Andy Jassy Just Made a Move That Should Excite Micron and Broadcom Investors

Amazon raised planned annual capital expenditure to $220 billion from $200 billion to expand AI and cloud capacity, while CEO Andy Jassy said AWS demand will exceed available capacity through at least 2027 and that 2028 demand is "striking." The spending supports a favorable demand outlook for Micron's AI memory products and Broadcom's networking and custom-chip businesses, whose data-center offerings are critical to hyperscaler AI infrastructure. The article argues that sustained AI demand reduces concerns that current infrastructure investment is outpacing eventual revenue opportunities.

Analysis

The key investable implication is that hyperscaler capacity scarcity shifts bargaining power upstream, but unevenly. AVGO is better positioned than MU to monetize sustained build-outs because networking and custom-silicon content is architected into clusters years ahead, while memory remains exposed to the eventual supply response from Samsung and SK Hynix; ANET, MRVL and CRDO are secondary beneficiaries if AI fabrics broaden beyond proprietary switching stacks. For AMZN, the revenue signal is positive but incremental capex raises depreciation, power, and lease burdens before utilization catches up, making AWS margin conversion—not demand commentary—the critical equity variable.

Over the next 1-3 months, consensus revisions should favor suppliers with confirmed hyperscaler backlog and improving gross-margin guidance. The more consequential 6-18 month issue is whether AI inference becomes sufficiently monetizable for cloud customers to renew capacity at attractive rates; a utilization shortfall would first pressure hyperscaler returns on invested capital, then drive order digestion in networking and memory. Treat multi-year demand visibility as a management claim rather than a contracted-revenue proxy until disclosed commitments, power availability, and delivery schedules corroborate it.

Contrarianly, this is not automatically bullish for every AI component vendor after a capex increase: constrained power and construction capacity can defer server deployment, creating a gap between equipment shipments and productive compute. MU has the highest near-term earnings torque to memory pricing, but also the greatest downside if HBM qualification/supply expands faster than demand; AVGO's premium multiple leaves it vulnerable to any custom-ASIC program delay. A sustained AWS operating-margin decline, reduced 2027 capex outlook, or weakening AI-cluster utilization would falsify the broader supply-chain thesis.

Key Decisions for Investors

• Initiate/maintain a 3-6 month long AVGO / short MU pair, sized beta-neutral: favor AVGO's higher visibility networking and ASIC content over MU's more cyclical memory pricing exposure. Reassess if MU raises supply materially or AVGO reports a meaningful hyperscaler program delay; target roughly 10-15% relative upside with a 7-8% stop on the spread.
• Buy AMZN only on post-results evidence that AWS revenue acceleration exceeds incremental depreciation and AWS operating margin is stable-to-up; absent that confirmation, keep AMZN as a watch item rather than chasing capex headlines. The catalyst is the next earnings release, with downside risk concentrated in free-cash-flow and margin guidance.
• Add ANET selectively on AI-networking order evidence, preferably after any broad semiconductor pullback; it offers a cleaner merchant-networking read-through than memory. Exit if cloud revenue growth decelerates for two consecutive quarters or inventory days rise materially.
• Monitor MU HBM pricing, Samsung/SK Hynix capacity announcements, and hyperscaler server-delivery timelines weekly. Do not add outright MU exposure if supply-growth guidance begins to exceed AI-server demand growth, as memory-price reversals can compress earnings rapidly.