Goldman Sachs Unveils Five Key Debates for Semiconductors: AI Capex, Equipment Cycles, and More
TREE NEWS reports: As hyperscale cloud providers continue to ramp up capital expenditures and AI computing demand expands from training to inference and Agentic AI, investors are increasingly focused on the durability of the AI infrastructure cycle and which segments will continue to benefit. In a preview of its upcoming Communacopia & Technology Conference (September 8–11 in San Francisco), Goldman Sachs has outlined five critical debates that will shape the semiconductor industry’s near-term outlook.
What Happened
Goldman Sachs analysts have identified five key debates ahead of the conference, which will feature 32 global companies spanning digital/AI chips, EDA software, analog semiconductors, semiconductor equipment, memory, and IT services. The debates center on: (1) AI computing demand and competitive dynamics, (2) the sustainability of the wafer fab equipment (WFE) upcycle through 2028, (3) memory and SSD supply-demand balance and valuation re-rating, (4) the persistence of the analog semiconductor recovery, and (5) the potential for Agentic AI to create a new growth curve for EDA software.
Market Implications
AI Compute Demand: Goldman expects overall optimism on AI demand, with hyperscaler capex remaining robust and Agentic AI emerging as a new catalyst. While merchant silicon (GPUs) will dominate near-term, ASICs and custom accelerators are expected to gain share as cost and power efficiency become key differentiators. Nvidia, Broadcom, and AMD are likely to discuss their respective product cycles and competitive positions.
Semiconductor Equipment: The WFE upcycle is projected to last at least through 2028, driven by DRAM, leading-edge logic/foundry, and advanced packaging. Non-traditional customers like Terafab and a re-accelerating Intel investment cycle could add further demand. Deposition and etch remain key focus areas, with advanced packaging and inspection/metrology also benefiting.
Memory & SSD: Goldman’s supply-demand model indicates DRAM and NAND shortages persisting through 2028, with gaps of 5.0%/5.9%/3.9% for DRAM and 4.4%/4.6%/3.0% for NAND in 2026-2028. This could support pricing and drive significant free cash flow, with companies potentially returning 50-100% of excess FCF to shareholders. HAMR and HBF technologies are also on the radar.
Analog Semiconductors: The recovery may be more durable than expected. Shipments are already ~5% above the long-term trend, but four years of under-shipment imply ample catch-up demand. Combined with improving auto end-markets and new AI data center demand, the cycle could extend further.
EDA Software: Agentic AI could automate design, verification, and optimization processes, addressing engineer shortages and creating a $3.7 billion annual incremental market by 2030. This opportunity is not yet reflected in sell-side estimates and may start to materialize in late 2026.
Key Takeaways for Investors
- AI capital expenditure remains the core theme, with resilience expected despite concerns about data center bottlenecks.
- Custom AI chips and Agentic AI are opening new incremental growth areas beyond traditional GPU scaling.
- Memory and equipment stocks could see sustained demand and potential re-rating if supply shortages persist.
- Analog and EDA names may offer diversified exposure to the AI cycle, with EDA presenting an underappreciated growth optionality.
- Investors should monitor conference commentary for updates on product roadmaps and capex guidance.
