Broadcom's Custom Chips Outpace Nvidia in AI Growth
Broadcom reported 221% year-over-year growth in AI semiconductor revenue, challenging the dominance of general-purpose GPUs in the next phase of infrastructure buildout.
Broadcom (NASDAQ:AVGO) has demonstrated that custom silicon is capturing a growing share of the artificial intelligence market. The company reported AI semiconductor revenue of $16.7 billion for the third quarter of fiscal 2026, marking a 221% increase compared to the same period last year. This performance positions Broadcom as a significant alternative to Nvidia, with management guiding fourth-quarter AI revenue to $21.7 billion, a 236% year-over-year jump.
The shift reflects a structural change in how hyperscalers approach compute requirements. While Nvidia continues to lead in general-purpose training workloads, Broadcom’s application-specific integrated circuits are gaining traction for inference tasks. CEO Hock Tan outlined a long-term trajectory, projecting AI revenue of approximately $115 billion in fiscal 2027 and $230 billion in fiscal 2028, indicating sustained demand from clients like Google and Meta.
Custom Silicon Targets Stable Inference Workloads
The economic advantage of Broadcom’s chips lies in their optimization for specific large language model workloads. Tan stated that co-developed chips outperform general-purpose GPUs in cost and power efficiency when the workload is stable. In the latest quarter, these custom accelerators accounted for 73% of Broadcom’s total AI revenue, highlighting the preference of major cloud providers for specialized hardware over generic solutions.
This strategy contrasts with Nvidia’s approach, which remains dominant in frontier model training where architectures shift frequently. However, for inference of shipped models at hyperscale, the stable nature of the workload favors the efficiency of custom silicon. Broadcom’s ability to deliver lower cost per token is becoming a key differentiator in capital expenditure decisions among major technology firms.
Supply Chain Partners Benefit From Volume Growth
Taiwan Semiconductor (NYSE:TSM) stands to benefit from this trend as the primary fabricator for Broadcom’s custom chips. The company reported August 2026 revenue up 53.3% year over year, driven by strong demand for advanced node semiconductors. Despite a forward P/E of 20, the growth trajectory remains robust, though investors must weigh the risks of geographic concentration in manufacturing.
Dell Technologies (NYSE:DELL) also reflects the broader infrastructure buildout, booking a record $60.9 billion in AI orders last quarter. With a backlog of $95 billion and a raised full-year guide of $192 billion, the server market is expanding rapidly. However, hardware margins remain thinner than those of chip designers, with Dell’s infrastructure solutions group operating at a 15% margin.
Market Consensus Splits on AI Leaders
Opinions on the next phase of AI leadership remain divided among market participants. Some argue for sticking with established leaders like Nvidia and Microsoft, citing their scale and ecosystem strength. Others, including podcast hosts Troy Millings and Rashaad Bilal, debate whether the value accrues to specialized players like Broadcom or to the broadest platforms. The data suggests the market is large enough to support multiple winners.
Broadcom’s financial results provide a concrete basis for the case for custom silicon. The company’s ability to secure multi-year commitments from hyperscalers ensures visibility into future revenue. As AI infrastructure continues to expand, the competitive dynamic between general-purpose and application-specific chips will define the profitability of the sector.
