Broadcom's $230 Billion Roadmap Rewrites The Valuation

Broadcom projects AI semiconductor revenue to jump from $58B in FY2026 to ~$230B by FY2028. In Q3, AI revenue surged 221% to $16.7B and operating margin reached 67.9% despite an unfavorable product mix. The outlook is supported by demand for custom XPUs plus networking/optics as frontier AI infrastructure scales toward tens of gigawatts.

Analysis

The key market implication is not just higher AI revenue, but Broadcom moving deeper into the value chain where pricing power is better and the addressable wallet is larger. When a supplier sells the compute engine, the switching layer, and the network/optics around it, it reduces dependence on any single component cycle and increases the odds that AI spend becomes a multi-year annuity rather than a one-off hardware burst.

Second-order, this is more negative for merchant GPU and standalone networking vendors than the headline suggests. If hyperscalers keep shifting some training workloads into custom silicon, the market may start discounting slower unit growth for names whose economics depend on being the default chip choice; at the same time, rack-scale integration should support a narrower set of “full-stack” winners with the best system economics. The main beneficiaries outside AVGO are likely foundry and select optical supply-chain names, but only if volume growth is real rather than just design-win optimism.

The near-term risk is that the stock already reflects some version of this story, so the next leg depends on follow-through from hyperscaler capex guides and incremental design wins, not the press-release trajectory. Over 1-3 months, watch for any flattening in customer capital budgets; over 6-18 months, the real falsifier is evidence that custom XPU adoption cannibalizes margin faster than the AI mix can expand it. Consensus may be underestimating how quickly one or two customers can dominate the revenue slope, which makes the forecast powerful but also more fragile than the long-term numbers imply.

Key Decisions for Investors

• Overweight AVGO versus SOXX/SMH over the next 1-3 months: if AI infrastructure spend stays intact, Broadcom should outperform the basket because of superior mix and margin leverage; thesis breaks if hyperscaler capex guides down.
• Consider a relative-value long AVGO / short MRVL pair into any post-rally consolidation: AVGO has broader monetization across compute, networking, and optics, while MRVL remains more reliant on the same custom-silicon theme with less balance-sheet and margin cushion.
• Use pullbacks to add AVGO via stock or call spreads for a 3-6 month horizon: best risk/reward is after a 3-5% giveback, with upside driven by follow-on AI capex disclosures and design-win confirmation.
• Reduce or hedge AI basket exposure if next hyperscaler earnings show capex deceleration: that would be the cleanest falsifier for the extrapolation trade and would likely hit SOXX more than AVGO.
• Watch for a long AVGO / short NVDA relative trade only if custom silicon share gains become more explicit; until then, keep it as an alert rather than a core position.