The custom ASIC maker is on track to hit $100 billion in AI revenue by fiscal 2027, trading at a steep discount to peers despite faster projected growth.
As AI chip valuations stretch to historic highs, with Nvidia Corp. (NVDA) dominating the training compute narrative and Intel Corp. (INTC) surging more than 500% on its turnaround story, one industry heavyweight has flown largely under the radar: Broadcom Inc. (AVGO). The semiconductor and infrastructure software leader has locked in the world’s largest AI and cloud players as customers, delivered blistering top-line growth, and built a durable competitive moat — all while trading at a fraction of the earnings multiple of its high-flying peers.
Wall Street’s AI conversation has long centered on general-purpose graphics processing units, the workhorse of large language model training that turned Nvidia into a trillion-dollar powerhouse. But Broadcom has carved out a fast-growing, high-margin niche with application-specific integrated circuits, or ASICs: custom-built chips optimized exclusively for AI inference workloads. Deployed at hyperscale, these purpose-built accelerators deliver materially lower cost per compute than off-the-shelf data center GPUs, making them a core strategy for cloud giants seeking to cut infrastructure expenses and curb long-term reliance on a single supplier.
Today, Broadcom counts Alphabet’s Google, Meta Platforms, OpenAI and Anthropic among its flagship AI chip customers. It co-developed Google’s tensor processing units, the custom accelerators powering the search giant’s internal AI systems and cloud services, and has secured multi-year supply agreements across the hyperscaler ecosystem as demand for inference capacity outpaces training growth.
That demand is already translating to record financial results. For fiscal 2025, which ended in November, Broadcom’s AI chip revenue jumped 65% year-over-year to $20 billion, accounting for 31% of the company’s total top line. Management projects that figure will climb to at least $100 billion by fiscal 2027, when AI silicon will make up more than 58% of projected total revenue.
Beyond AI chips, Broadcom’s sprawling portfolio includes semiconductors for mobile devices, data center networking, wireless connectivity, storage and industrial applications, plus infrastructure and cybersecurity software. While those segments grow at a slower pace, they allow the company to bundle end-to-end solutions for enterprise and cloud clients, deepening customer stickiness and widening its competitive moat across multiple end markets. Consensus analyst estimates peg Broadcom’s revenue and earnings per share compound annual growth at 53% and 66%, respectively, from fiscal 2025 through fiscal 2028 — rates that rank among the highest across global large-cap chipmakers.
Despite that explosive growth trajectory, Broadcom remains one of the most attractively valued mega-cap AI stocks on the market. Shares trade at roughly 25 times next year’s expected earnings. Measured against adjusted earnings before interest, taxes, depreciation and amortization — a metric that strips out near-term noise from recent acquisitions — the company trades at just 16 times forward EBITDA based on its $1.98 trillion enterprise value.
The valuation gap with peers is stark. Nvidia, expected to post 46% compound annual growth in both revenue and EPS from fiscal 2026 through fiscal 2029, trades at about 16 times forward earnings. Intel, whose rally has been fueled by agentic AI tailwinds for server CPUs and foundry momentum, commands a forward price-to-earnings ratio above 120. Even Advanced Micro Devices Inc. (AMD), a favored pure-play AI growth name, trades at roughly 73 times forward earnings. Put simply: Broadcom delivers faster projected growth than Nvidia at a moderate premium, and offers dramatically better value than both Intel and AMD.
The stock has pulled back recently amid a broader tech sell-off, triggered in part by investor disappointment that management did not raise its fiscal 2027 AI revenue outlook at its latest update. That reaction misses the mark: Broadcom is still in the middle of its fiscal 2026 year, and its $100 billion long-term target remains firmly intact.
For long-term investors, the pullback looks like a compelling entry point. As AI inference workloads continue to explode and agentic AI drives more demand for specialized compute, hyperscaler orders for Broadcom’s custom ASICs are set to accelerate. The company’s leading positions in co-packaged optics and data center networking also act as natural tailwinds, complementing its custom chip business and locking in deeper customer relationships.
The AI compute market is not a winner-takes-all dynamic dominated by general-purpose GPUs. Custom silicon is carving out a massive, fast-growing slice of the market as operators scale production deployments and prioritize cost efficiency. Broadcom, with its deep engineering expertise, blue-chip customer base and entrenched partner relationships, already sits at the center of that shift. In a sector where valuations have become stretched across the board, it stands out as a rare combination of high-growth certainty and compelling value — making it one of the strongest long-term plays in the AI chip space.
