Qualcomm and Amazon have struck a multigenerational agreement covering customized AI-inference silicon and high-speed optical networking for Amazon Web Services infrastructure, giving Qualcomm a significant opening into a data-center market it has been targeting beyond its traditional mobile-chip business.

The widely reported $4 billion figure is not the value of the supply agreement. Qualcomm’s SEC filing says Amazon received warrants to acquire as many as 25 million Qualcomm shares at $161.26 each—an exercise value of about $4.03 billion. Vesting is tied to commercial arrangements, binding purchase orders and actual purchases of Qualcomm server-chip products, technology, systems and manufacturing services, with the milestones extending to as much as $60 billion in payments over the warrant term.

The filing does not establish that Amazon has already committed to spend $60 billion. It says 3.75 million warrant shares vested at issuance based on initial purchase commitments, but does not disclose the dollar value of those commitments. The remaining warrants vest as additional commercial and purchasing milestones are met. Reuters reported that the arrangement runs over roughly a decade.

The technology agreement spans both computation and the network connecting it. In its announcement, Qualcomm said the companies will develop customized silicon for AI inference and optical-connectivity systems extending to 1.6 terabits per second, using Qualcomm SerDes and optical DSP technology. Qualcomm will also expand its use of AWS infrastructure and Amazon Bedrock for electronic-design-automation workloads.

The optical component reflects the widening infrastructure contest around AI data centers. Scaling inference increasingly requires not only accelerators but also high-bandwidth links among processors, memory and racks. Qualcomm strengthened that part of its portfolio through its approximately $2.4 billion acquisition of Alphawave Semi, announced in 2025 specifically as part of its data-center expansion.

For AWS, the arrangement broadens the supply chain behind its custom infrastructure rather than simply adding another GPU supplier. For Qualcomm, it puts a major hyperscaler behind its attempt to compete in AI inference, where power efficiency and system-level economics can be as consequential as raw processing performance. Nvidia remains the dominant incumbent, but cloud operators’ growing use of custom silicon and proprietary interconnect architectures gives alternative chip and networking suppliers a larger role.

Neither Qualcomm’s announcement nor its filing specifies the number of AWS data centers involved, deployment locations, electrical capacity or resulting power demand. Those figures will determine how much of the agreement ultimately translates into new physical infrastructure rather than replacement or diversification of existing supply.