Broadcom Forecasts AI Chip Revenue to Top $100 Billion by 2027
The semiconductor leader expects a surge in custom AI accelerators and networking hardware as hyperscalers scale LLM infrastructure.
Broadcom is forecasting a massive surge in its AI semiconductor revenue over the next few years. The company expects its AI-driven chip sales to grow exponentially as the industry shifts toward specialized hardware for large-scale artificial intelligence.
According to CEO Hock Tan, Broadcom projects that AI chip revenue will significantly exceed $100 billion by 2027. This growth is primarily driven by an escalating demand for custom AI accelerators, known as Application-Specific Integrated Circuits (ASICs), and high-performance networking hardware. Some reports suggest this trajectory could push revenues as high as $230 billion by 2028, representing a multi-fold increase from current levels.
The Infrastructure Backbone
Broadcom occupies a critical position in the AI supply chain, providing the essential plumbing that allows AI clusters to function. While GPUs handle the primary computation, Broadcom’s custom ASICs and high-speed networking chips—such as the Tomahawk and Jericho lines—are vital for connecting thousands of GPUs into a cohesive system. This networking capability is necessary to prevent data bottlenecks during the training of massive large language models (LLMs).
Shifting Market Dynamics
A revenue target of over $100 billion signals a pivotal shift in the semiconductor market. For the past several years, the AI boom has been dominated by general-purpose GPUs. However, Broadcom's projections suggest that the market for custom silicon is expanding rapidly. As hyperscalers seek more energy-efficient and cost-effective ways to run specific AI workloads, they are increasingly turning to custom-designed chips tailored to their unique software stacks rather than relying solely on off-the-shelf hardware.
Future Outlook
Industry observers will be watching whether Broadcom can maintain its lead in the ASIC space as more cloud providers attempt to bring chip design entirely in-house. The company's ability to hit these multi-billion dollar targets depends on the continued build-out of AI data centers and the sustained appetite for LLM scaling. While the growth trajectory is steep, the company's role as a primary beneficiary of the infrastructure build-out positions it as a central pillar of the AI hardware ecosystem.