Nomura Securities Research Report Analysis: SOX Up 85% Is Not the Peak, WoS Bottleneck Pushes AI Cycle to 2028
TechFlow Selected TechFlow Selected
Nomura Securities Research Report Analysis: SOX Up 85% Is Not the Peak, WoS Bottleneck Pushes AI Cycle to 2028
The deployment of computing power has been pushed to after 2027, fundamentally impacting the extension of the entire cycle.
Navigating Web3 tides with focused insightsWritten by: Rita
TechFlow Guide
The Philadelphia Semiconductor Index (SOX) has risen 85% cumulatively since March, but Nomura Securities clearly believes the peak of this cycle has not yet been reached. The most critical change is that the number of global data center construction projects has increased from 240 to 280, and GW-level projects have increased from 40 to 50. This indicates that computing power deployment has been pushed beyond 2027, fundamentally affecting the extension of the entire cycle.
Data Center Deployment Delays Extend Cycle
Nomura Securities' tracked deployment capacity forecast is: 2026 26.7GW (Gigawatts), 2027 32.3GW, 2028 22.9GW. This corresponds to an annual demand of 4-6 million AI chips. 2027 is the deployment peak, after which it begins to decline. This timeline indicates that the peak of capacity pressure has been pushed to 2027-2028.
The global server market growth forecast has been上调 from 43% to 74% in 2026 and 65% in 2027, and AI server growth has been上调 from 58% to 78% and 76%. This magnitude of upward adjustment is rare. If realized, 2026-2027 will be the two most tightest years for the entire server industry chain.
CPU Demand Severely Underestimated
The most easily overlooked point in the report is the growth in CPU demand. Arm targets the total addressable market for data center CPUs to exceed $100 billion by 2030. AMD estimated the server CPU TAM at $60 billion in November 2024, which doubled to $120 billion by May 2026, only 5 months later. NVIDIA and Qualcomm are each targeting $20 billion.
The ratio of CPU to GPU is undergoing a fundamental change. In inference scenarios, the ratio is from 3:1 to 4:1. But in agentic-type workloads, the ratio reverses to 1:4 (1 CPU to 4 GPUs). This means CPU demand far exceeds market expectations and is a key variable extending the cycle.
WoS Bottleneck is a Real Constraint
The most valuable insight in the report is the micro-analysis of CoWoS capacity. TSMC raised its CoWoS capacity target from 1,300-1,350kpcs to 2,000kpcs (2027), but Nomura Securities only modeled 1,800kpcs. This is an accurate estimate of the actual situation.
TSMC's wafer-level packaging (CoW) expansion is controlled by itself, but WoS expansion relies on outsourced semiconductor assembly and test (OSAT) vendors such as ASE Technology and SPIL, with capacity adjustment speeds significantly lagging. Starting from the second half of 2026, WoS capacity will become a constraint on the entire CoWoS output. TSMC's capacity targets have historically not been met, and actual capacity is closer to 1,800kpcs.
Intel's EMIB-T (Embedded Multi-die Interconnect Bridge) technology impacts TSMC. Google v9 Tensor Processors may adopt Through-Silicon Via (TSV) solutions in 2028, threatening TSMC's advanced packaging pricing power.
Nine Target Prices Raised 15-50%
Nomura Securities raised target prices for nine AI semiconductor companies. Among them, MediaTek raised 49.5% to NT$5,800, TSMC raised 46.4% to NT$3,425, Realtek raised 33.9% to NT$2,115, NXP raised 30.9% to NT$6,880, GlobalWafers raised 28.2% to NT$1,200, KYEC raised 26.6% to NT$390, Bihe Technology raised 24.1% to NT$720, Kinwong Integrated raised 22.3% to NT$19,100, ASE Technology raised 15.5% to NT$730.
TechFlow Perspective
Nomura Securities' micro-insight on the WoS bottleneck is the core value of this report. But investors need to be alert: TSMC's capacity targets have historically not been met, if OSAT expansion is slower than expected, 2027 CoWoS actual output may be lower than 1,800kpcs.
Nomura Securities already adjusted TSMC's rating in the May AI report, and raised the target price again only three months later; this high frequency of change is worth alerting to. However, Nomura Securities' tracking accuracy has always been relatively high.
Pay attention to the progress of "non-NVIDIA" routes such as Intel EMIB-T, Google TPU, and Amazon Trainium. These alternative solutions may change the entire supply chain landscape.
From an investment perspective, the top of the AI cycle is extended at least to 2028, and any pullback during this period is worth buying. But at the same time, realize that capacity constraints are evolving from the supply side to the pricing side, which may make chip manufacturers' profit margins lower than expected.
Disclaimer
This article is a compilation and interpretation by TechFlow Research of third-party brokerage research reports. The ratings, target prices, earnings forecasts, and related judgments cited in the text are the views of the brokerage analysts, represent only the position of their affiliated institutions, do not represent the views of TechFlow Research, and do not constitute any investment advice.
The market carries risks; decisions must be made independently. This article should not be used as a basis for buying or selling any securities.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
@chaoxiangooo
