FilippoBacci
Goldman Sachs began coverage of several U.S Digital Semiconductor and Electronic Design Automation Software sector companies and noted that they are most constructive on merchant (and select custom) silicon and EDA vendors tied to the most sustainable elements of AI-related capital expenditure.
Buy ratings:
The firm has started coverage of Nvidia (NASDAQ:NVDA) with a $185 price target and Broadcom (NASDAQ:AVGO) with a price target of $315. In addition, Cadence Design Systems with a $380 price target and Synopsys (NASDAQ:SNPS) with a $620 price target.
Neutral ratings:
Goldman began coverage of Advanced Micro Devices (NASDAQ:AMD) with a $140 price target, British chip architecture designer Arm (NASDAQ:ARM) with a price objective of $160 and Marvell Technology (NASDAQ:MRVL) with a $75 price target.
Analysts led by James Schneider said they believe the AI investment cycle is in a state of transition, with over $350bn in capital expenditure spent on AI infrastructure. They added that although monetization has been elusive, they see early signs of incremental revenue and much clearer evidence of cost takeout to justify these investments, and believe AI CapEx can sustain growth from current levels.
The analysts noted that the magnitude of the AI infrastructure opportunity has resulted in a rapid re-ordering of semiconductor markets and technology leadership.
“We believe the balance between leading-edge AI model training and lower-cost inference should drive a “barbell” approach to semiconductor investing: (1) performance and software ecosystem leaders; (2) low-cost leaders enabling greater affordability,” the analysts added.
Schneider and his team said that leading-edge model training favors performance and ecosystem winners, plus scale leaders in custom silicon. The analysts believe AI remains at an early stage of development, which should translate to a premium for performance and software ecosystem breadth. Commoditization remains some time away, and they think the custom silicon market — although rapidly ramping — will remain a minority for now.
More customers means more design, more designers, greater complexity, and more EDA. The analysts added that as the industry has moved from the era of client/server to cloud computing to generative AI, the number of systems and chip designers and customers has rapidly increased — along with chip design complexity and now multi-chip systems — thereby heightening the importance of EDA Software providers.
In addition, Schneider and his team said that geopolitical risks and opportunities remain — from China and beyond. The analysts noted that geopolitical risks tied to technology remain central for the semiconductor industry — and they quantify the risks within. At the same time, a new rangeof opportunities tied to sovereign infrastructure needs has arisen, which can serve as a partial offset.