Big win for @Qualcomm - finally can talk about this 😃 Qualcomm just locked in Amazon as its second major hyperscaler partner for data-center silicon building upon the Investor Day roadmap and promised outlook which is starting to look healthy Qualcomm has issued a warrant to ‌Amazon, giving it the right to buy an about $4 billion stake, as part of a collaboration to develop custom chips for AI data centers. The warrant lets Amazon buy Qualcomm shares at $161.26 apiece and is linked to up to $60 billion of business under the deal At #QualcommInvestorDay in June we highlighted how Qualcomm laid out Dragonfly, HBC (stacked DRAM as an HBM alternative to break the memory wall for inference), an annual AI200/AI250/AI300 cadence, the Meta multi-gen CPU deal, Modular acquisition, Microsoft partnership and a path to $15B data-center revenue by FY29. Today’s announcement with Amazon is a good validation for Qualcomm’s datacenter portfolio and roadmap. - Multi-generational collaboration on customized silicon at scale for @awscloud AI Inference infrastructure. - Joint work on optical connectivity up to 1.6T (and next-gen) using Qualcomm’s SerDes + optical DSP for high-bandwidth data-center interconnects. This is where @alphawavesemi acquisition shines and the foundation for the deal. Congrats @TonyPialis - Qualcomm will also lean harder on AWS (including Bedrock) for its own EDA workloads to shorten chip design cycles makes it a nice symbiotic relationship. This pairs Amazon’s scale, security, and price-performance infrastructure with Qualcomm’s power-efficient processing and system integration bringing in the “compute + connectivity” combo @cristianoamon highlighted as essential as token demand explodes. So the timeline : - Meta was the first named multi-gen data-center CPU partner (Dragonfly C1000). Amazon now adds custom silicon + optics for inference clusters. - HBC and the Dragonfly accelerator roadmap were designed for exactly this kind of energy-efficient, high-bandwidth inference at hyperscale. - Microsoft to add HBC architecture into Azure AI infrastructure - endorsement + planned deployment - Two global-scale hyperscalers + custom ASIC programs shipping later this year = the diversification thesis (non-handset to ~$40B by FY29, data center as a third pillar) is moving from slides to contracts. - Optical 1.6T traction: AI clusters are becoming interconnect-bound as much as compute-bound. Qualcomm (w/ Alphawave Semi) bringing its SerDes/DSP expertise here is a natural extension of the “full-stack from device to data center” offering. $QCOM #AI #DataCenter #HBC #Dragonfly $AMZN $TSM

Will be in Bay Area next week, look forward to meet many of you!! #AppleEvent $AAPL

LG Display (@lgdisplayoled ) just unveiled FLiPP (FMM-Less innovative Pixel Patterning), a breakthrough that could solve OLED’s biggest manufacturing bottleneck. For over a decade, OLED manufacturing has relied on Fine Metal Masks (FMM) essentially micro-stencils that are costly, limit panel sizes due to metal sagging, and restrict glass substrate efficiency. FLiPP replaces mechanical masks with UV photolithography patterning. LG claims following performance leap: 🔆 +60% Brightness (1.6x) ⏳ +140% Lifespan (2.4x) ⚡ -13% Power Consumption 📐 +55% Aperture Ratio -8.5th-Gen Full Glass Processing: Enables single whole-piece 8.5G mother glass production without dividing substrates thus boosting utilization efficiency by up to 64%. -Total Scalability: Eliminates size/resolution constraints, scaling seamlessly from 1-inch micro-displays (AR/VR/wearables) up to 100-inch ultra-large TVs. A massive moment for RGB OLED production economics and display longevity across consumer tech.📱💻📺 Might bring some reprieve for struggling Display makers being squeezed due to Chipflation counterpointresearch.com/en/… counterpointresearch.com/en/… $OLED