AI data-center demand is broadening, while memory, SPE, and semiconductor materials remain in an upcycle

Overview

This report is Nomura's June 2026 technology monthly, covering Japan electronics, semiconductors, memory, SPE, electronic components, consumer electronics, semiconductor and display materials, and IT service opportunities related to AI agent deployment. The core conclusion is that AI models are advancing faster than system implementation, but major IT companies, private equity funds, and system integrators are driving AI agent deployment through FDE capabilities; meanwhile, rapid growth in shipments of AI servers and networking equipment is spreading supply-demand tightness from GPUs and HBM to ASICs, SSDs, CPUs, analog chips, and MLCCs. Overall, the report is optimistic about the structural opportunities created by AI demand, while also warning that price-sensitive markets such as PCs and smartphones may be hit by CPU and memory shortages as well as price increases.

Core views

Nomura believes AI should not be viewed as a threat to Japan's large IT companies, but rather as an important growth opportunity. As AI agents penetrate B2B scenarios at a faster pace, companies with AI platforms, system architecture capabilities, and industry know-how are expected to increase service revenue. In memory, DRAM and NAND prices remain supported by structural supply shortages, and short-term share-price pullbacks may instead offer buying opportunities, though the report also expects upward momentum in commodity memory prices to slow materially in 2H26. In SPE, front-end equipment demand still has upside in 2026 and 2027, with improving investment in advanced logic and advanced memory enhancing demand visibility. In materials, demand for wafers, photoresists, packaging substrate materials, and packaging materials is strong, and price hikes in some materials could improve margins. In consumer electronics, diversified companies such as Sony Group, Fujifilm Holdings, and Canon remain favored, but price-sensitive end markets such as PCs and smartphones require caution over cost pass-through pressure.

Analysis framework

The report combines a top-down industry-cycle view with bottom-up company tracking: it first identifies the demand theme through AI agents, data-center investment, and the performance of major chip companies, then uses WSTS/SIA, METI, SEAJ, MOF, monthly sales of Taiwanese electronics companies, TSMC monthly sales, memory prices, and company results to validate supply-demand, pricing, and inventory trends, and finally maps these into investment implications for sectors such as Japan electronics, memory, SPE, materials, consumer electronics, and IT services.

Methodology notes

• Industry cycleSupply-demand cycle analysis

  Determine the semiconductor cycle position through prices, inventories, shipment value, and lead times.

  The report uses rising memory prices, changes in inventory turnover, WSTS shipment value, SPE exports, and materials prices as evidence of supply-demand tightness, and distinguishes AI data-center demand from traditional end demand such as PCs and smartphones.

• Thematic investingAI agent deployment chain

  AI agent implementation drives demand for servers, networks, chips, memory, components, and IT services.

  The report believes FDE capabilities help accelerate the deployment of AI agents in customer systems, thereby extending the AI infrastructure investment cycle and creating growth opportunities for Japanese IT services and industrial electronics companies.

• Supply-chain mappingDiffusion from AI training to inference and then to the agent stage

  Demand is first concentrated in GPUs and HBM, then spreads to ASICs, SSDs, CPUs, analog chips, and MLCCs.

  The report emphasizes that AI demand is moving from the training stage into inference and agent stages, so the beneficiaries of supply-demand tightness are expanding from a few core computing components to a broader range of semiconductors, materials, and electronic components.

Asset mapping & comparison

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

• Memory and AI-related semiconductors

  Directly benefit from the spread of demand across AI servers, inference, and the AI agent stage.

  Strengths

  Demand for DRAM, HBM, NAND, and SSDs is being driven by AI, improving price and contract visibility, while structural supply shortages persist.

  Weaknesses

  Margins for commodity DRAM and NAND are already at historical highs, and upward price momentum may slow in 2H26.

  Comparison

  Compared with traditional PC and smartphone demand, AI server customers are more willing to accept price hikes to secure procurement volumes.

  Risks

  Customer price pressure, expectations of government intervention, changes in Nvidia Vera Rubin configurations, and subsequent supply releases could all trigger share-price volatility.

• Japanese IT services and industrial electronics

  Participate in AI agent implementation through FDE, AI platforms, system architecture, and industry knowledge.

  Strengths

  Companies such as Hitachi, Fujitsu, and NEC are expected to benefit from AI agent deployment, physical AI, defense, and infrastructure demand.

  Weaknesses

  AI model progress is faster than customer system deployment, and implementation speed still depends on FDE capability and the pace of customer IT transformation.

  Comparison

  Compared with pure model suppliers, the strengths of Japan's large IT companies lie in industry expertise, system integration, and on-site customer deployment.

  Risks

  Slower-than-expected commercialization of AI agents, shrinking enterprise IT budgets, or longer project delivery cycles.

• Electronic components

  AI data centers and high-end industrial applications are driving demand for MLCCs, BBUs, probes, EMIB, and electrostatic chucks.

  Strengths

  Sales and profits of 10 major Japanese electronic component suppliers exceeded expectations, with strong demand from data centers, general industry, high-end autos, smartphones, and wearables.

  Weaknesses

  Some consumer-side and price-sensitive applications may still be affected by rising costs.

  Comparison

  New AI-related products offer stronger growth and profit contribution than traditional consumer electronics components.

  Risks

  Yen volatility, customer inventory adjustments, weaker end demand, and supply-chain bottlenecks.

• Consumer electronics

  Costs are affected by rising memory prices and the Middle East situation, but some companies can remain resilient through price pass-through, a weak yen, and portfolio optimization.

  Strengths

  Sony Group, Fujifilm Holdings, and Canon are recommended, corresponding respectively to entertainment platforms, portfolio transformation driven by proprietary technology, and stable cash flow from printers and cameras.

  Weaknesses

  Price-sensitive markets such as PCs and smartphones may face pressure from memory and CPU shortages.

  Comparison

  Diversified companies need to be assessed case by case by business mix rather than judged simply as a broad consumer-electronics group.

  Risks

  Inability to pass through costs, contraction in end demand, exchange-rate reversal, and geopolitical shocks.

• SPE

  Advanced logic, advanced memory, and long-term contracts support demand for front-end equipment.

  Strengths

  Sales and exports of Japan-made SPE have improved, front-end equipment demand is rising in 2026-2027, and Tokyo Electron has confidence in passing through price increases.

  Weaknesses

  Back-end SPE is improving, but market preference remains concentrated in front-end, and the period of declining exports to China has lasted longer than expected.

  Comparison

  Demand improvement is more evident in Taiwan and South Korea, while China demand may skew toward 2H26.

  Risks

  Further delays in China demand or demand coming in below expectations, changes in the pace of advanced-process investment, and volatility in equipment delivery timing.

• Semiconductor and display materials

  AI demand is spreading to advanced logic, HBM, ASICs, CPUs, NAND, analog, and power semiconductors, driving materials demand.

  Strengths

  Wafer and photoresist exports are close to historical highs, and price hikes for some materials may lift margins in 2H26.

  Weaknesses

  Demand for small-sized display materials is weak, and smartphone and notebook production may decline year over year.

  Comparison

  Semiconductor materials benefit more directly from structural AI growth than small-sized display materials, while large-sized display materials are relatively stable.

  Risks

  Whether material price hikes can be passed through smoothly, long-term contract renewal pricing, the pace of downstream capacity expansion, and weak consumer end demand.

Key data

• Global semiconductor shipment valueApril 2026: +106.3% YoY; three-month moving average: +94.0% YoYSIA/WSTS data show AI and data-center demand remains strong, while memory price increases also lifted shipment value.
• Japan semiconductor shipmentsApril 2026: +8.1% YoY, +3.2% MoMMETI data show production +3.6% YoY, inventories -2.3% YoY, and inventory ratio +8.9% YoY.
• Sales of Taiwanese electronics companiesMay 2026: +42.2% YoY, +0.7% MoMThe sample covers 265 companies; Hon Hai Precision Industry posted sales of TWD859bn, up 39.6% YoY and 3.3% MoM.
• TSMC monthly salesMay 2026: TWD417bn, +30% YoY, +2% MoMThis marked the 29th consecutive month of year-over-year growth, reaching 66% of the midpoint of 2Q26 sales guidance.
• DDR5 DRAM pricesEnd-May spot price: $26.3/GB, +3.1% MoM; contract price: $12.8/GB, +2.0% MoMThe spot price carries about a 105% premium to the contract price, reflecting still-tight supply-demand conditions.
• 512Gb TLC NAND priceMay spot price: $0.36/GB, flat MoMPrices remain elevated, and Adata's monthly data imply recent NAND price gains may be faster than those of DRAM.
• Japanese electronic component suppliersJan-Mar 2026 aggregate sales: ¥2,862.6bn, +12.8% YoY; operating profit: ¥269.8bn, +66.2% YoYBoth sales and profit exceeded Nomura's pre-release estimates, supported by strong demand from data centers and general industrial applications.
• Sales of Japan-made SPEApril 2026 three-month moving average: ¥510.2bn, +14% YoY, +6% MoMSEAJ data reflect improving SPE demand, while MOF statistics show April SPE exports of ¥236.2bn, up 6% YoY.
• Japan semiconductor inventory turnoverAnalog 39.9 days, logic 35.9 days, memory 27.2 days, discrete 10.3 daysApril 2026 data; memory turnover days rose from 19.5 days in March.
• Semiconductor material price hikesCopper-clad laminate material prices up at least 30%, packaging material prices up about 10-20%The report expects margins for materials companies to improve in 2H26 on price hikes and strong sales, while wafer contract prices may have room to rise about 20% upon renewal in 2027-2029.

Impact & implications

From an investment perspective, AI data centers remain the most important demand anchor in the technology chain. Tight supply and rising prices are positive drivers for memory, front-end SPE, advanced-packaging-related components, semiconductor materials, and Japanese IT companies with AI platform capabilities; however, they may also lead to higher costs, procurement difficulties, and demand contraction for price-sensitive end markets such as PCs, smartphones, and home appliances. The report prefers companies that can benefit from structural AI demand, possess pricing power, or are improving their business mix, while also recommending continued monitoring of delayed China SPE demand, slower pricing momentum after high memory profitability, and spillover effects of CPU and memory shortages on traditional end markets.

Risks

• Tight memory and CPU supply may suppress demand in price-sensitive end markets such as PCs, smartphones, and home appliances.
• Margins for commodity DRAM and NAND are already at historical highs, and price increase momentum may slow in 2H26.
• China SPE demand and equipment deliveries may continue to be delayed, and export data remain uncertain.
• Geopolitical risks and the Middle East situation may increase macro uncertainty and supply-chain costs.
• If AI agent deployment is slower than expected, data-center investment and demand for related components may fall short of market expectations.
• If price increases for materials and components cannot be passed through downstream, customer margins may be squeezed and demand may decline.

What to watch

• WSTS/SIA monthly semiconductor shipment value and growth in AI-related categories.
• Changes in spot and contract prices for DDR5 DRAM, HBM, NAND, and SSDs.
• Demand guidance related to AI accelerators such as Nvidia Vera Rubin, Broadcom, and Marvell.
• Monthly sales of Taiwanese electronics companies, especially Hon Hai Precision Industry, Quanta Computer, Wiwynn, ASPEED, and TSMC.
• Japan-made SPE sales and exports released by SEAJ and MOF, as well as demand from China, Taiwan, and South Korea.
• METI data on Japan semiconductor shipments, inventories, and inventory turnover days.
• Implementation of semiconductor material price hikes and wafer long-term contract renewal prices for 2027-2029.
• Expansion of FDE teams and progress in AI agent commercialization at Japanese IT companies such as Hitachi and Fujitsu.