Key Notes
- Amazon is discussing a potential Decart acquisition valued at around $7 billion, according to the Wall Street Journal.
- Decart already works with AWS on Trainium and builds software for optimizing AI workloads across different chips.
- The reported price exceeds its $4 billion funding valuation, but final terms and completion remain unconfirmed.
Amazon.com Inc. (NASDAQ: AMZN) is in talks to acquire artificial intelligence startup Decart in a transaction that could value the company at around $7 billion, The Wall Street Journal reported, citing people familiar with the matter.
The potential purchase would bring an existing Amazon technology partner inside the company. Decart develops software for optimizing AI workloads across competing chips, alongside models that transform live video and generate interactive environments.
The discussions remain negotiations. Bloomberg, citing the Journal, said a deal could come together soon if the talks do not collapse. A reported valuation does not establish a final purchase price, payment structure or closing timetable.
An Existing AWS Relationship Could Deepen
Amazon’s connection with Decart predates the takeover report. In its May 18 funding announcement, Decart identified Amazon as a strategic customer and described a commercial agreement and joint go-to-market initiative with Amazon Web Services.
The startup said the collaboration included AWS’s Annapurna Labs team and deployment of its Lucy 2 model on Trainium3, Amazon’s custom AI chip. That gives the reported acquisition discussions a concrete technical backdrop rather than a relationship starting from scratch.
For Amazon shareholders, the potential benefit would be greater control over software that helps turn expensive computing capacity into useful output. That is a possible strategic rationale, not a forecast of savings or earnings from a transaction that has yet to be announced.
Why Chip Optimization Matters
Decart’s optimization platform, known as DOS, is designed to improve training and inference across different hardware architectures. Training builds a model’s capabilities; inference is the computing work required when that model serves users.
The company offers workload-specific optimization, custom kernel development, compiler tuning and tools for examining performance. These are less visible than a chatbot or video demonstration, but they can influence how much useful work a chip performs and the cost of delivering an AI service.
Decart says its software works across graphics processors, Google’s tensor processing units, Amazon Trainium and other accelerators. Its performance claims remain company-reported and depend on the model, hardware and benchmark being tested.
For a cloud provider, the economic question is whether those improvements can be repeated across customer workloads. Faster demonstrations alone do not establish the commercial return from buying the business.
Trainium Connects the Software to Amazon’s AI Strategy
AWS describes Trainium as a purpose-built AI processor integrated with its servers, networking, software and cloud services. Its stated objective is to improve the economics of both training models and running them at scale.
The AWS product page lists Decart among Trainium customers. Amazon’s own materials therefore corroborate the technology relationship, independently of the reported acquisition negotiations.
Owning an optimization specialist could help Amazon develop its hardware and software together. However, Decart’s ability to support rival processors is also part of its proposition. How an acquisition would affect that broad compatibility, external customers and future product development remains an important unanswered question.
A Higher Price After the Latest Funding Round
Radical Ventures confirms that it led Decart’s $300 million Series B in 2026 at a $4 billion valuation. It identifies the company’s founders as Dean Leitersdorf, Moshe Shalev and Orian Leitersdorf, with Dean serving as chief executive.
The roughly $7 billion figure now being reported is about 75% above that financing valuation. The comparison illustrates the scale of the potential step-up, although a private funding valuation and an acquisition price can reflect different rights, terms and financial structures.
Decart’s May announcement put total funding above $450 million and named Nvidia Corp. (NASDAQ: NVDA) among its new investors. Financial backing from a chipmaker is distinct from ownership of the company or an agreement to acquire it.
Lucy and Oasis Add a Model Business
Decart also builds applications on top of its infrastructure. Its current product overview describes Lucy as a system for transforming video in real time, with uses including virtual try-on, advertising, social platforms and gaming.
Oasis focuses on interactive environments for physical AI, including robotics, autonomous vehicles, manufacturing and drones. Those are intended applications; the product descriptions do not establish adoption levels or revenue for each market.
The reported Amazon discussions follow Anthropic’s decision to abandon a potential $6 billion Decart acquisition, as we previously reported. That earlier outcome is a reminder that serious interest and due diligence do not guarantee a signed transaction.
The next material development would be confirmation from the companies, followed by details on consideration, integration and any closing conditions. Until then, the investment question is whether combining Decart’s infrastructure and model research with AWS would justify the reported price.
