Advanced Micro Devices disclosed a 6-gigawatt GPU partnership with Meta in February 2026. Seven weeks later, Broadcom announced a multi-gigawatt extension covering Meta's custom MTIA silicon. Both companies' IR press releases — sourced directly from SEC-registered feeds — land in the same 90-day window as four additional gigawatt-scale AI-infrastructure partnerships from NVIDIA and Caterpillar. For retail investors tracking who's winning hyperscaler AI spending, Meta's two chip picks tell a specific story.
Broadcom reported Q1 FY2026 revenue of $19.3 billion (+28.7% Y/Y) and non-GAAP EPS of $2.05, but the 10-Q reveals a more complex picture. A single distributor now accounts for 42% of total revenue — up from 29% a year ago — as AI semiconductor revenue doubled to $8.4 billion. Meanwhile, infrastructure software grew just 1.4% Y/Y, GAAP EPS declined 14.3% Q/Q due to tax normalization, and capital returns of $10.9 billion exceeded operating cash flow by 32%, requiring new borrowings to bridge the gap.
AVGO's top 5 customers account for 40% of revenue. NVDA's largest customer is 22%. ANET depends on Microsoft (20%) and Meta (15%). Standard AI stock screens show none of this. Here's our 2-signal framework for stress-testing AI beneficiaries.
Qualcomm's consolidated margins hide a secret: the QTL licensing segment earns 72% operating margin vs QCT hardware's 30%. When Apple moves fully to in-house modems, what happens to QCOM's profitability? Our semiconductor ROIC comparison reveals the answer.