Profit-Driven AI Bull Market: The Semiconductor Rally Is Far From Over

Overview

This in-depth report aims to unpack the drivers behind the semiconductor sector’s ~72% rally since the beginning of 2026 and assess its sustainability. The core conclusion is that this rally is not driven by speculative valuation expansion but is underpinned by solid earnings growth. The AI infrastructure build-out cycle remains in its early and durable phase. Bank of America maintains a highly optimistic outlook on the AI-driven semiconductor industry, particularly favoring sub-sectors such as compute, storage, analog chips, and semiconductor capital equipment. It forecasts the total AI market opportunity to triple by 2030, reaching approximately $1.7 trillion.

Core views

The report’s central argument refutes market concerns about a 'valuation bubble' in semiconductors. Data shows that despite the Philadelphia Semiconductor Index (SOX) surging significantly year-to-date, its forward P/E ratio remains around 25.6x—essentially flat compared to the start of the year and well below its previous peak of ~30x. This indicates that the index’s gains are primarily driven by upward revisions in earnings expectations, not blind P/E expansion, providing a solid foundation for sustainability. On the demand side, the report argues that the sustainability of AI infrastructure investment is underestimated by the market and identifies four key supports: 1) Leading AI labs (e.g., OpenAI, Anthropic) are experiencing explosive revenue growth of 3–5x YoY, creating a commercial feedback loop for massive capex; 2) AI commercialization is accelerating—as exemplified by Google, whose token consumption has grown 7x YoY. Agentic workflows can generate 10–1,000x more tokens than chat interactions, shifting the business model from 'per-seat' to 'per-usage' and offering clear monetization paths for cloud providers; 3) Supply-side constraints persist, with nearly all deployed infrastructure operating at full utilization; and 4) Sovereign, enterprise, and industrial demand is underestimated. Based on this, Bank of America forecasts the TAM for AI data center systems to grow from $264 billion in 2025 to over $1.7 trillion by 2030, with AI servers accounting for ~75%, networking equipment ~20%, and storage ~5%. From a valuation perspective, the report uses the PEG ratio (P/E relative to earnings growth) to identify investment opportunities across sub-sectors. In compute and storage, companies like NVIDIA (NVDA), Micron (MU), and Credo (CRDO) trade at PEGs below 1x and below their historical P/E averages, indicating attractive valuations. While semiconductor capital equipment trades at relatively higher valuations, this is justified by expectations of >20% annual sales growth over the next 2–3 years—far exceeding the historical 10% rate. The analog chip segment offers a unique 'low-beta AI play,' with only ~10% average AI revenue exposure, while also benefiting from cyclical recoveries in aerospace, defense, and automotive markets—combining defensiveness with growth potential.

Analysis framework

The report employs a combined top-down and bottom-up analytical framework to build its thesis. First, at the macro and market level, it decomposes index performance into two core drivers: earnings and valuation. By comparing the SOX Index’s ~72% YTD price gain against the stability of its forward P/E ratio, it clearly demonstrates that the rally is almost entirely earnings-driven—effectively countering 'bubble' narratives. Next, on the industry demand side, the report constructs a rigorous logic chain: 'Explosive revenue growth at leading AI labs' → 'Strong growth in cloud providers’ RPO' → 'Sustainable AI capex' → 'Long-term strength in semiconductor demand.' It cites specific revenue forecasts for OpenAI and Anthropic, along with RPO and capex data from hyperscalers like Microsoft, Amazon, and Google, transforming abstract 'AI demand' into verifiable financial metrics. Finally, at the stock-selection level, the report uses the PEG ratio (P/E divided by earnings growth rate) to compare valuation attractiveness across sub-sectors and individual stocks. A PEG below 1 is typically seen as undervalued. Using this metric, the report identifies attractive opportunities in compute and storage and provides differentiated investment rationales based on valuation levels and growth outlooks across segments. This approach systematically integrates macro views, industry trends, and micro-level valuation into a clear decision-making roadmap for investors.

Methodology notes

• Valuation MethodPE/PEG valuation

  Combined use of PE (Price-to-Earnings) and PEG (PE-to-Growth) ratios

  PE measures the price relative to earnings per share, while PEG further links PE to earnings growth rate: PEG = PE / earnings growth rate. A PEG below 1 is often interpreted as potentially undervalued. The report heavily relies on this metric—for example, noting that NVIDIA, Micron, and others have PEGs below 1x—to argue that their valuations are attractive relative to their high-growth potential.

• Industry/Market Analysis FrameworkPrice-Volume Decomposition

  Decomposing stock/index movements into earnings and valuation components

  This is a common analytical approach to determine whether market gains are driven by healthy earnings growth or speculative valuation expansion. The report uses this by contrasting the SOX Index’s ~72% YTD gain with the fact that its forward P/E has remained 'essentially flat,' leading to the core conclusion of 'earnings-driven' growth—a direct application of price-volume decomposition thinking.

• Industry/Market Analysis FrameworkS-Curve of Penetration

  S-curve of AI technology adoption and industry lifecycle

  The report posits that the AI infrastructure investment cycle remains in an 'early and durable' phase, implicitly invoking the S-curve theory of technology diffusion—where new technologies initially penetrate slowly, then accelerate rapidly before saturating. The forecast of a tripling TAM by 2030, alongside emerging applications like token consumption and agentic workflows, supports the view that AI is still in its early acceleration phase.

• Quantitative/Factor/Portfolio TheoryBeta/alpha analysis

  Application of low-beta strategies in sector rotation

  Beta measures a stock’s volatility relative to the broader market (e.g., S&P 500). The report notes that the analog chip segment has low AI revenue exposure (~10%), implying a lower beta, making it a 'defensive' way to participate in the AI theme. This means that during periods of heightened market volatility, low-beta assets may hold up better, offering investors a balanced growth-and-defense strategy.

• Cyclical and Sentiment Framework

  Attributing stock gains to 'earnings-driven' rather than 'valuation-driven' to assess market cycle health

  The report’s core logic distinguishes between a 'profit bull market' and a 'liquidity/valuation bull market.' In a healthy bull market, price gains should be primarily driven by corporate earnings growth. If P/E multiples expand rapidly in tandem, it may signal overheated sentiment and potential correction risk. By demonstrating stable SOX P/E ratios, the report argues this semiconductor rally is fundamentally sound and more sustainable.

• Fixed Income and Credit AnalysisSpreads and Asset Quality

  Financial metrics (e.g., RPO, capex intensity) and sustainability of capital expenditure

  The report analyzes hyperscalers’ commercial RPO (Remaining Performance Obligations—signed but unearned revenue) and capex intensity to validate the sustainability of AI-related capex. Rapid RPO growth implies secured future revenue, supporting current large-scale investments. This parallels credit analysis, where a borrower’s future income certainty is used to assess debt-servicing capacity and the prudence of current spending plans.

Asset mapping & comparison

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

• AMD (Advanced Micro Devices)

  Benefits from growing AI CPU/GPU market share and potential for >50% annual EPS CAGR.

  Strengths

  Strong potential for AI CPU/GPU market share gains and robust EPS growth.

  Weaknesses

  Slower growth in cyclical PC, embedded, and gaming console markets.

  Risks

  Uncertainty around timing/scale of Middle East AI projects; irregular consumer and enterprise spending; heavy reliance on a single outsourced manufacturing partner; maturing gaming console cycle.

• ADI (Analog Devices Inc.)

  Benefits from data center exposure (~20% revenue) and AI power demands, plus best-in-class free cash flow generation.

  Strengths

  Data center exposure, best-in-class profitability and FCF, differentiated communications business.

  Weaknesses

  Near-term concerns over rising costs and tariffs.

  Risks

  Recession reducing auto/industrial demand and pressuring margins; failure to realize Maxim merger synergies; competition from lower-cost producers; US-China trade tensions/tariff risks.

• AVGO (Broadcom Inc)

  Core supplier in AI through custom ASICs and networking, directly benefiting from AI compute demand.

  Strengths

  Double-digit EPS growth, best-in-class profitability, FCF, and returns.

  Weaknesses

  High exposure to Apple and Google creates replacement risk; faces competition from NVIDIA in networking.

  Risks

  Semiconductor cyclicality; customer concentration; competition in networking/smartphones/storage; financial and integration risks from frequent M&A; ~$60B net debt.

• NVDA (NVIDIA Corporation)

  Leader in AI accelerated computing and networking, direct beneficiary of AI capex.

  Strengths

  Dominant share in fast-growing AI compute/networking markets.

  Weaknesses

  Faces competition from large public companies, in-house cloud projects, and private firms.

  Risks

  Weakness in consumer gaming; intensifying AI competition; escalating China export restrictions; irregular sales in enterprise/data center/auto segments; slowing capital return; antitrust scrutiny.

• MU (Micron Technology, Inc)

  Memory leader benefiting from AI-driven HBM demand and cyclical recovery in traditional memory.

  Risks

  DRAM ASP declines worse than expected; intensifying competition from Chinese entrants; market share loss; weak end-market demand from data centers, smartphones, and PCs.

Key data

• SOX Index YTD Gain~72%Demonstrates strong sector momentum.
• SOX Index Forward P/E Ratio~25.6xFlat since the start of the year and below the prior peak of ~30x, indicating earnings-driven gains.
• AI Data Center Systems TAM (2030 Estimate)~$1.7 TrillionGrowing from $264 billion in 2025 at a 45% CAGR.
• OpenAI 2030 Revenue Estimate~$284 BillionPer BofA analyst Justin Post; agentic business contributes ~20%.
• Anthropic Annualized Recurring Revenue (ARR)Increased from $9B in December to $19B in MarchImplied quarterly increase potentially exceeding $2.5B, recent evidence of AI monetization.
• Microsoft Commercial RPO~$627 BillionUp ~99% YoY, including a new $250B contract from OpenAI for Azure services.
• Amazon AWS Backlog$364 BillionExplicitly excludes the recent >$100B Anthropic deal.

Impact & implications

The report contends that, amid sustained AI demand, the semiconductor industry is in a multi-year structural growth cycle. For investors, the 'earnings-driven' nature of this rally means that despite significant gains, dangerous valuation bubbles have not formed, and fundamentals continue to support further upside. Within the sector, value creation from AI investment will broaden beyond core compute chips and HBM to include networking, storage, analog chips, and semiconductor equipment. The report specifically notes that traditional CPUs have relatively limited economic value in the AI era, while accelerators (GPUs/ASICs), memory, and foundries remain superior ways to capture AI growth leverage.

Risks

• Execution Risk: Bottlenecks in power, data center construction, financing, and component supply could delay actual AI capex deployment.
• AI Monetization Risk: Despite rapid revenue growth at leading labs, low margins and high cash burn remain profitability challenges for the sector.
• Semiconductor Cyclicality: Shifts in market sentiment or fundamentals around the AI theme could trigger stock corrections.
• Geopolitical Risk: US-China trade tensions, tariff escalations, or China chip export restrictions could impact multiple companies.
• Competitive Risk: Intensifying competition in AI accelerators, custom ASICs, and networking equipment could shift market shares.

What to watch

• Upcoming Computex event (early June): Expected to feature more CPU announcements, including NVIDIA’s detailed breakdown of its $200B TAM by 2030.
• Hyperscaler capex and RPO data: Key indicators to validate sustained AI demand.
• Quarterly DRAM pricing: As long as prices continue to rise sequentially, the 'pain trade' in memory remains compelling.
• Commercialization of Agentic AI workflows: These workflows generate far more token consumption than chat, representing the next core driver of compute demand growth.