Amazon’s $7B Decart Bid Is Really About Beating Nvidia

Close-up macro photograph of a computer processor with golden pins on a dark background

Amazon’s Biggest Check Was Never for Hollywood

Amazon is reportedly in talks to buy AI startup Decart for around $7 billion, according to The Wall Street Journal, in a deal that would eclipse its $6.5 billion MGM purchase as the largest acquisition in the company’s history. The catch: this time the prize isn’t content — it’s the software layer that could loosen Nvidia’s stranglehold on AI computing.

Decart, founded in 2023, builds software that lets developers move AI workloads between different AI chips with minimal friction — the kind of translation layer that matters enormously as companies scramble for compute and try to avoid depending on a single GPU supplier. The startup is also working on “world models”: Lucy, which edits live video in real time, and Oasis, which generates simulated 3D environments for training robots and autonomous vehicles, Reuters reports.

The Journal says a deal could come together soon if talks don’t collapse and no rival bidder emerges. Neither Amazon nor Decart has commented.

The Trainium Connection

Here is the detail that reframes the whole story. According to the Journal’s reporting, Amazon Web Services gave Decart early access to Trainium3 — Amazon’s third-generation custom AI chip, built to compete with Nvidia — last year, after meeting the company and coming away impressed by its founders.

That matters because Amazon’s chip problem was never really about silicon. Trainium hardware exists; the hard part is getting developers to actually use it. AI models are built for specific chips — overwhelmingly Nvidia’s — and porting them elsewhere requires painful performance tuning. Decart’s software is designed to erase that burden. Buying it would hand AWS a ready-made portability layer purpose-built to bridge its proprietary silicon to customer workloads.

In other words, Amazon isn’t just buying an AI startup. It’s buying the on-ramp to its own chips.

Why this matters: the bidding war tells the story

Amazon is not the first giant to circle Decart. Anthropic walked away from a potential $6 billion acquisition in September after conducting due diligence, Reuters reports, citing Bloomberg News — it had explored the deal in August. Before that, Decart raised $300 million in May at a valuation near $4 billion, in a round led by Radical Ventures with Nvidia itself joining as a new investor. The Journal also noted that Israeli outlet Calcalist had reported the startup was in advanced talks with an unnamed major technology company.

Read that sequence: Nvidia invests, Anthropic does due diligence and passes at $6 billion, Amazon bids $7 billion. Everybody in the AI infrastructure stack has now priced the same insight — the scarcest asset in the compute war isn’t GPUs anymore, it’s the software that makes any GPU interchangeable.

There is a world-models subplot too. Lucy and Oasis put Decart in the same arena as the simulation platforms training the next generation of robots and self-driving systems — a second growth story stapled to the infrastructure one. If the deal closes, Amazon gets both: a wedge into Nvidia’s moat and a foothold in physical AI.

The usual caveats apply: talks can collapse, and Reuters noted it could not independently verify the Journal’s report. But the direction of travel is unmistakable — the AI giants have stopped fighting only over chips and started fighting over the software that decides which chips get used.

Meanwhile in AI Tech:

Also this week: TypeSafe’s $870M raise for Jev decision models — another bet that decision-making AI is where the money goes.

And: Tavus’s Griffin avatar fooled 48% of job interviewers — the video Turing test claim, and what the fine print actually says.

Written by
Ryan covers artificial intelligence and enterprise tech — from foundation models and AI chips to the business of machine intelligence. He tracks model releases, funding rounds, and the policy moves shaping the AI industry.