Broadcom’s stock surged 50% in 2025, outpacing its largest rival in AI semiconductors by 14 percentage points, and has climbed nearly 700% over the past five years. The company’s custom AI accelerators have won business from some of the industry’s biggest names: Alphabet leveraged Broadcom-designed Tensor Processing Units for its Gemini models, and Anthropic placed an initial $10 billion order for Ironwood TPUs followed by an additional $11 billion commitment for delivery by late 2026. Broadcom counts four major AI developers among its accelerator customers. At the same time, the firm’s Tomahawk Ethernet switches continue to lead the data center networking market, offering industry-leading low latency and high throughput that support ever-larger AI workloads.

In its fiscal fourth quarter, Broadcom generated $18 billion in revenue—beating management’s $17.4 billion forecast—and achieved 28% year-over-year growth, marking a second consecutive quarter of acceleration. AI semiconductor revenue soared 74% to $6.5 billion, up from 63% growth in the prior quarter. For the first quarter of fiscal 2026, Broadcom projects AI chip and networking revenue of $8.2 billion, implying 100% year-over-year expansion. On the bottom line, the company reported a GAAP profit of $8.5 billion, a 97% increase from the year-ago period, and delivered $23.1 billion in GAAP earnings for fiscal 2025—nearly four times the prior year.

Broadcom’s rapid expansion carries a premium valuation, with a trailing price-to-earnings ratio of 73.3—more than double the tech index average—and a price-to-sales multiple of 26.5, nearly three times its 10-year norm. While the company’s growth profile could justify this premium over a multi-year horizon, the stretched valuation suggests limited upside in the next 12 months unless multiples contract or revenue growth accelerates further. Investors seeking exposure to AI infrastructure are advised to maintain a multi-year holding period to capture potential normalization of valuation metrics and continued strength in chip and networking demand.