Micron, Intel, two other stocks to lead pack as BofA sees $3.2T semis market by 2030

Bank of America analyst Vivek Arya forecasts the semiconductor industry's total addressable market will expand to $3.2 trillion by 2030 from an estimated $1.7 trillion this year. The long-term outlook remains constructive for chips and data-centre infrastructure despite investor concerns that a slowdown in AI development could reduce near-term demand.

Analysis

The relevant investable question is not the long-run industry growth estimate, but whether AI infrastructure spending remains broad enough to support utilization and pricing beyond the current accelerator cycle. A slower model-training cadence would first hit high-beta compute supply chains (NVDA, AMD, AVGO) through order visibility and valuation de-rating, while leading-edge equipment demand (ASML, AMAT, LRCX) could prove more resilient because foundry capacity decisions are made on a multi-quarter horizon. Memory is the highest operating-leverage expression: sustained HBM tightness supports MU margins, but any hyperscaler digestion would reverse spot-pricing expectations faster than logic-chip demand.

BAC has no material direct earnings sensitivity to this theme; the analyst commentary is a sentiment input rather than a bank-specific catalyst. Consensus is likely underestimating that a larger semiconductor profit pool may accrue disproportionately to bottleneck owners—TSMC (TSM), ASML, and HBM suppliers—rather than evenly across the industry. Conversely, the long-duration demand framing risks masking a near-term air pocket if cloud capex shifts from training clusters to lower-cost inference, reducing the mix benefit currently embedded in premium compute and networking multiples.

Over the next 1-3 months, hyperscaler capex guidance, TSM monthly sales, HBM lead-time commentary, and AI-server shipment revisions matter more than multi-year TAM projections. The structural thesis is falsified if AI-related capex remains elevated but semiconductor content per dollar of cloud spend declines, or if MU/TSM report inventory normalization without corresponding pricing strength; either would imply excess capacity is being built ahead of monetizable demand.

Key Decisions for Investors

• No directional BAC trade: treat this as low-impact sector commentary rather than an earnings-relevant bank catalyst.
• Maintain a quality AI-infrastructure bias through a pair trade: long TSM / short SOXX for a 3-6 month horizon. TSM is better positioned to capture leading-edge node scarcity and customer diversification; exit if TSM cuts full-year capex or guides wafer utilization materially lower.
• Use MU as the higher-beta tactical long only after confirmation of sustained HBM pricing and firm DRAM contract prices; target a 3-6 month holding period, with risk defined by any evidence that hyperscaler orders are being pushed out rather than merely reconfigured.
• Avoid adding to richly valued accelerator exposure (NVDA, AMD) solely on long-term industry forecasts. Add only if next earnings validate both backlog conversion and customer concentration stability; a downward revision to hyperscaler capex or weaker networking attach rates would be the near-term de-risk trigger.
• Monitor ASML, AMAT, and LRCX order commentary as the 6-18 month confirmation signal. A broad equipment recovery supports the structural thesis; continued concentration in AI-related leading-edge tools without improving mature-node demand argues for narrower positioning rather than a broad semiconductor overweight.