Amazon in Talks to Buy AI Startup Decart for About $7 Billion

Amazon is in talks to acquire Decart, a fast-growing AI startup, in a deal valued at around $7 billion, according to a Wall Street Journal report on October 11. A deal "could happen soon if talks do not fall apart or another suitor emerges," the Journal reported, citing people familiar with the matter.

If it closes, it would be one of Amazon's largest AI acquisitions. More importantly, it would hand Amazon a piece of software that helps AI models run fast on many kinds of chips, including rivals' hardware.

Who is Decart?

Decart is a young company, founded in 2023, that has moved remarkably fast. In May, it raised $300 million at a valuation of nearly $4 billion in a round led by Radical Ventures, with Nvidia, Adobe Ventures, and Toyota Ventures participating. OpenAI co-founder Andrej Karpathy also invested as an angel.

A $7 billion price would mean Decart's value nearly doubled in about five months.

Two businesses in one

Decart sells two very different things.

The Decart Optimization Stack (DOS) is software that speeds up AI training and inference (running a trained model to answer requests). It works across Nvidia GPUs, Google TPUs, and Amazon's own Trainium chips. Decart says it earns "significant revenue" licensing DOS to cloud providers and AI labs, and that DOS 2.0 lets AI agents process more than 1,600 tokens per second.

World models are AI systems that generate and simulate visual environments in real time. Decart's Lucy can edit live video on the fly, swapping objects or outfits in a stream, which makes it useful for things like virtual try-ons. Its Oasis model generates interactive 3D environments, such as a simulated warehouse for training robots.

Why this matters: Amazon cares about both. It runs one of the world's biggest clouds, and it operates huge, robot-heavy logistics networks.

The chip angle

The most strategic piece is likely DOS. Amazon has invested heavily in Trainium, its custom AI chip, to reduce its dependence on Nvidia.

But custom chips only pay off if customers can easily move their models onto them. Software that makes switching between chips painless is exactly what Trainium needs. RuntimeWire framed the open question well: would owning Decart strengthen Trainium, or undercut AWS's pitch as a neutral, hardware-agnostic cloud?

That tension matters to Decart's existing customers. A neutral optimization layer becomes less neutral once one chip vendor owns it.

Amazon wasn't the first to look

Bloomberg previously reported that Anthropic explored acquiring Decart for about $6 billion in August, then walked away after due diligence (the deep review buyers do before committing). Amazon is one of Anthropic's largest backers, so it's notable that Amazon is pursuing a deal Anthropic passed on.

Nvidia, already an investor, has also shown interest in the company, according to reports.

Who this affects

AWS customers could eventually see faster, cheaper inference on Trainium if DOS gets folded into Amazon's cloud tools.

Nvidia faces another push by a hyperscaler to make its chips easier to replace. This comes in the same week Microsoft detailed its Maia 300 chip.

AI startups get another data point: specialized infrastructure software is commanding premium prices as compute stays scarce.

What insiders are saying

So far, the companies themselves aren't talking. Amazon declined to comment outside business hours, and Decart didn't respond to requests for comment, according to Reuters, which said it could not independently verify the report.

Industry coverage has focused on the compute crunch. When chips are scarce and expensive, software that squeezes more work out of each one becomes extremely valuable.

What's next

Nothing is signed. Watch for an official announcement, a competing bid, or reports that talks have stalled. Any deal of this size would also likely draw regulatory scrutiny.

If the deal closes, the key question is whether DOS stays cross-platform or tilts toward Trainium. Decart's customers will want clear answers on that quickly.

The bottom line

Amazon's reported interest in Decart shows where the AI arms race is heading: beyond models and chips into the software layer that connects them. If you build on AWS, this could eventually mean better performance on Amazon's own hardware. If you rely on DOS today, keep a close eye on how its neutrality holds up under new ownership.