Amazon in talks to inject over $10B into OpenAI

The artificial boundaries separating AI alliances in Silicon Valley are officially falling apart.

OpenAI is in talks with Amazon over a massive investment that could top $10 billion, CNBC confirmed on Tuesday following an initial report from The Information. While details remain fluid and subject to change, the deal under discussion includes a crucial twist: OpenAI would agree to use Amazon's custom artificial intelligence chips.

The talks became possible after OpenAI completed a corporate restructuring in October that redefined its long-standing partnership with Microsoft. While Microsoft has backed OpenAI with more than $13 billion since 2019, it surrendered its right of first refusal to be OpenAI's compute provider. With those handcuffs removed, OpenAI gained the freedom to raise outside capital and develop products with third parties.

The great AI cross-pollination

The landscape is turning into an all-out corporate entanglement. Amazon has already poured at least $8 billion into OpenAI's fierce rival, Anthropic. Yet the e-commerce giant is clearly eager to broaden its bets across frontier generative models.

Microsoft did something similar just last month, announcing plans to inject up to $5 billion into Anthropic alongside Nvidia's $10 billion commitment. Today, the tech titans are increasingly backing both sides of the race, making the concept of exclusive AI alliances look ancient.

OpenAI's appetite for compute explains why. The company has made over $1.4 trillion in infrastructure commitments in recent months, striking deals with Nvidia, AMD, and Broadcom. Just last month, OpenAI signed its first-ever contract with AWS, buying $38 billion in capacity. It also completed a $6.6 billion secondary share sale in October that pegged its valuation at $500 billion.

Amazon chips finally get their moment

Beyond the cash, the centerpiece of this negotiation is hardware. AWS has been quietly building internal AI processors since around 2015, launching its Inferentia inference line in 2018 and rolling out the latest generation of its Trainium training silicon earlier this month.

Until now, Nvidia has maintained an iron grip on frontier model training, but supply shortages and sky-high pricing have pushed labs to seek alternatives. Winning OpenAI as a flagship customer for Trainium would instantly legitimize Amazon's chip pipeline, proving it can handle the most demanding workloads alongside Claude.

Why it matters

For developers and everyday users, this shift could ultimately lower inference costs and ease reliability bottlenecks. Spreading workloads across Microsoft Azure, AWS, and custom silicon makes OpenAI far less vulnerable to single-cloud outages, though end users should not expect an immediate overhaul overnight.

Across the industry, this is a clear warning sign for Nvidia. OpenAI is aggressively diversifying its supply chain across AMD, custom Broadcom designs, and now AWS silicon. At the same time, the old playbook of choosing a single cloud partner is dead; OpenAI is playing the hyperscalers off one another to secure capital and subsidized compute.

We think this maneuvering is extraordinarily clever from a bargaining perspective, but it is not without risk. Piling on more than $1.4 trillion in long-term compute obligations creates immense financial pressure. If commercial revenue cannot keep pace with these staggering capital commitments, OpenAI could find itself locked into infrastructure it struggles to monetize.