Broadcom's AI Chip Business Just Proved Nvidia Isn't the Only Way to Bet on AI
Broadcom's latest earnings show something the market has been waiting to see: real, massive revenue from AI chips that aren't made by Nvidia.
The numbers from its fiscal Q3:
→ Overall quarterly revenue: $29.6 billion, up 86% year-over-year
→ Quarterly profit: roughly $13.1 billion — more than tripled from a year earlier
→ AI semiconductor revenue alone: $16.7 billion, up 221% year-over-year and 54% sequentially
→ Custom AI chips now make up about 73% of that AI semiconductor revenue
That last number is the real story. This isn't Broadcom selling general-purpose accelerators — it's Broadcom designing custom silicon built to the exact specifications of the hyperscalers buying it. Google, Meta, OpenAI, and other major AI players increasingly want chips tuned to their own specific workloads instead of general-purpose GPUs built for everyone.
Here's why that shift matters more than the headline growth number: for the past few years, "AI infrastructure" has largely meant one company — Nvidia. Broadcom's results are the clearest evidence yet that the market is actively diversifying away from that single point of dependency. Nvidia still dominates raw AI compute, and that isn't changing overnight. But the fact that custom silicon revenue is compounding this fast tells you hyperscalers are willing to pay a premium for chips designed around their exact needs, rather than defaulting to whatever's on the shelf.
It also reframes where the real bottleneck in AI is heading. Broadcom itself is signaling that the constraint isn't just chip access anymore — it's manufacturing capacity, advanced packaging, networking, and memory. Compute is becoming a supply chain problem, not just a design problem.
For anyone investing in or building on AI infrastructure, the takeaway is simple: the "who wins AI hardware" question isn't a one-horse race anymore, and betting everything on a single supplier is starting to look riskier than it did a year ago.
Do you think custom silicon eventually catches up to Nvidia's ecosystem advantage — or does Nvidia's software moat keep it ahead regardless of who makes faster chips?
#AI #Broadcom #Semiconductors #Nvidia #TechNews #AIInfrastructure #CustomSilicon
— Sandeep Raiza