$QCOM just entered the custom silicon and optical connectivity race, and Amazon is the customer. They announced a multi-generation collaboration to build custom silicon for AWS AI infrastructure, plus optical connectivity solutions extending to 1.6T and beyond, using Qualcomm's SerDes and optical DSP technology. Think about what this means for the full-stack thesis. Google has TPUs with Broadcom. OpenAI has Jalapeño with Broadcom. Anthropic and Meta too. Amazon already has Trainium. Now Qualcomm gets a seat at the table for AWS's next generations of silicon, in a market that was starting to look like Broadcom's alone. And it's not just compute. The optical piece matters just as much. High bandwidth interconnect at 1.6T and beyond is exactly where the whole industry is racing, and Qualcomm just confirmed it has real technology in that layer, not just phone chips. There's also a quieter detail buried in the release. Qualcomm is deepening its own use of AWS infrastructure, including Bedrock, for chip design workloads to shorten design cycles. That's Qualcomm becoming a customer of AWS at the same time it becomes a silicon supplier to AWS. Circular, but it tells you how seriously they're leaning in. Amon's framing: decades of power-efficient processing leadership applied to AI infrastructure. This is the same pattern playing out everywhere. Every hyperscaler wants custom silicon, and now there are two credible partners instead of one.