AI demand extends semiconductor cycle visibility; BofA raises U.S. semiconductor and WFE forecasts

Overview

This report is Bank of America's updated outlook on the U.S. semiconductor industry. The core view is that the AI industry is shifting from an early phase of needing to prove ROI to a phase constrained by structural limits in chips, power, memory, and capacity. As visibility improves for AI data centers, HBM/DRAM/NAND, server CPUs, advanced process nodes, and semiconductor equipment demand, the report raises industry TAM, WFE forecasts, and price targets for multiple companies.

Core views

The report believes AI could add about $1tn in semiconductor industry sales over the next five years. Key drivers include expansion in AI data center system TAM, long-term memory agreements improving supply-demand and price visibility, higher WFE driven by advanced process and packaging complexity, analog chip demand driven by AI power requirements, server CPU market expansion, and incremental long-term demand from physical AI. At the same time, PCs, smartphones, and some consumer end markets remain drags, while automotive and industrial demand is improving modestly.

Analysis framework

The report uses a combination of top-down semiconductor TAM forecasting, end-market breakdowns, WFE intensity and WFE-per-wafer analysis, memory supply-demand balance, pricing trends, and individual stock valuation frameworks. For the equipment cycle, the report argues that simply looking at WFE as a percentage of sales can be distorted by rising memory ASPs, so it places greater emphasis on WFE investment per 12-inch wafer start.

Methodology notes

• Industry ForecastingTAM and CAGR Framework

  Derive semiconductor industry size through end markets, product categories, and annual growth rates.

  The report models total semiconductor TAM, core semiconductors, AI data centers, server CPUs, HBM, and WFE separately, and uses CY25-CY30E CAGR to measure the growth slope.

• Equipment Cycle AnalysisWFE Intensity and WFE-per-wafer

  Compare WFE as a share of industry sales with equipment investment per wafer.

  The report believes that rising memory ASPs currently depress the apparent WFE intensity metric, so WFE per wafer better reflects true equipment demand driven by process complexity, capacity expansion, and technology migration.

• Supply-Demand AnalysisDRAM/NAND supply-demand adequacy ratio

  Use the supply-demand adequacy ratio to judge whether memory has a clear oversupply.

  The report expects DRAM and NAND supply-demand adequacy ratios to remain above 110% through CY28, and therefore sees no obvious near-term oversupply, with pricing cyclicality potentially lower than in past downcycles.

• Valuation methodsPE, PEG, and Sum-of-Parts valuation

  Apply different valuation benchmarks and sum-of-the-parts methods to different semiconductor companies.

  The report uses a PE framework based on discounted CY30 EPS power for INTC, a sum-of-the-parts approach for ARM's IP business and chip business, and shifts some AI beneficiaries to a CY28E valuation basis.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• MU

  Core beneficiary of AI memory and HBM.

  Strengths

  The report maintains Buy and raises the price target to $1,500, arguing that the HBM business makes the company's fundamentals different from the traditional commodity memory cycle.

  Weaknesses

  Valuation is already above the upper end of its historical P/B range, requiring greater confidence in HBM durability and price stability.

  Comparison

  Compared with the traditional DRAM/NAND cycle, the report places greater emphasis on the long-term supply-demand and earnings leverage created by HBM.

  Risks

  Faster-than-expected HBM supply expansion, slower customer demand, or price declines could weaken valuation support.

• INTC

  Potential beneficiary in server CPUs, IDM, and external foundry.

  Strengths

  The report maintains Buy and raises the price target to $160, believing CY30 EPS power could reach about $6.24, with server CPU and foundry opportunities expanding long-term upside.

  Weaknesses

  There is still uncertainty around external foundry orders and specific deal details.

  Comparison

  The report shifts from a prior CY28 sum-of-the-parts approach to full IDM company valuation to capture further-out CY29-CY30 potential.

  Risks

  Process execution, foundry customer wins, capex, and the competitive landscape are the main risks.

• ARM

  Beneficiary of server CPU expansion and ecosystem share gains.

  Strengths

  The report expects the ARM ecosystem to reach about 50% server CPU value share in CY30E, including commercial CPUs and custom CPUs from hyperscale cloud providers.

  Weaknesses

  The report believes the roughly $420 share price already largely reflects the opportunity, so it maintains Neutral.

  Comparison

  ARM has faster share gain potential versus the x86 ecosystem, but valuation is already fuller.

  Risks

  Server ecosystem adoption speed, customer in-house development pace, valuation multiples, and competitive pushback could affect returns.

• CRDO

  Beneficiary of AI interconnect and high-speed connectivity demand.

  Strengths

  The report maintains Buy and raises the price target to $340, benefiting from expansion in AI system scale and rising network connectivity complexity.

  Weaknesses

  High growth expectations may already be reflected in an elevated valuation.

  Comparison

  Compared with broader semiconductor companies, CRDO is more sensitive to AI data center interconnect demand.

  Risks

  Customer concentration, product iteration, and fluctuations in data center capex are the main risks.

• MRVL

  Beneficiary of AI data centers, custom chips, and networking demand.

  Strengths

  The report maintains Buy and raises the price target to $365, believing CY30E EPS power could exceed $15.

  Weaknesses

  Forward EPS power depends on AI project ramp and sustained customer demand.

  Comparison

  MRVL sits at the intersection of AI ASICs, networking, and data center semiconductors, giving it broad exposure to beneficiaries.

  Risks

  Delays in AI ASIC projects, intensifying competition, and customer capex contraction would affect expectations.

• ALAB

  Beneficiary of AI infrastructure connectivity and system-level scaling.

  Strengths

  The report raises the price target to $450 and points to higher long-term EPS power.

  Weaknesses

  The report maintains Neutral, implying current valuation is already fairly full.

  Comparison

  Compared with Buy-rated names, ALAB has a clear growth story but a lower valuation margin of safety.

  Risks

  Valuation pullback, competition, and slower AI server deployment pace are the main risks.

• TER

  Beneficiary of semiconductor testing and rising complexity in AI/advanced chips.

  Strengths

  The report maintains Buy and raises the price target to $525, benefiting from advanced chips and an upcycle in equipment.

  Weaknesses

  Demand for test equipment may still be affected by swings in the semiconductor capex cycle.

  Comparison

  Compared with front-end equipment companies, TER is more exposed to testing, but it is similarly driven by rising complexity.

  Risks

  Customer order volatility, an equipment cycle downturn, and valuation multiple compression are the main risks.

• AMAT/LRCX/KLAC/MKSI/AEIS

  Equipment-chain basket benefiting from higher WFE, advanced processes, EUV, etch/deposition, and increased advanced packaging complexity.

  Strengths

  The report raises the WFE forecast to $250bn in CY28E, with structural upside drivers in WFE per wafer across logic, DRAM, and NAND.

  Weaknesses

  Equipment demand is meaningfully affected by capacity ramps, customer budgets, and geopolitical restrictions.

  Comparison

  Compared with end-chip companies, the equipment chain is more directly driven by process complexity and capacity expansion cycles.

  Risks

  Lower WFE intensity, delayed non-China capacity expansion, digestion of China demand, and export restrictions are the main risks.

Key data

• CY30E total semiconductor TAMabout $2.7tnRaised from the prior about $2.3tn, implying about 28% CAGR for CY25-CY30E.
• AI data center system TAMabout $1.7tn in CY30EUp sharply from about $273bn in CY25, and the main source of incremental semiconductor sales.
• Core semiconductors (excluding memory) sizeabout $1.1tn in CY30EUp from about $567bn in CY25, about 14% CAGR, mainly driven by server silicon and wired communications.
• CY26 semiconductor/core semiconductor growth+103%/+27% YoYMemory, data centers, and improvements in industrial and automotive are the main contributors, while PCs, smartphones, and consumer remain under pressure.
• CY26 memory growthabout +298% YoYDRAM about +309% YoY, NAND about +295% YoY.
• WFE forecastCY27E $190bn; CY28E $250bn; CY29E $268bn; CY30E $292bnCY28E forecast was raised 23% from the prior $203bn, driven by cleanroom availability, long-term memory agreements, and technology inflection points.
• WFE CAGRabout 20% for CY25-CY30EMemory WFE about 21% CAGR, non-memory about 19% CAGR.
• HBM TAMabout $246bn in CY30Up from about $35bn in CY25, about 34% CAGR; HBM capacity per accelerator is expected to rise from about 187GB in CY25 to about 464GB in CY30.
• HBM pricing assumptionabout $17.5/GB in CY27-CY28Above about $14.3/GB in CY26, reflecting tight supply-demand conditions.
• Server CPU TAMabout $170bn in CY30EThe combined value share of the ARM-related ecosystem in commercial CPUs and custom CPUs is expected to reach about 50%.
• Memory supply-demand viewDRAM/NAND adequacy ratios remain above 110% through CY28Based on this, the report sees no obvious oversupply and believes near-term downside risk to ASPs is relatively manageable.
• Major price target changesMU $1,500; INTC $160; ARM $460; CRDO $340; MRVL $365; ALAB $450; TER $525; ACLS $156The report also raises price targets for multiple semiconductor equipment and AI beneficiary companies.

Impact & implications

If the report's view plays out, the investment theme across the semiconductor supply chain will broaden from AI GPUs alone to memory, advanced process equipment, advanced packaging, server CPUs, network interconnects, and power-related chips. Memory and WFE could become more persistent bottlenecks in the AI cycle, while weak consumer electronics recovery would continue to drive structural divergence.

Risks

• AI data center demand or ROI may fall short of expectations, causing the higher semiconductor TAM and WFE assumptions to fail.
• HBM, DRAM, and NAND supply may expand faster than expected, potentially leading to price declines and lower earnings leverage.
• Constraints in cleanrooms, advanced packaging, power, and capex may delay CY27-CY28 capacity ramps.
• Persistently weak demand in PCs, smartphones, and consumer electronics may drag on the recovery in core semiconductors.
• China WFE digestion, export restrictions, and geopolitical changes may affect the equipment-chain order mix.
• Price targets for many companies are based on forward CY28-CY30 valuations; if rates, risk appetite, or AI multiples fall, valuation pressure could be significant.
• Execution uncertainty remains around INTC foundry, ARM server share, and the rollout of AI projects at MRVL/CRDO/ALAB.

What to watch

• Long-term supply agreements, HBM capacity, and pricing guidance from MU and other memory makers.
• Whether DRAM/NAND spot and contract prices can remain stable after 2026.
• The pace of cleanroom, advanced packaging, and front-end equipment capex at TSMC, Samsung, INTC, MU, and others.
• Whether CY27-CY28 WFE orders track toward the $190bn/$250bn path.
• Changes in AI data center system TAM, server CPU TAM, and capex by hyperscale cloud providers.
• Whether demand in PCs, smartphones, automotive, and industrial semiconductors improves as the report assumes or continues to diverge.
• Progress in INTC external foundry deals, ARM server CPU share, and the custom CPU ecosystem.
• The pace of reshoring projects in the U.S. and outside China, as well as EUV and High-NA EUV adoption.