Hyperscalers’ $720 B AI Capex Spurs $5.5 B ETF Inflows
MU•Retail investors poured $3.2 billion into semiconductor ETFs since January 2025, with April seeing a record $5.5 billion inflow as hyperscalers forecast $600 billion–$720 billion in 2026 AI capex, driving memory chip shortages. Meanwhile, Marvell’s $2 billion Nvidia investment and inference silicon focus signal potential outperformance relative to larger peers like Micron.
1. AI Capex Boom and ETF Inflows
Retail investors have poured $3.2 billion into semiconductor ETFs since January 2025, including a record $5.5 billion in April, driven by expectations of $600 billion–$720 billion in AI infrastructure expenditure by major hyperscalers.
2. Memory Chip Supply Dynamics
Soaring AI workloads are consuming high-bandwidth memory at scale, creating shortages that underpin upside for memory-focused firms like Micron as demand outstrips production capacity.
3. Competitive Outlook with Marvell
Marvell’s $2 billion strategic investment from Nvidia and its focus on inference and power-efficient silicon position it to potentially outpace larger peers, signaling intensified competition for Micron in the AI semiconductor segment.
