Qualcomm partners with Amazon to co-develop AI chips for AWS

Qualcomm just landed the kind of deal that transforms a company’s entire trajectory. The chipmaker announced a collaboration with Amazon to design custom silicon for AI data centers running on AWS, a partnership that could funnel up to $60 billion in commercial purchases to Qualcomm through September 2036.

That’s not a licensing agreement or a vague memorandum of understanding. It’s a production-ready engagement, with Qualcomm already manufacturing the new chips and expecting revenue to flow starting in its fiscal Q1 2027, which maps to the December 2026 quarter.

The deal structure

The partnership centers on AI inference chips, the silicon that runs trained models at scale rather than training them from scratch. Qualcomm’s new chips integrate high-speed optical connectivity capable of hitting 1.6 terabits per second, a specification that matters enormously when you’re shuttling data across warehouse-scale computing clusters.

Beyond chip sales, the deal has a software dimension. Qualcomm gets access to AWS AI infrastructure, including Amazon Bedrock, to improve its electronic design automation workflows.

The financial engineering is equally telling. Qualcomm issued warrants giving Amazon the right to purchase up to 25 million shares of QCOM stock at $161.26 per share. An initial tranche of 3.75 million shares has already vested. If Amazon exercises the full allotment, the warrants would be worth roughly $4 billion at current prices.

Qualcomm’s stock rose approximately 3-7% on the announcement day.

Why this matters for the AI chip landscape

This is Qualcomm’s first substantial engagement with a Western hyperscaler for data center silicon. Qualcomm laid the groundwork at its June 2026 Investor Day, where it introduced the Dragonfly C1000 CPU and outlined a target of $15 billion in data center sales by fiscal 2029. The company also acquired Alphawave Semi, a move that brought critical high-speed connectivity IP in-house.

AWS’s custom-chip revenue run-rate exceeded $25 billion as of June 2025, driven largely by its in-house Graviton CPUs and Trainium AI accelerators.

Competitive implications and what to watch

The $60 billion ceiling on the deal spans a full decade, so the near-term revenue impact will be modest relative to Qualcomm’s overall business. Amazon choosing Qualcomm as a co-development partner, rather than simply designing everything in-house or buying from established data center vendors, suggests that Qualcomm’s technology offers something the existing options don’t.

For Nvidia, this is another data point in a growing trend of hyperscalers diversifying their chip supply chains. Google has its TPUs, Amazon has Trainium and Graviton, and Microsoft has been developing custom AI accelerators.

Marvell Technology, which has built a significant business designing custom chips for cloud providers, may feel the competitive pressure most directly from Qualcomm’s entry into co-developed hyperscaler silicon.

As Amazon’s 25 million shares vest over time, Qualcomm’s share count will dilute modestly. But if the deal generates anything close to its $60 billion potential, that dilution will be a rounding error against the revenue it produces.