The Terafab Math

研报英文原文证据摘录

The Terafab Math

TD SECURITIES (USA) LLC INDUSTRY UPDATE

June 9, 2026

■Semiconductor Capital Equipment The Terafab Math

Krish Sankar THE TD COWEN INSIGHT

415 646 7372

While it might be a bit early to evaluate Terafab, the surrounding excitement against a

krish.sankar@tdsecurities.com

backdrop of tight semiconductor capacity, is worthy enough of framing its implications to

Eddy Orabi WFE spend and for the ceiling this cycle. We estimate the initial buildout could support ~ $20B

415 646 7371 of WFE over 2028–29. More importantly, it may incentivize other customers to place orders

eddy.orabi@tdsecurities.com

earlier — especially in lithography.

Steven Chin

415 646 7374 SpaceX’/TSLA's goal is to build a semiconductor facility spanning logic, memory, and

steven.chin@tdsecurities.com packaging, ultimately capable of producing up to 1 TW per year of AI compute (~50x this

Robert Mertens, CFA year's industry deployment). That is obviously far beyond anything the industry is deploying

646 562 1338 today, which is why we think the right way to look at it is as a ceiling thought experiment, not a

robert.mertens@tdsecurities.com base-case forecast. Based on our 1GW framework, we estimate that 1TW would be equivalent

to roughly 6M NVL72s of Vera Rubin, requiring ~ 3.5M wspm of front-end wafer capacity for

GPUs/CPUs and ~10M wspm of HBM. We estimate that meeting that goal would require ~$1.5T

of WFE.

ASML capacity is key to watch. Our existing work already suggests that ASML would require

90+ EUV tools to support $220B+ of WFE in 2028, which is why we think ASML capacity

commentary (guided for +80 EUV units in 2027) and any updates around High-NA are the most

important indicators from here (Terafab uses Intel 14A, which uses High NA). More broadly, our