Broadcom Stock Is Down More Than 22% From Its Peak and Worth a Closer Look

Broadcom’s stock has fallen more than 22% from its early-June record high of $495 after management did not raise its fiscal 2027 AI chip revenue outlook. The article argues the sell-off is an overreaction, citing Broadcom’s path to over $100 billion in AI chip revenue in fiscal 2027, a forward P/E of about 19.5x fiscal 2027 estimates, and growing demand from hyperscalers for custom AI chips, TPUs, networking, and co-packaged optics.

Analysis

The market is treating a lack of near-term upside in a long-dated AI revenue view as if it were evidence the whole custom-chip franchise is peaking. That’s backwards: the important variable is not whether management “stretches” a 2027 number today, but whether hyperscalers keep shifting marginal AI spend from merchant GPUs toward proprietary silicon and the attached networking stack. If that substitution keeps happening, the earnings power compounds through both chip attach rates and a higher mix of systems-level spend, which is a better setup than a one-off revenue beat.

The second-order effect is that Broadcom is increasingly a picks-and-shovels beneficiary of AI capex optimization. When buyers pursue inference efficiency, the winners are not only the ASIC designers but also the vendors that solve power, bandwidth, and intra-cluster bottlenecks; that should support networking and optics demand even if unit growth in accelerators normalizes. This also creates a subtle pressure point for NVIDIA: the more inference workloads move to bespoke silicon, the more the market has to think about long-duration share loss at the margin rather than broad AI demand destruction.

The consensus is likely over-penalizing a valuation reset while underestimating earnings durability. At ~20x forward earnings for a business with multi-year visibility and expanding content per cluster, the stock is pricing in some combination of customer concentration risk and a stall in AI monetization that would require a much sharper slowdown than the current evidence suggests. The real risk is not the next quarter; it’s whether custom-chip economics get commoditized faster than expected over the next 12–24 months, especially if hyperscalers use Broadcom as a design partner to internalize more value over time.

From a trading perspective, this looks better as a medium-horizon relative value long than as a naked momentum bet. The catalyst path is likely incremental: new customer wins, higher networking content, and any commentary that confirms AI capex remains biased toward custom silicon rather than generic accelerators. Near term, the tape can stay noisy, but the setup favors owning the pullback while the market is still extrapolating a guidance miss into a structural slowdown.

Key Decisions for Investors

• Go long AVGO on a 1–3 month horizon with a stop if AI networking/custom-chip commentary deteriorates; risk/reward favors 20–25% upside to prior highs versus mid-teens downside if multiple compression persists.
• Pair trade: long AVGO / short NVDA as a 3–6 month relative-value expression of custom ASIC share gain versus merchant GPU intensity; best if hyperscaler capex keeps shifting toward inference efficiency.
• Buy AVGO Jan-2027 call spreads on continued weakness to capture re-rating from ~20x forward earnings if the market normalizes the guidance reaction; defined risk is preferable given headline volatility.
• For more conservative exposure, scale into AVGO only after the next earnings/industry capex check confirms networking and CPO attach rates; the business is more durable there than the current tape implies.
• Avoid chasing the move in NVDA as a hedge against AVGO strength; if anything, use NVDA as the funding leg for a basket that expresses custom-chip substitution risk over the next 12–24 months.