2026 Global Technology Outlook: AI and memory remain strong in the first half, but beware of demand destruction in the second half

Overview

This is a Morgan Stanley outlook report on the global technology sector for 2026, covering AI semiconductors, memory, logic chips, semiconductor equipment, foundry, hardware, and the AI server supply chain. The report’s core view is that the AI capex and commodity price uptrend that began in 2025 will continue through the first half of 2026, but may face challenges in the second half of 2026 due to the pass-through of high costs, weakening end-demand elasticity, and pressure on AI investment returns.

Core views

The report argues that the technology sector in 2026 will tell 'a tale of two halves': in the first half, AI spending, memory prices, and logic semiconductor earnings will continue to support the market; in the second half, investors need to watch for demand destruction caused by price increases, as well as margin pressure on chip designers and end-hardware from wafer/OSAT/memory costs. From an investment perspective, it favors stocks with pricing power, AI infrastructure exposure, earnings upgrades, and end-market momentum, while also recommending some reasonably valued, cash-flow-strong, previously overlooked analog and consumer-related technology companies as diversification tools.

Analysis framework

The report uses a bottom-up stock-picking framework, focusing on growth at reasonable valuation, earnings forecast revision momentum, end-market trends, and changes in year-over-year revenue. Industry analysis revolves around the AI capex cycle, memory supply constraints, Agentic AI diffusion, China AI chip self-sufficiency, ABF substrate supply-demand, and the rising content value in AI servers, combined with cross-sectional comparisons of valuation, P/B, ROE, P/E, target price, and upside.

Methodology notes

• Stock-picking frameworkGARP and earnings revision momentum

  Growth at a reasonable price

  The report explicitly states that stock selection is based on three dimensions: growth at a reasonable price, momentum in earnings estimate revisions, end-market momentum, and the expected change in year-over-year revenue growth.

• Cycle analysisAI investment cycle and semiconductor cycle positioning

  Buy-the-dip opportunities within a long AI capex cycle

  The report views the current technology cycle as a transformative and long-lasting AI investment cycle, and believes pullbacks can be used to reassess more attractive entry points.

• Supply chain bottleneck analysisMigration of AI infrastructure bottlenecks

  Bottlenecks are expanding from GPU to CPU, memory, foundry, substrates, and equipment

  With the growth of Agentic AI and inference demand, the report argues that AI systems require not only higher compute power but also smarter resource coordination across the full stack; therefore, CPU, memory, wafer capacity, substrates, and equipment could all become critical bottlenecks.

Asset mapping & comparison

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

• NVIDIA / Broadcom / MediaTek

  Key beneficiaries in AI processors and custom computing

  Strengths

  The report maintains OW; among them, NVIDIA is preferred for its leading ROI in cloud scenarios and expected volume ramp of Vera Rubin in 2H26, while Broadcom and MediaTek benefit from ASIC and AI computing demand.

  Weaknesses

  Rising enthusiasm for ASICs may alter the competitive landscape, and a high-investment AI cycle with lower-than-expected returns would pressure valuations.

  Comparison

  Compared with EW-rated names such as AMD, Marvell, and Intel, the report shows a stronger preference for NVIDIA, Broadcom, and MediaTek.

  Risks

  A slowdown in AI capex, customer in-house substitution, supply chain bottlenecks, and valuation digestion.

• Samsung / SK Hynix / Micron / SNDK / KIOXIA

  Beneficiaries of the memory and NAND cycle

  Strengths

  The report believes memory remains an important AI bottleneck, with long order visibility and rapidly shifting pricing power; memory stocks may outperform in a cautious market.

  Weaknesses

  The memory cycle may still ultimately be affected by additional capacity and demand elasticity.

  Comparison

  The report favors Samsung, SK Hynix, Micron, and SNDK, while Winbond and Nanya Tech are shown in the charts as relatively less preferred.

  Risks

  Overexpansion, slowing AI demand, peaking prices, and pressure on consumers from rising memory prices.

• TSMC

  Beneficiary of advanced-node foundry and potential HBM4/4e base die demand

  Strengths

  The report believes AI demand can help TSMC sustain about 20% revenue CAGR over the next five years, while China AI GPU foundry opportunities may also provide incremental upside.

  Weaknesses

  Advanced-node capacity expansion requires high capex, while geopolitics and customer concentration are long-term constraints.

  Comparison

  Compared with legacy foundry, leading-edge foundry has more structural advantages in AI demand.

  Risks

  Customer order cuts, changes in advanced-node utilization, geopolitics, and capex returns.

• ASML / ASMI / AMAT / Advantest / DISCO / Tokyo Seimitsu / Micronics

  Beneficiaries in semiconductor equipment and test equipment

  Strengths

  The report believes bottlenecks are shifting toward front-end wafer capacity, DDR5 DRAM, and advanced logic, and SPE upside will broaden in 2H26; AMAT is seen as having the most leverage to greenfield DRAM.

  Weaknesses

  Equipment orders depend on the pace of foundry capex, and if AI or memory expansion is weaker than expected, revenue recognition may be delayed.

  Comparison

  The report favors equipment companies with high DRAM exposure and those benefiting from China localization and TSMC advanced-node expansion.

  Risks

  EUV or other equipment bottlenecks, export controls, and a downturn in the capex cycle.

• ABF substrate / SEMCO / Unimicron / Nanya PCB / Gold Circuit / Isu Petasys

  Substrate supply chain for AI servers and custom chips

  Strengths

  The report expects ABF substrate supply to become insufficient from 2027 onward; high-end ABF is used in AI accelerators such as Trainium/TPU, and network PCB also benefits from TPU diffusion.

  Weaknesses

  The predicted supply-demand gap depends on AI accelerator shipments and discipline in substrate capacity expansion.

  Comparison

  Compared with substrates used in traditional PCs and consumer electronics, a rising share of AI GPU, AI ASIC, and networking applications offers stronger growth potential.

  Risks

  New capacity exceeding expectations, slower AI server orders, and customer design changes.

• Wiwynn / Accton / King Slide / BizLink / Delta / SEMCO / Murata

  AI server assembly, networking, mechanical parts, connectors, power, and MLCC value chain

  Strengths

  AI server demand is expected to remain strong throughout 2026, while higher MLCC content and growing server assembly demand create structural opportunities.

  Weaknesses

  Hardware company margins may be squeezed by rising memory and component costs.

  Comparison

  The report advises caution toward hardware stocks with heavy memory leverage and high valuations, but still favors companies with rising AI server content and strong supply chain positioning.

  Risks

  AI server shipment timing, customer concentration, and inability to fully pass through component price increases.

• China AI chip ecosystem

  Rising self-sufficiency and localization opportunities in China AI chips

  Strengths

  The report expects China AI chip TAM to reach US$67bn by 2030, with self-sufficiency rising from 33% in 2024 to 76% by 2030e, benefiting semiconductor manufacturing localization and AI infrastructure spending.

  Weaknesses

  The report also warns that industry commoditization may emerge, suppressing long-term margins.

  Comparison

  Compared with dependence on overseas GPUs, the China market places greater emphasis on local substitution in foundry, GPU, and infrastructure.

  Risks

  Price competition, technology iteration, supply chain constraints, and policy and capex volatility.

Key data

• 2026e global semiconductor revenueabout US$1.6 trillionThe report says this is mainly driven by memory and logic, and is expected to generate record earnings.
• 2026e global semiconductor revenue YoYabout +96% YoYThis figure comes from the report's executive summary and is used to illustrate the strength of the semiconductor cycle.
• Edge AI semis CAGR22% CAGR, 2023-2030The report lists Edge AI semiconductors as one of the AI semiconductor growth subsegments.
• Inference AI semis CAGR68% CAGR, 2023-2030The report believes inference chips in cloud AI will grow faster than training chips.
• Custom AI semis CAGR65% CAGR, 2023-2030The report expects custom AI chips to grow faster than general-purpose chips.
• China AI chip TAMUS$67bn by 2030The report expects China’s total addressable market for AI chips to grow to US$67bn by 2030.
• China AI chip self-sufficiency rate33% in 2024, reaching 76% by 2030eThe report also warns that commoditization risk may emerge in the industry.
• DRAM bit growth25%+ in 2026-27, versus 16% in 2022-25The report says DRAM suppliers are preparing for stronger bit growth demand.
• ABF substrate supply-demandSupply shortage expected starting in 2027The report believes the era of AI diffusion will drive ABF substrate demand and lead to shortages after 2027.

Impact & implications

For portfolios, the report implies that 2026 technology allocation should not rely solely on a single AI leader, but should expand around full-stack AI infrastructure bottlenecks: GPU/ASIC, memory, foundry, SPE, ABF substrates, MLCC, server assembly, and thermal/power management may all benefit. At the same time, greater caution is warranted for highly valued hardware companies overly dependent on the memory cycle, chip design companies sensitive to consumer demand, and end companies with weaker ability to pass through costs.

Risks

• Demand destruction and worsening price elasticity for technology products in 2H26.
• Rising wafer, OSAT, and memory costs compressing margins for chip design companies and end hardware.
• Unclear AI investment returns, potentially leading to a rational pullback after an initial round of overinvestment and overconsumption.
• Power bottlenecks, financing constraints, and deployment limitations may slow the buildout of AI infrastructure.
• Excessively fast memory capacity expansion or peaking prices may weaken the earnings leverage of memory stocks.
• Hardware companies with high valuations and strong dependence on a single AI or memory theme face greater drawdown risk.
• Commoditization and price competition may emerge during the localization push in China AI chips.

What to watch

• Whether AI capex and cloud vendor orders continue to exceed expectations in the first half of 2026.
• Whether end demand in 2H26 shows clear demand destruction due to rising memory, wafer, and packaging/testing prices.
• Progress in adoption of Vera Rubin, AI ASICs, inference chips, and Agentic AI.
• The tightness of DRAM, HBM, and DDR5 supply-demand and the pace of new capacity additions.
• TSMC advanced nodes, HBM4/4e base die, and China AI GPU foundry opportunities.
• Whether ABF substrate enters shortage after 2027 as the report expects.
• China AI chip TAM, self-sufficiency, and the intensity of domestic semiconductor manufacturing investment.
• Higher content value in AI servers from MLCC, thermal management, power management, connectors, and server assembly.