MarketMoodz
Broadcom Custom Chips Test Nvidia Dominance, Not Breakthrough Yet
Broadcom is piloting a wave of hyperscaler custom AI chips, a move that tests Nvidia's dominance but stops short of a breakthrough. The collaboration with Google on Tensor Processing Units powering Gemini 3 highlights ASIC-based accelerators' rising viability, even as Nvidia remains broadly dominant.
Key Takeaways
- Hyperscalers are increasingly adopting custom ASICs, with Broadcom positioned as a leading player.
- Google co-developed TPUs with Broadcom to train Gemini 3, launched in November.
- TPUs offer an AI-accelerator alternative to Nvidia GPUs but do not replace Nvidia's overall dominance.
- OpenAI's October deal with Broadcom signals supply-chain diversification away from Nvidia-dominated sourcing.
- Analysts see potential upside for Broadcom from AI revenue and TPU adoption, but Nvidia's share could stay sizable for years.
People Involved
- Hock TanBroadcom CEO
- Jensen HuangNvidia CEO
- Gil LuriaD.A. Davidson Analyst
- Jim CramerCNBC Host
Entities Involved
- Broadcom Inc. (AVGO)Semiconductor and infrastructure software company
- NVIDIA Corp. (NVDA)GPU and AI accelerator leader
- Google LLCDeveloper of TPUs; Gemini 3 training milestone
- OpenAIAI research lab with Broadcom partnership
- TSMCFoundry partner facing capacity constraints
MarketMoodz Analysis
For investors, the story signals a potential re-rating of Broadcom as AI-related revenue expands. Broadcom reported AI revenue up 65% year over year to $20 billion, with total semiconductor revenue at a record $37 billion, underscoring how chip suppliers can ride AI-driven demand beyond pure GPU enzymes.
Nvidia remains the dominant force in AI compute, but the path toward diversification—via ASICs and co-developed accelerators—could temper the pace of Nvidia's share gains. The manufacturing bottlenecks at leading foundries like TSMC and the high barriers to scalable ASIC production are persistent risks that could delay a wide shift in supply dynamics.
What to watch next: Google's continued collaboration with Broadcom on TPUs and any expansions in TPU shipments could lift Broadcom's AI-related revenue; Wolfe Research projects 7 million TPU shipments by 2028 with a $400 Broadcom price target (~21% upside). Analysts remain cautiously constructive on both names, with Nvidia favored for cloud ROI via Vera Rubin ramp in H2 2026, and Broadcom seen as a potential beneficiary if TPU adoption expands.
Source: Original Article
Get AI-Powered Market Insights
Stay ahead of market-moving events with our real-time analysis and stock ratings.
Start Your Free Trial