Menu
Broadcom is leveraging AI-specific chips (ASICs) to cement a leadership role in hyperscale data center infrastructure.
CEO Hock Tan’s acquisition strategy transformed Broadcom from a semiconductor roll-up into a high-margin hybrid of chip and software prowess.
While Nvidia dominates headlines, Broadcom quietly powers many of the hyperscalers driving AI’s next wave.
Broadcom (NASDAQ: AVGO) has spent the past two decades crafting an empire not through buzz or bravado, but through calculated, transformative acquisitions—and under-the-radar dominance in a sector now central to AI: custom chips.
Once best known as a supplier for iPhone components, Broadcom’s pivot to high-margin software and application-specific integrated circuits (ASICs) places it squarely in the AI conversation. CEO Hock Tan, a methodical leader with a background that spans General Motors and PepsiCo, has executed a long-game vision that’s now paying off in the AI buildout.
The company’s history is stitched together from the leftovers of legacy tech. Hewlett-Packard’s 1999 spinoff, Agilent, passed off its volatile semiconductor arm to private equity in 2005, where it was renamed Avago. Tan was tapped to lead the unit. Under his stewardship, Avago bought Broadcom in 2016 and took its name, retaining the AVGO ticker. Since then, he’s acquired software assets like CA Technologies and VMware—controversial moves that now provide recurring revenue to balance hardware cyclicality.
Today, the AI boom is shifting the chip narrative. While Nvidia (NASDAQ: NVDA) dominates with its GPUs, which serve a wide range of AI applications, hyperscalers like Google, Meta, and OpenAI are turning to ASICs—tailored chips that optimize performance and cost for specific tasks. Broadcom leads in this niche. Its deep design expertise and strong ties with Taiwan Semiconductor Manufacturing (TSMC) have secured it a top position in this crucial supply chain.
The industry’s evolving lexicon—XPUs, TPUs, and other derivatives—can confuse even insiders. But the essence is clear: custom hardware is essential for differentiated AI performance, and Broadcom is powering much of it.
A year ago, Broadcom’s AI ASIC client roster included three major players. Today, it’s seven, including ByteDance and others believed to be scaling aggressively. Broadcom’s projections suggest those original three customers could each deploy a million chips by 2027, hinting at a $60–90 billion addressable market. Analyst Ben Reitzes at Melius Research forecasts $36 billion in AI chip revenue by 2027, with upside toward $70 billion as new customers mature.
Financially, the momentum is real: last quarter, Broadcom posted $4.4 billion in AI-related revenue—a 46% increase—and expects $5.1 billion next quarter. Overall revenue hit $15 billion, with free cash flow rising 44% to $6.4 billion.
But valuation is catching up. Trading at 32x forward free cash flow, Broadcom is no longer a deep-value play. Future gains hinge not just on fundamentals, but on continued investor appetite for AI exposure.
Still, as hype-driven names swing wildly, Broadcom’s calculated growth and diversified revenue streams make it a compelling story. It might not double overnight—but it’s positioned to quietly power the AI economy for years to come.
