Semiconductor Profit Super Cycle: AI Demand Drives Stock Prices Up Over 66%, Earnings Growth Becomes Main Factor
AI summary card
Semiconductor Profit Super Cycle: AI Demand Drives Stock Prices Up Over 66%, Earnings Growth Becomes Main Factor
The semiconductor industry rose 66% in 5 months, mainly driven by earnings rather than valuation expansion, with AI demand pulling all segments of the supply chain
- The Philadelphia Semiconductor Index (SOX) rose 66% in the past 5 months, with a year-over-year increase of 162%
- The rise was mainly driven by earnings, with the P/E multiple slightly decreasing by 2%
- Earnings outlook for the memory segment improved nearly 5 times, with stock prices rising over 3 times
- GPU/ASIC and semiconductor equipment segments still have room for valuation expansion
- Institutions are bullish on companies like AMD, NVIDIA, and Applied Materials, giving them Overweight ratings
Report interpretation
Overview
The Bernstein report explains the current significant earnings-driven rally in the semiconductor industry. The SOX index rose 66% in 5 months, but the P/E multiple slightly decreased by 2%, showing that earnings growth contributed to 70% of the total gain. AI demand is driving a sector rotation across the entire supply chain, from memory and equipment to optical components. The report is optimistic about the growth prospects of companies like AMD, NVIDIA, and Applied Materials.
Core views
The core views focus on the earnings-driven logic of the semiconductor industry rather than valuation expansion. In the recent 66% rise of the SOX index, earnings growth contributed as much as 69%, while the P/E multiple decreased by 2%. The speed of this earnings improvement far exceeds historical cycles, making sustainability a key concern. AI demand is creating a cascading effect across the supply chain: the memory segment led the charge (average stock price YTD up over 3 times), followed by the equipment segment, then optical and analog chip segments, with CPU-related companies recently gaining momentum. The drivers vary significantly across sub-segments: the memory segment saw a 386% surge in earnings expectations while valuations contracted by 21%; the GPU/ASIC segment saw only an 11% valuation expansion, suggesting potential for catch-up; the CPU segment showed dual drivers of earnings and valuation (56% earnings growth + 47% valuation expansion). In terms of specific stock recommendations: NVIDIA (AI data center demand) and Applied Materials (equipment demand) remain core bullish picks; AMD is newly favored, with optimism for its CPU business and AI accelerator opportunities, expecting 2028 EPS to reach $20; Broadcom's valuation is attractive (CY27 forward P/E in the teens). The memory segment's short-term momentum exceeds expectations, with long-term optimism for DRAM companies; TSMC is seen as a high-quality core asset.
Analysis framework
The report uses an earnings-driven factor analysis method, breaking down index movements into contributions from earnings growth and changes in valuation multiples. Historical cycle comparisons show that while the current 69% earnings growth rate has not surpassed historical highs, the speed of achievement is record-breaking. The report also employs a sub-industry comparison framework (GPU/ASIC vs. memory vs. equipment) to reveal differences in valuation and earnings alignment across segments. Industry analysis uses a demand transmission model, tracking how AI demand spreads from memory (HBM acceleration) → semiconductor equipment → optical modules → analog chips → CPUs, supplemented by fund flow data for validation. Finally, the conclusions are cross-verified through segmented earnings forecasts (e.g., AMD's 2028 $20 EPS) and relative valuation (Broadcom's CY27 mid-term P/E).
Methodology notes
Supply Chain Transmission Analysis
The report tracks how AI demand spreads sequentially from memory to equipment, optical modules, analog chips, and finally to CPU companies. This transmission analysis reveals the internal rotation logic of the industry, similar to a domino effect.
Return Attribution Model
Breaking down index returns into contributions from earnings growth and valuation expansion (e.g., the report shows that 69% of YTD gains came from earnings), helping to isolate fundamental drivers from sentiment-driven factors in the market.
Forward PE Anchoring
The report uses CY27 forward P/E to evaluate true valuations (e.g., Broadcom's mid-term forward P/E in the teens), which better reflects long-term rationality compared to static P/E, especially for high-growth segments.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- AMD.USBenefits from AI accelerator demand and CPU business growth
- Strengths
- Diversified competitor in the AI accelerator market, stable CPU business
- Weaknesses
- Faces strong competition from NVIDIA
- Comparison
- Valuation has room to rise compared to NVIDIA
- Risks
- AI demand realization falls short of expectations
- NVDA.USCore beneficiary of AI data centers
- Strengths
- Absolute leadership in AI chips
- Weaknesses
- Valuation already partially reflects expectations
- Comparison
- CY27 forward P/E around 15x, still attractive
- Risks
- Customer in-house chip substitution
- AMAT.USDirect beneficiary of semiconductor equipment demand
- Strengths
- Full-process equipment coverage
- Weaknesses
- Reliant on capital expenditure cycles
- Comparison
- Among the top beneficiaries in equipment providers
- Risks
- Geopolitical policy changes
Key data
- SOX Index Gain66%5-month gain
- Year-over-Year Gain162%Far exceeds Nasdaq's 45%
- Earnings Contribution69%Percentage of YTD index gain from earnings growth
- Memory Segment Earnings Outlook386%NTM EPS growth
Impact & implications
The report believes the current semiconductor rally is driven by real earnings, with AI infrastructure investments creating sustained demand. Equipment manufacturers directly benefit from wafer fab expansion needs; GPU companies, though recently lagging, have room for valuation recovery. While the memory segment's short-term gains may be excessive, new demand structures like HBM provide long-term support. TSMC, as a technology platform provider, will continue to dominate advanced node红利.
Risks
- Earnings sustainability: Whether the current growth rate can be maintained
- Valuation pullback risk in the memory segment
- Smartphone demand falls short of expectations
What to watch
- Validation of actual AI chip shipments
- Signals of memory price inflection points
- Changes in wafer fab capital expenditure intensity