Under a newly signed agreement, the two technology giants will deepen their collaboration on ASICs (Application-Specific Integrated Circuits). According to a filing submitted by Broadcom on Monday local time, these ASICs will be used across multiple generations of Apple products. ASICs are silicon chips designed for specific purposes and are becoming increasingly important for developing AI chip components and clusters used in artificial intelligence training and inference. Apple is currently developing more advanced AI chips and complete AI servers for AI training and inference workloads.

Broadcom is one of Apple's core chip suppliers, as well as a key provider of high-performance Ethernet switch chips, high-speed optical interconnect components for large-scale AI data centers, and customized AI chips for major cloud computing companies.

Through its involvement in Google TPU, OpenAI, Anthropic, and other hyperscalers' in-house AI ASIC accelerator initiatives, Broadcom has positioned itself at the center of the largest incremental market beyond Nvidia GPUs—the custom AI chip market. AI ASICs have undoubtedly become the primary pillar driving Broadcom's higher valuation and long-term earnings growth. The strong earnings recently reported by both Broadcom and Marvell provide financial evidence that the unprecedented growth trajectory of AI ASICs is rapidly being validated.

Broadcom expects AI chip revenue to exceed US$100 billion by 2027, while AI semiconductor revenue for Q2 FY2026 is projected to reach US$10.8 billion, representing 143% year-over-year growth. This indicates that demand for AI ASICs and XPUs has become the company's core growth engine.

On another front, regardless of whether AI computing follows the GPU, TPU, or ASIC route, all architectures rely on high-speed Ethernet switching and optical interconnects. Broadcom's Tomahawk 6 has already achieved 102.4 Tbps switching capacity on a single chip. At OFC 2026, the company also showcased AI networking solutions including 102.4T Ethernet, Co-Packaged Optics (CPO), 400G/lane optical DSPs, and PCIe Gen6, positioning itself precisely at the key bottleneck as AI data centers evolve from single-chip computing power toward clusters of tens of thousands or even hundreds of thousands of accelerators.

As a result, Broadcom's next major bullish investment thesis is not simply about challenging Nvidia. Instead, it is about becoming the foundational supplier of computing resources for massive cloud-based AI inference workloads, data center optical interconnect systems, and networking infrastructure at the AI factory level. Broadcom stands to benef

it not only from partnering with Apple to develop ASIC components and working with cloud providers to build proprietary AI ASICs that significantly reduce costs, but also from the robust demand for optical interconnects, Ethernet switching, and data center networking infrastructure driven collectively by all forms of AI computing clusters.

Market Analysis:

The continued expansion of business from major customers such as OpenAI, Meta, and Anthropic is transforming Broadcom from a supplier primarily tied to Google's TPU ecosystem and an AI ASIC leader into a comprehensive AI infrastructure company spanning custom AI ASICs, networking infrastructure, data center optical interconnect systems, and the VMware virtual cloud platform. This is also the core rationale behind Wall Street's continued upward revisions to Broadcom's target share price.

Powered by Froala Editor