Broadcom's $120B AI Bet: Can Hock Tan Outpace Nvidia With Next-Gen Chips?

JPMorgan analyst Harlan Sur highlights Broadcom's (AVGO) potential to significantly challenge Nvidia's (NVDA) AI dominance, driven by its forthcoming 2nm custom AI XPU, which is on track for tape-out this year and could offer a technological leap. Broadcom's strategy extends beyond chips to include rack-scale deployments and leveraging its Ethernet networking solutions, aiming to capture a broader share of the AI infrastructure and erode Nvidia's InfiniBand advantage. With CEO Hock Tan incentivized by a $120 billion AI revenue target by 2030, Broadcom is positioning itself as a formidable long-term competitor in the AI semiconductor market.

Analysis

A JPMorgan analyst report indicates that Broadcom (AVGO) is strategically positioning itself as a significant, long-term competitor to Nvidia's (NVDA) dominance in the artificial intelligence market. The core of this challenge rests on Broadcom's custom AI XPU program, which is on track to tape-out the industry's first 2nm AI accelerator this year, a milestone that could provide a technological lead over merchant GPU providers. This initiative is supported by tangible customer traction, with three hyperscalers already ramping up and four more in the pipeline, underpinning an ambitious $120 billion AI revenue target by 2030 that is directly tied to CEO Hock Tan's compensation. Broadcom's strategy extends beyond silicon to full rack-scale AI systems and a direct challenge to Nvidia's networking advantage by promoting its high-bandwidth Tomahawk6 Ethernet solutions over Nvidia's favored InfiniBand. While the rack-scale deployments carry lower margins, they are expected to drive higher absolute profit dollars and cement Broadcom's position as an integrated solutions provider, not merely a component supplier.

Key Decisions for Investors

• Investors should evaluate Broadcom's long-term potential based on its AI systems strategy, monitoring the successful tape-out of its 2nm XPU this year and customer ramp-ups as key validation points for its $120 billion revenue goal.
• Given the direct challenge to Nvidia's structural advantages in both GPUs and networking, investors holding NVDA should closely monitor the adoption rate of Broadcom's Ethernet-based AI solutions by hyperscalers as a leading indicator of potential market share erosion.
• Consider the strategic shift to full rack-scale systems, which, while diluting gross margins, could significantly increase absolute profit and create a stickier ecosystem, potentially warranting a valuation re-rating for Broadcom beyond its traditional component business model.