joanne feeney

Portfolio manager explicitly includes Broadcom in short list of preferred AI infrastructure beneficiaries along with Nvidia and AMD, citing rapidly growing total addressable market.

Nvidia's dominance will erode as more competitors enter, but the total AI infrastructure market is expanding fast enough that Nvidia, Broadcom, AMD, and hyperscalers all benefit.

Adding Qualcomm as a second custom silicon partner gives Amazon pricing leverage, reduces single-source risk from Marvell, and provides protection against Nvidia's pricing power, with Broadcom as a potential third source.

Portfolio manager explicitly includes AMD in short list of preferred AI infrastructure beneficiaries, citing rapidly growing total addressable market.

Amazon's $4B equity stake creates long-term contracts and skin-in-the-game that de-risks Qualcomm's large design team investments, accelerating innovation through shared risk.

Memory remains cyclical despite AI-driven HBM demand; three leading players will collectively add excess capacity to gain share, causing prices to collapse and earnings to turn negative, making Micron's optically cheap 6x forward multiple a trap.

Nvidia and Broadcom benefit from durable competitive advantages in data center buildout — security, stability, reliability and data handling — that sustain profit margins better than model-layer companies facing commoditization.

Broadcom shares Nvidia's durable moat in data center infrastructure — custom silicon, networking, and software — providing diversified revenue beyond just running models, which protects margins as model-layer competition intensifies.

Micron trades at 6x forward earnings but profits are driven by hundreds of percent price increases that are cyclical and unsustainable; three leading-edge players will collectively add excess capacity, causing prices and stocks to collapse well ahead of earnings turning negative.

Chip suppliers like Broadcom and Nvidia have more durable moats than AI model companies because they provide security, stability, reliability and data handling beyond just running models, sustaining profit margins as model margins compress.