Qualcomm issued a warrant to an Amazon affiliate on Tuesday allowing the purchase of 25 million shares for $4 billion, locking the cloud giant into a framework that could see it spend up to $60 billion on Qualcomm silicon over the next decade. The warrant expires Sept. 3, 2036, and is tied to purchase orders covering multiple generations of custom data-center chips, including AI inference processors and optical connectivity running at 1.6 terabits per second.
The stake is the signal
The $4 billion equity commitment functions less as an investment than as a down payment on supply. Amazon gets the right to buy shares at a fixed price while Qualcomm gets a guaranteed demand pipeline that stretches past the current AI capital-expenditure cycle. The structure mirrors the stakes chipmakers have been offering hyperscalers to secure long-term offtake, but the $60 billion ceiling, if fully drawn, would represent a material share of Qualcomm’s projected revenue base.
Data center ambition gets a second anchor
Qualcomm’s June reveal of the Dragonfly C1000 server CPU landed Meta as a launch customer. Amazon now becomes the second hyperscaler to commit to Qualcomm’s data-center roadmap, which management has said should generate more than $15 billion in annual revenue by fiscal 2029. At the company’s investor day, Chief Financial Officer Akash Palkhiwala framed the target as the final leg of a diversification strategy spanning handsets, automotive, industrial IoT and now the server rack.
Optical speed as the differentiator
The 1.6 terabit-per-second optical digital signal processor technology is the spec that separates this deal from a standard CPU supply agreement. AWS plans to pair Qualcomm’s connectivity silicon with its Bedrock cloud service, betting that interconnect bandwidth will be the bottleneck in the next wave of AI cluster scaling. Whether the custom silicon delivers a measurable advantage over merchant-market alternatives remains unproven; the warrant’s value to Amazon ultimately depends on Qualcomm hitting performance and volume targets that have not yet been demonstrated in production.