With AI Revenue Set to Surge 400% Over the Next 2 Years, Broadcom Stock Looks Like a Buy on Recent Dip

Broadcom projects AI revenue to double to $115B in fiscal 2027 (from $100B earlier guidance) and then double again to $230B in fiscal 2028. In fiscal Q3 (ended Aug. 2), revenue rose 86% YoY to $29.6B and adjusted EPS jumped 96% to $3.32, topping estimates ($29.36B revenue; $3.24 EPS). AI revenue guidance for fiscal Q4 calls for +236% YoY to $21.6B, alongside a gross margin expected at 73%; the article frames the stock as “incredibly cheap” at a 11.5x forward P/E versus rapid AI growth.

Analysis

Broadcom is increasingly a financing and integration layer for hyperscaler AI capex, not just a chip designer. That matters because the market is still discounting the business like a high-quality semiconductor compounder, while the economics are drifting toward a quasi-software infrastructure model with long-duration customer commitments and less unit volatility. The first-order winner is AVGO; the second-order winner is any customer using custom silicon to lower inference cost, especially GOOGL and META, because cheaper model serving expands room for product investment and margin defense.

The bigger competitive implication is pressure on merchant GPUs at the margin, but mainly in inference rather than frontier training. That means the near-term risk to NVDA is not a sudden share collapse; it is slower incremental demand growth as more workloads are routed into bespoke ASICs, which can cap upside multiple expansion even if absolute GPU demand stays healthy. If Broadcom’s networking attach keeps scaling alongside compute, the ecosystem takeaway is that AI capex is becoming more vertically integrated, which tends to favor the vendors closest to the customer roadmap and punish generic hardware suppliers.

The main watch item is supply, not demand. Management can talk up 2027-2028 revenue, but if advanced packaging, power availability, or customer deployment timing slips, the stock could de-rate quickly because the current valuation already prices a very clean ramp. The contrarian view is that the cheap multiple may be misleading: gross margin mix is moving the wrong way, and if AI spend rotates from build-out to optimization, the revenue base can still grow while the stock stops working.

Key Decisions for Investors

• Long AVGO on 3-6 month weakness, but size for execution risk; the thesis breaks if next guidance implies AI revenue growth is slowing materially below current forward expectations or if gross margin trends persist below the low-70s.
• Pair trade: long AVGO / short NVDA over 6-12 months to express the shift from merchant GPUs to custom inference silicon; this works best if hyperscaler capex stays strong but mix continues moving toward ASICs. Falsifier: NVDA re-accelerates on training demand or software attach widens further.
• Add small tactical longs in GOOGL and META into earnings over the next 1-3 quarters as indirect beneficiaries of lower inference cost and better AI unit economics; cut if AI capex starts crowding out core margin expansion.
• Use a watchlist alert on AVGO if the stock re-rates above its current implied FY28 multiple without supply confirmation; that would be a setup to fade, since the market would then be paying upfront for capacity that has not yet been de-risked.