Amazon is weighing a roughly $7 billion purchase of the artificial intelligence startup Decart, according to a report from PYMNTS — a figure that, if completed, would place the transaction among the e-commerce and cloud giant's largest bets on generative AI.

The report describes the deal as under consideration rather than agreed. Neither Amazon nor Decart has confirmed the talks, and no terms beyond the reported $7 billion valuation have surfaced. For a company of Amazon's scale, the number is modest against its capital expenditure on data centers, but it is significant as a signal of where the company believes AI value will concentrate.

Why would Amazon buy rather than build?

Amazon already develops its own foundation models through its Bedrock platform and its Anthropic partnership, so an acquisition of Decart would raise an obvious question: what does a startup bring that internal teams cannot?

Acquisitions of AI startups typically deliver three things — model architectures and training know-how, a concentrated team of researchers, and shipping products with early user traction. For a buyer, purchasing a team outright can be faster than competing for the same talent in the open market, where senior AI researchers command compensation packages that rival senior executives.

A $7 billion price tag for a young company also reflects the current economics of the model layer. Investors have pushed valuations of AI startups sharply higher on the assumption that a small number of model developers will capture durable margins, even as the cost of training frontier models climbs into the hundreds of millions of dollars per run.

What does the reported interest signal about the AI market?

The reported deliberation lands during a period of intense consolidation pressure across the AI supply chain. Large cloud providers — Amazon, Microsoft and Google's parent Alphabet — have pursued a mix of minority stakes, deep partnerships and outright acquisitions to secure model capability and the workloads that come with it.

For startups, the calculus is equally stark. Training competitive models requires enormous compute budgets, and access to that compute increasingly runs through the hyperscalers. Selling to one of them offers capital and infrastructure at the cost of independence.

The $7 billion figure, if accurate, also illustrates how quickly the market reprices AI capability. Companies founded only in the current AI cycle are now commanding sums that, a few years ago, would have been reserved for mature businesses with established revenue.

What happens next?

Deliberation-stage reports frequently precede one of three outcomes: a signed deal at or near the reported figure, a walked-away negotiation, or a competitive bid from another buyer that drives the price up. Regulatory scrutiny of large technology acquisitions adds a further variable, as antitrust authorities in the United States and Europe have paid close attention to hyperscaler purchases of AI firms.

Nothing is confirmed until the parties announce it. But if Amazon proceeds, the acquisition would mark another step in the consolidation of foundation-model development under the major cloud platforms — and would narrow the field of independent AI startups that hyperscalers have left to acquire.