How Broadcom Became the System-Level Winner in AI Infrastructure - Chips, Networking, and VMware

TL;DR

• Broadcom isn't a single AI product company — it's a system-level player spanning chips, networking, and software
• AI networking revenue jumped 60% in one year, making up ~1/3 of AI business → expected to reach ~40% in Q2
• VMware Cloud Foundation serves as the software control layer between AI applications and physical hardware
• "Many people can design a chip, but producing 100,000 of them quickly at yields you can afford is a different game"

Broadcom's Real Competitive Edge Is Multi-Layered

Viewing Broadcom as just an AI chip company is an incomplete perspective. The company's true strength lies in simultaneously dominating multiple layers of the AI infrastructure stack.

Most investors focus on the custom AI chips, and rightfully so — it's a massive growth driver. Management confirmed that five major customers are actively ramping custom AI chips, with OpenAI joining as the sixth customer. OpenAI is expected to deploy at over 1 gigawatt of compute capacity by 2027.

But that's only part of the story.

AI Networking — The Quietly Exploding Second Engine

AI networking revenue surged 60% in a single year and already accounts for roughly one-third of Broadcom's AI business. By Q2, networking is projected to make up approximately 40% of AI revenue.

AI Business Mix	Q1 Share	Q2 Outlook
Custom AI Chips	~67%	~60%
AI Networking	~33%	~40%

Why does this matter? As AI clusters scale, bandwidth becomes critical, switching becomes critical, and the entire network starts generating revenue alongside the chips. Broadcom isn't just profiting from the chips themselves — it's profiting from the fabric that connects them.

This is why I see Broadcom as one of the biggest non-Nvidia winners in AI infrastructure. It's a dominant custom AI player with a widening moat in networking and software. Not a one-trick pony — multiple layers of competitive advantage.

VMware — The Software Control Layer for AI Infrastructure

Infrastructure software revenue was $6.8 billion, up just 1% year-over-year. Software wasn't the hero of this quarter.

However, VMware revenue itself grew 13% YoY, Q1 total contract value (TCV) exceeded $9.2 billion, and ARR grew 19% year-over-year. Management emphasized that VMware Cloud Foundation is the software layer that sits between AI applications and physical hardware inside data centers.

This means Broadcom isn't just touching the hardware buildout — it's touching the control layer too. Computing, networking, software, and multi-year integration into customer roadmaps — that's Broadcom's positioning.

"Producing 100,000 Chips Quickly Is a Different Game"

Management made a critical point: many people can design a chip, but producing 100,000 of them quickly at yields you can afford is a completely different game.

That's manufacturing reality. That's execution. That's a moat. When a company starts talking in gigawatts, this isn't a small product cycle — it's industrial-scale demand. And the number of companies that can actually execute at that industrial scale is vanishingly small.

Investment Implications

• Broadcom's moat is system-level (chips + networking + software), not single-product
• AI networking is rapidly growing into a core pillar of the AI business
• VMware still needs to show cleaner growth acceleration, but its positioning as the AI infrastructure software layer is clear
• Manufacturing execution at scale serves as a significant barrier to entry
• If you're looking for the biggest non-Nvidia AI infrastructure winner, Broadcom is the strongest candidate

FAQ

Q: What share of Broadcom's AI business comes from networking? A: As of Q1, networking accounts for roughly one-third (33%) of AI revenue, with expectations to reach approximately 40% by Q2. AI networking revenue grew 60% year-over-year.

Q: How does VMware relate to AI? A: VMware Cloud Foundation serves as the software control layer between AI applications and physical hardware in data centers. Q1 TCV exceeded $9.2 billion, and ARR grew 19% year-over-year.

Q: How many custom AI chip customers does Broadcom have? A: Six major customers. In addition to the original five, OpenAI was added as the sixth, expected to deploy at over 1 gigawatt of compute capacity by 2027.

Q: Why is Broadcom considered the biggest non-Nvidia AI beneficiary? A: Because it dominates the custom AI chip market while simultaneously spanning AI networking (60% growth) and the VMware software layer — a system-level positioning with multiple layers of competitive advantage.