Goldman Sachs Raises Global Wafer Fab Equipment Spending Forecast, Expected to Reach $281 Billion by 2028
nashnova research
Goldman Sachs lifted its 2026–2028 global wafer-fab equipment spending forecasts to $150 billion, $218 billion, and $281 billion, signaling that the semiconductor equipment upcycle may run two years longer than the market assumed.
Why did Goldman raise forecasts so sharply?
Two triggers: Q2 earnings revealed broadly higher-than-expected semiconductor capex, and equipment suppliers guided more optimistically than before.
This means → the upgrade is not a single-company bet — order visibility is improving across the supply chain.
Goldman maintains its overall bullish stance on semiconductor equipment, projecting the upcycle to extend through 2028.
Foundry: how much does TSMC alone account for?
Foundry WFE forecasts for 2026–2028 were raised to $58 billion, $84 billion, and $109 billion, with year-on-year growth jumping to 45% / 45% / 30% from the prior 30% / 30% / 16%.
The main driver is TSMC — Goldman raised its annual capex estimate by $8 billion per year, citing stronger-than-expected equipment intensity for the N2 node — TSMC's next-generation 2 nm process — and sustained customer demand.
In plain terms = TSMC's expansion plan alone lifted the entire foundry equipment market's growth rate by double digits.
DRAM: why does the HBM4 migration keep capacity tight?
DRAM WFE forecasts were raised to $48 billion, $72 billion, and $97 billion; the 2028 growth rate jumped from 10% to 35%.
Analyst Giuni Lee raised Samsung's average DRAM capex forecast by 22% and SK Hynix's by 19%.
This means → even with sharply higher spending, capacity will stay tight through 2028. The migration to HBM4 — fourth-generation high-bandwidth memory, used mainly for AI training — itself consumes significant capacity, so more spending does not equal looser supply.
NAND and logic: which is the real surprise?
NAND adjustments were modest: the 2027 forecast was actually cut to $15 billion from $17 billion, while 2028 was raised to $22 billion. Near-term spending focuses on upgrades, not greenfield expansion.
The real surprise is logic and other: the 2027 forecast jumped from $35 billion to $47 billion (+34%), and 2028 from $37 billion to $53 billion (+43%).
Two drivers: Intel demand came in above expectations, and SpaceX and Tesla committed roughly $16.8 billion to the initial phase of the Terafab project — a mega-scale wafer-fab initiative now folded into the logic forecast.
Can the upcycle really last until 2028?
Goldman's core thesis: capex beats have spread from a single company to multiple sub-segments — this is no longer a one-name story.
Three variables will decide whether the thesis holds: TSMC's N2 ramp timeline, the pace of HBM4 migration, and the actual execution of Terafab.
In plain terms = Goldman has drawn an upward curve, but whether it completes depends on these three milestones landing on schedule.
市场有风险，内容仅供研究参考，不构成投资建议。