Qualcomm announced on September 8 that it is collaborating with Amazon across multiple generations of customized silicon for large-scale AI data centers. The companies are focusing on AI inference and on optical connectivity intended to carry growing volumes of data within Amazon's infrastructure.
Compute and optical links move together
The product announcement says Qualcomm and Amazon will work on optical solutions extending to 1.6 terabits per second and later generations, using Qualcomm's SerDes and optical digital-signal-processing technology. Qualcomm also plans to expand its use of AWS infrastructure, including Amazon Bedrock, for electronic-design-automation workloads aimed at shortening chip-development cycles.
The strategic shift is broader than a single accelerator. AI inference at hyperscale depends on server compute, memory movement, networking, and power efficiency. Qualcomm is positioning its custom-silicon and connectivity portfolio as an alternative supply path in a market where NVIDIA remains the dominant accelerator vendor.
The filing reveals the commercial ceiling
Qualcomm's Form 8-K says it issued Amazon.com NV Investment Holdings a warrant for as many as 25 million Qualcomm shares at an exercise price of $161.26. The warrant expires in September 2036. Vesting is tied to commercial arrangements, binding orders, and actual purchases of Qualcomm server-chip products, technology, systems, and manufacturing services, up to $60 billion in payments.
The filing says 3.75 million shares vested when the warrant was issued based on initial purchase commitments. Reuters independently corroborated the collaboration, warrant terms, customized inference focus, and 1.6-terabit optical work.
Large ceiling, limited deployment detail
The $60 billion figure is a maximum purchase threshold connected to warrant vesting, not booked revenue or a guarantee that Amazon will spend that amount. Neither the release nor the filing names the first production silicon, provides a deployment schedule, or supplies independently tested performance-per-watt results. Those execution details remain open.