actions.

For AWS, the possible attraction is a combination of infrastructure software and applications that use it. Decart’s models could demonstrate what optimised computing enables, while its software could support additional customer workloads.

Commercial success would still require dependable performance, customer adoption and sustainable operating costs. Technical demonstrations alone do not establish the revenue potential of either product.

Anthropic’s Earlier Talks Highlight Deal Uncertainty

Anthropic reportedly explored acquiring Decart for approximately $6 billion before withdrawing in September after conducting due diligence.

The earlier discussions show that Decart has attracted interest beyond Amazon. They also demonstrate that acquisition negotiations can end without an agreement, even after a prospective buyer has examined the business.

The available reporting does not establish a definitive reason for Anthropic’s decision. Its withdrawal should therefore not be treated as proof of a technical problem or a conclusion about Decart’s value.

Decart separately announced a $300 million funding round in May, led by Radical Ventures. The company identified Nvidia among its new investors, alongside returning backers including Sequoia Capital and Benchmark.

Nvidia’s participation adds nuance to the competitive picture. Decart works across several hardware platforms, so its role extends beyond supporting Amazon’s alternative chips.

What Amazon Investors Should Watch Next

For Amazon shareholders, the immediate catalyst would be confirmation of a signed agreement and disclosure of its terms.

The reported valuation does not reveal how a purchase would be financed, whether consideration would include stock, or what commitments might be needed to retain key employees. Those details would help investors assess the transaction’s cost.

The longer-term question is whether Decart would increase Trainium adoption, improve customer economics or create additional demand for AWS services.

Integration would also matter. Decart’s support for several computing platforms raises questions about how Amazon might manage its existing partnerships and product strategy after an acquisition. No confirmed plan has been disclosed.

The reported talks point to software becoming a larger part of competition in AI infrastructure. A completed purchase could strengthen Amazon’s tools for attracting workloads to AWS, but its value would ultimately depend on customer adoption and financial returns.