Broadcom Inc is currently capitalizing on a significant competitive advantage within the rapidly evolving field of artificial intelligence (AI) chip technology. According to JPMorgan analyst Harlan Sur, the company holds an 18-month lead over chip programs that are internally designed and managed by customers, a factor that is accelerating its AI Application-Specific Integrated Circuit (ASIC) outlook through fiscal years 2026 and 2027.
The key driver behind this momentum is Broadcom’s collaboration with Alphabet Inc’s Google. Sur highlights that Broadcom's next-generation 3nm Tensor Processing Unit (TPU), codenamed Sunfish, has already completed evaluation and qualification phases with Google, establishing a timing advantage over Google's own internal chip development efforts. In contrast, Google's internal 3nm TPU project, known as Zebrafish, has encountered repeated delays and may only reach first silicon production toward the end of the current quarter or early next quarter.
Sur points out that customer-owned tooling in chip design generally involves customers directing approximately 90% of the design work themselves. This approach, while widely used, typically cannot match the efficiency and speed of Broadcom's dedicated ASIC development programs. The Sunfish TPU represents a tailored solution that supports Google’s objective to scale AI workloads more effectively.
Market expectations according to Sur indicate that Google aims for substantial deployment of TPU processors, increasing its target from an earlier estimate of 5 million units to a range between 6 million and 7 million TPU processors by calendar year 2027. Significantly, most of these shipments are anticipated to serve external customers, utilizing Broadcom’s hardware contributions. Furthermore, even for Google's internal AI demands during the same period, the bulk of TPU utilization is expected to rely on Broadcom's Sunfish ASIC, underscoring Broadcom's dominant role in Google's AI infrastructure.
Broadcom’s production schedule features a ramp-up of its Sunfish chipset starting in the second half of this year, backed by multi-billion dollar purchase orders. Sur notes that the AI backlog associated with Broadcom’s chipsets has grown by several billions since the company’s December earnings announcement. This increase spans both current-generation Ironwood chips and the next-gen Sunfish units, reflecting strong demand and confidence in Broadcom’s AI product stack.
The analyst further elaborates on potential supply chain constraints that could impact the broader AI ASIC landscape. He emphasizes the critical importance of early qualification of wafers and advanced packaging technologies, such as Chip-on-Wafer-on-Substrate (CoWoS), High Bandwidth Memory (HBM), and advanced substrates. These components are essential for successful high-volume production but are subject to tightening lead times in the market. This factor is likely to curtail the volume potential for programs that start later in the development cycle, exemplified by Google's delayed Zebrafish initiative.
Looking ahead, Broadcom is already engaged in designing Google’s subsequent TPU generation utilizing a 2nm process node, targeted for volume manufacturing in calendar 2028. This ongoing innovation pipeline underlines Broadcom’s commitment to maintaining technological leadership and long-term partnership strength with Google.
Financially, Sur recommends purchasing Broadcom shares aggressively at current market prices. His projections include first-quarter revenue of $19.14 billion along with adjusted earnings per share (EPS) of $2.02. At the time of publication, Broadcom’s stock trades near $320, reflecting a notable increase consistent with positive investor sentiment.
Key Points:
- Broadcom holds an 18-month lead in custom AI chip development versus customer-led internal programs, reinforcing its market position through 2026 and 2027.
- The partnership with Alphabet’s Google is central to Broadcom’s AI ASIC growth, particularly with the deployment of the 3nm Sunfish TPU expected to support millions of units by 2027.
- Broadcom's multi-billion dollar purchase orders underpin a robust backlog covering current and next-generation AI chip shipments, indicating strong demand momentum.
- Tightening supply conditions for wafers, advanced packaging, and memory components highlight the importance of early qualification and may limit late-starting chip initiatives’ production volumes.
Risks and Uncertainties:
- Potential delays in Google’s internal 3nm TPU development program, Zebrafish, which has experienced repeated delays, could impact timing and volume expectations.
- Supply chain constraints related to wafers, CoWoS packaging, High Bandwidth Memory, and substrates could restrict scaling opportunities for Broadcom and competitors alike.
- Reliance on the successful qualification and volume ramp of the Sunfish 3nm TPU is critical; any setbacks in this process could affect Broadcom's projected AI chipset leadership.
Disclosure: This analysis is based solely on publicly available information and does not constitute investment advice. Investors should perform their own due diligence before making any investment decisions.