Broadcom’s AI Play: Infrastructure, Not LLMs in 2026

Misinformation abounds when discussing investor expectations for Broadcom’s AI investments and its role in the burgeoning LLM market. Many assume a direct, consumer-facing play, overlooking the nuanced but critical infrastructure Broadcom provides. How deeply integrated is Broadcom into the core of AI growth?

Myth 1: Broadcom is building its own large language models.

There’s a persistent misconception that Broadcom is directly competing with firms like OpenAI or Google in developing proprietary large language models (LLMs). This simply isn’t true. Broadcom’s strategy is far more foundational. Their strength lies in the underlying hardware and software infrastructure that makes LLMs, and AI in general, possible. Think of it this way: while some companies are designing the intricate neural networks that power conversational AI, Broadcom is manufacturing the high-speed roads, the strong power grids, and the essential communication lines these networks run on.

Broadcom’s core business in this space revolves around high-performance networking chips and custom silicon. For instance, their Tomahawk 5 Ethernet switches are important for hyperscale data centers that train and deploy LLMs. These switches handle the immense data traffic generated by AI workloads. Without such specialized hardware, the sheer volume of data moving between GPUs and compute clusters would create insurmountable bottlenecks, rendering even the most sophisticated LLMs impractical. According to a recent market analysis by Dell’Oro Group, the Ethernet switch market for AI applications is expected to see significant growth, with Broadcom positioned as a key supplier for these critical components.

Plus, Broadcom designs and manufactures application-specific integrated circuits (ASICs) for major AI developers. These custom chips are optimized for specific AI tasks, offering superior performance and energy efficiency compared to general-purpose processors. While the specifics of these partnerships are often proprietary, the trend toward custom silicon in AI is undeniable, as highlighted in reports from Deloitte’s Technology, Media, and Telecommunications predictions for 2026. Broadcom isn’t creating the LLMs. They’re providing the tailored engines that power them, an often-overlooked but utterly indispensable role.

Myth 2: Broadcom’s AI growth is solely dependent on chip sales.

Many investors focus narrowly on Broadcom’s semiconductor division when assessing its AI potential. While chips are undeniably a major component, ignoring the significant role of their software and infrastructure solutions is a mistake. The acquisition of VMware in 2023 dramatically reshaped Broadcom’s portfolio, positioning it as a powerhouse in software-defined infrastructure, which is increasingly vital for AI deployments.

AI workloads require scalable, flexible, and secure environments. VMware’s virtualization and cloud management platforms provide exactly that. Hyperscalers and large enterprises using AI need to orchestrate vast compute resources, manage complex data pipelines, and ensure high availability for their AI models. VMware’s software stack allows for the efficient provisioning and management of virtual machines and containers, optimizing the utilization of expensive AI hardware. This means that a data center investing in Broadcom’s networking chips to support AI is also very likely to be running VMware software to manage those resources.

Consider the operational challenges of deploying an LLM, for example. It’s not just about having powerful GPUs. It’s about smoothly integrating those GPUs into a larger ecosystem, ensuring data integrity, and providing strong security. VMware’s solutions, such as VMware Cloud Foundation, offer a unified platform for compute, storage, and networking, simplifying the deployment and management of AI infrastructure. This integrated approach reduces operational overhead and accelerates the time-to-value for AI initiatives, making Broadcom’s combined offering far more compelling than its chip sales alone might suggest. The software side of the business provides recurring revenue and deep customer lock-in, which is a powerful differentiator in the competitive AI market.

Myth 3: Broadcom is a latecomer to the AI market.

Some observers mistakenly believe Broadcom is only now scrambling to catch up in the AI race. This perspective ignores decades of foundational work in networking and custom silicon that are now directly benefiting from the AI boom. Broadcom has been a silent but critical enabler of high-performance computing and data center infrastructure for a very long time. Their expertise in designing complex system-on-a-chip (SoC) solutions and high-speed interconnects predates the current LLM frenzy by many years.

Broadcom’s Jericho series of network processors, for instance, have been evolving for over a decade, designed for the demanding requirements of hyperscale networks. The latest iteration, Jericho3-AI, is specifically optimized for AI workloads, featuring advanced capabilities like congestion management and ultra-low latency communication, which are paramount for distributed AI training. These are not new inventions. They are the culmination of continuous research and development, strategically adapted for the most pressing needs of AI. A recent article in EE Times detailed the engineering advancements in Jericho3-AI, underscoring years of development.

Plus, Broadcom’s long-standing relationships with major cloud providers and enterprise clients give them an inherent advantage. These are the companies building the massive data centers required for AI. They already trust Broadcom for their existing networking and compute infrastructure. Transitioning to Broadcom’s AI-specific solutions is a natural extension of these established partnerships, not a leap of faith into an unknown vendor. This deep integration into the existing technology ecosystem means Broadcom isn’t playing catch-up. They’re capitalizing on their established position as a foundational provider.

Myth 4: Broadcom’s AI play is too niche to drive significant investor returns.

The idea that Broadcom’s focus on infrastructure and custom silicon is too niche to generate substantial returns for investors misjudges the scale and foundational nature of the AI build-out. While Broadcom may not have the consumer-facing glamour of an LLM developer, their products are indispensable to every single AI initiative, from research labs to global enterprises. The market for AI infrastructure is vast and growing exponentially.

Consider the sheer capital expenditure involved in building and expanding AI data centers. Companies like Meta, Google, and Amazon are pouring billions into these facilities. Each new rack of GPUs requires high-speed networking, strong power management, and sophisticated software orchestration. Broadcom supplies critical components across these layers. The demand for high-bandwidth interconnects and custom AI accelerators is not a niche market. It’s the very bedrock of the AI revolution. According to financial reports from major tech companies, capital expenditures related to AI infrastructure are projected to continue their upward trajectory through the end of the decade.

On top of that, the recurring revenue streams from Broadcom’s software portfolio, particularly VMware, add a layer of stability and predictability often absent in pure hardware plays. As AI workloads become more complex and widespread, the need for strong management and security tools only increases. This creates a continuous demand for Broadcom’s software services and licenses. The combined strength of their semiconductor leadership and their enterprise software dominance positions Broadcom to capture significant value from the entire AI ecosystem, not just a small segment of it. It’s a strategic move that I believe will pay dividends for years to come, despite what some might consider a less “exciting” role.

Myth 5: AI growth will be stifled by competition, limiting Broadcom’s upside.

While competition in the broader AI space is fierce, Broadcom operates in segments where its competitive advantages are deeply entrenched, making it less susceptible to rapid market shifts. The complexity and capital intensity of designing and manufacturing advanced networking chips and custom ASICs create significant barriers to entry. This isn’t a market where a startup can easily disrupt established players overnight.

Broadcom’s expertise in specialized silicon design, coupled with its extensive intellectual property portfolio, gives it a strong moat. Developing a new generation of high-performance Ethernet switches, for example, requires billions in R&D and years of engineering talent. This makes it difficult for new entrants to compete effectively on performance, power efficiency, or cost at scale. Established competitors like Marvell and Cisco are strong, yes, but the overall market for AI infrastructure is expanding so rapidly that there’s ample room for multiple key players to thrive. The pie is growing, not merely being re-sliced.

Plus, Broadcom’s long-term relationships with hyperscale customers are built on trust, proven performance, and deep integration. These customers are unlikely to switch critical infrastructure providers lightly, especially given the mission-critical nature of AI workloads. The cost and risk associated with migrating from a proven Broadcom solution to an unproven alternative often outweigh any perceived benefits. Therefore, while competition exists, Broadcom’s established position, technical leadership, and strategic customer relationships insulate it considerably, ensuring continued upside from the massive wave of AI investments.

Broadcom’s strategic positioning within the AI ecosystem, focusing on foundational infrastructure rather than direct LLM development, offers a compelling investment thesis. Their deep expertise in high-performance networking and custom silicon, augmented by their strong software portfolio, places them at the core of the ongoing AI revolution. Investors should look beyond the surface-level narratives and recognize Broadcom’s critical, indispensable role in powering the future of artificial intelligence.