Nvidia's Hidden $60 Billion Business Is About to Overtake Broadcom

Nvidia disclosed $15 billion in fiscal Q1 networking revenue, implying a $60 billion annualized run rate and roughly 3x year-over-year growth. The article argues this makes Nvidia far more than a GPU supplier, with its Mellanox-backed InfiniBand and Spectrum-X platforms competing directly with Broadcom, Cisco, and Arista in AI networking. The key implication is a potentially higher valuation for Nvidia as it captures a larger share of AI data center infrastructure.

Analysis

The market is still pricing Nvidia as a component supplier, but the bigger second-order shift is that it is turning into the control plane for AI infrastructure. Once the networking layer is integrated with the GPU, software stack, and system design, the economics become stickier: customers are no longer just buying silicon, they are buying a closed architecture that raises switching costs and compresses the addressable share for standalone networking vendors over the next 12-24 months.

That creates a subtle winner/loser dynamic. Broadcom and Arista can still grow, but their upside increasingly depends on winning the “open” portion of AI buildouts, while Nvidia captures the highest-value integrated deployments where latency and tuning matter most. Cisco is more exposed to enterprise refresh cycles and is least able to monetize the bleeding-edge cluster architecture, so it risks becoming a slower-growing incumbent as AI capex migrates toward hyperscaler-native designs.

The key risk is not competitive response so much as digestion. If hyperscalers or sovereign buyers push back on vendor concentration, or if AI capex growth normalizes after the current buildout wave, Nvidia’s networking ramp could decelerate faster than investors expect. That would matter because the stock’s multiple is now increasingly anchored to a belief in sustained platform share gains; any evidence of order lumpiness would likely hit the name harder than the peers over a 1-2 quarter horizon.

Contrarian read: the market may be underestimating how much of Nvidia’s networking opportunity is already embedded in sentiment, even if the revenue mix shift is real. The best risk/reward may not be chasing NVDA outright after the breakout, but expressing the theme through relative value where the winner/loser gap can widen without requiring perfect macro conditions.

Key Decisions for Investors

• Long NVDA / short CSCO for 3-6 months: best expression of AI infrastructure migration versus legacy enterprise networking; target continued relative underperformance of Cisco as AI spend concentrates in hyperscaler-native architectures.
• Maintain core long NVDA but add on pullbacks only; use 6-12 month horizon and size down versus a normal momentum winner because the valuation now discounts sustained execution across multiple business lines.
• Pair trade: long NVDA / short ANET on strength if hyperscaler capex data stays hot; this captures the integrated-stack premium while hedging against a rerating of pure-play networking multiples.
• For investors already long AVGO, use it as a hedge against NVDA-specific execution risk rather than a pure AI upside proxy; AVGO remains a high-quality beneficiary, but its share of the upside is more exposed to open-networking competition.
• If seeking convexity, buy NVDA call spreads 3-6 months out rather than outright calls; the thesis is still intact, but implied volatility and post-breakout consolidation argue for defined-risk exposure.