JPMorgan analyst Harlan Sur's recent assessment underscores Broadcom's considerable advantage in the custom artificial intelligence (AI) chip sector. He highlights the company's 18-month lead in this specialized field, projecting an accelerated outlook for its AI Application-Specific Integrated Circuit (ASIC) business through fiscal years 2026 and 2027. This growth is primarily fueled by a pivotal partnership with Alphabet Inc.'s Google.
Broadcom's Strategic Edge in AI Chip Development
In a detailed analysis, JPMorgan's Harlan Sur maintained an "Overweight" rating on Broadcom, affirming a robust price target of $475. Sur elaborated on Broadcom's significant head start in the development of custom AI chips, noting that the company surpasses internal customer-led tooling initiatives by over 18 months, where clients typically manage about 90% of the design tasks. He cited Google's 3nm Zebrafish program as an illustrative case, which has encountered recurring delays and is not expected to yield initial silicon until the close of the current quarter or early next quarter.
In stark contrast, Broadcom's advanced 3nm Tensor Processing Unit (TPU), codenamed Sunfish, underwent earlier evaluation and qualification by Google. This provided Broadcom with a crucial timing advantage. Sur indicated that Google is experiencing enhanced success in supporting external AI workloads through its TPU platform. Consequently, Google now aims to deploy between 6 million and 7 million TPU processors by the calendar year 2027, an increase from previous estimates of 5 million units. The analyst anticipates that the majority of these 2027 shipments will cater to external clients, with all deployments exclusively powered by Broadcom and Google's Sunfish 3nm TPU ASIC. Furthermore, even Google's internal TPU utilization in 2027 is projected to predominantly rely on Sunfish, meaning over 95% of Google's 6 million to 7 million TPU units in 2027 will be equipped with Broadcom silicon.
Sur also pointed out that Broadcom has already secured multi-billion dollar purchase orders tied to the Sunfish production ramp, scheduled to commence in the latter half of the current year. He believes that Broadcom's AI backlog has continuously expanded by billions since the December earnings call, encompassing both current-generation Ironwood and next-generation Sunfish shipments. The analyst emphasized that the tightening lead times for wafers, CoWoS, HBM, and advanced substrates render early qualification indispensable. This dynamic is expected to restrict the volume potential for programs that begin later, such as Zebrafish. Broadcom has already initiated the design of Google's subsequent TPU ASIC on a 2nm process, with volume production targeted for calendar year 2028. Given these insights, Sur advocated for aggressive purchasing of Broadcom shares at current valuation levels. He forecasts first-quarter revenue of $19.14 billion and adjusted earnings per share of $2.02. On a recent Monday, Broadcom shares saw a 1.93% increase, trading at $320.10.
This analysis by JPMorgan highlights the intense competition and rapid innovation within the AI chip industry. Broadcom's strategic foresight and technological prowess in securing long-term partnerships, especially with a tech giant like Google, position it favorably in this evolving landscape. The projected increase in TPU deployments and Broadcom's involvement in future generations of AI ASICs suggest a significant impact on the semiconductor market. Investors paying close attention to this sector might consider Broadcom's strategic advantages and robust financial outlook as a compelling opportunity.