Qualcomm has signed a deal to supply Amazon with custom AI chips, according to a report carried by Yahoo Finance. The agreement makes the San Diego-based fabless designer a supplier to one of the largest buyers of AI compute in the world and signals a fresh competitive push into data center silicon for a company long identified with smartphone modems and application processors.

The report did not disclose financial terms, chip specifications, process nodes, or shipment volumes, and neither Qualcomm nor Amazon has publicly detailed the scope of the engagement. What is confirmed at this stage is the existence of the commercial arrangement itself: Qualcomm will provide Amazon with custom AI chips under the newly signed deal.

The arrangement matters because it positions Qualcomm against a specific and well-defined slice of the AI silicon market. Amazon, through its Annapurna Labs unit, designs its own training and inference accelerators in-house and has built its AI infrastructure strategy around internally developed parts rather than merchant GPUs. A deal to bring in Qualcomm as a custom silicon supplier implies a second sourcing decision — one that gives Amazon an additional design partner for workloads its internal roadmap may not cover economically.

For Qualcomm, the win extends a custom silicon business the company has cultivated for several years. That business previously produced co-developed processors for customer devices, and Qualcomm has repeatedly told investors that data center AI represents a diversification target as smartphone chip sales mature. Landing a hyperscaler as large as Amazon is the strongest evidence to date that the strategy is converting into design wins.

The competitive read-through is straightforward. Nvidia dominates merchant AI accelerators in the data center, and hyperscalers including Amazon, Google, Microsoft and Meta have responded by designing their own chips to reduce dependence and cost. Qualcomm's entry as a custom design partner places it in the same competitive lane as Broadcom and Marvell, the two merchant designers most associated with hyperscaler custom silicon engagements. A win at Amazon suggests that lane now has a third credible contender with modem-scale RF, power management and SoC integration expertise behind it.

Pricing and margin dynamics were not part of the disclosed information. Custom silicon engagements of this type typically trade unit volume for negotiated design and supply economics, and without disclosed figures, the revenue impact on Qualcomm cannot be quantified. The deal's significance at this stage is strategic rather than immediately financial: it establishes Qualcomm as a data center AI supplier to a top-four hyperscaler.

Questions that will shape the commercial outcome remain open. Which process node the chips will target, whether production runs on TSMC's advanced nodes as Qualcomm's flagship products do, what performance class Amazon is sourcing, and over what timeline shipments begin were all absent from the announcement as reported.

Investors and supply chain watchers will now look for follow-on disclosure — volume commitments, revenue contribution guidance, or an Amazon statement on where the custom parts sit relative to its in-house accelerator families — before the deal can be sized against Qualcomm's existing handset-dominated revenue base.