Broadcom Builds Custom Chips for Google, Meta, Anthropic, and OpenAI. At 25 Times Forward Earnings, It's the Cheapest Mega-Cap AI Stock Nobody Talks About.

Broadcom's AI chip sales surged 65% to $20 billion in fiscal 2025, and management expects AI revenue to reach at least $100 billion by fiscal 2027, implying AI will exceed 58% of projected revenue. Analysts expect revenue and EPS CAGRs of 53% and 66% from fiscal 2025 to fiscal 2028, while the stock still trades at 25x next year's earnings and 16x next year's adjusted EBITDA. The article argues Broadcom remains an attractive long-term AI beneficiary despite Nvidia's lower multiple.

Analysis

The market is still pricing Broadcom as a good AI beneficiary; the setup is that it may be becoming the second-order infrastructure toll collector. If custom accelerators keep gaining share in inference, the value pool shifts away from pure compute performance and toward system-level integration: networking, storage, software, and long-lived design wins. That favors AVGO’s bundled architecture and makes its gross margin profile more durable than a simple chip-cycle multiple would imply.

The key underappreciated implication is competitive pressure on Nvidia is more about customer spend allocation than outright displacement. Hyperscalers will likely keep Nvidia for frontier training while incrementally re-routing inference budgets to custom ASICs, which can cap incremental growth in the GPU franchise without requiring a headline share-loss event. That means the more important read-through is on capex mix: semiconductor dollars increasingly favor custom silicon, interconnect, and packaging ecosystems, while general-purpose compute gets a lower share of incremental wallet.

The main risk is timing mismatch. The forward revenue math assumes a very steep adoption curve, but custom silicon ramps are lumpy and depend on a narrow set of customers hitting deployment milestones on schedule; any slip in model inference efficiency, software integration, or design tape-outs could push the inflection out 2-4 quarters. In that scenario, the stock’s premium to the broader semiconductor group can compress even if the secular thesis remains intact.

Consensus is likely underestimating how much of AVGO’s upside is already tied to customer concentration and how little that concentration matters if the chips become embedded in mission-critical inference workloads. The bull case is not just AI unit growth, but pricing power from switching costs and multi-product attachment. The more interesting contrarian angle is that NVDA may still be the better trading vehicle on near-term AI excitement, while AVGO is the better compounding vehicle if investors can tolerate fewer headlines and more execution risk.

Key Decisions for Investors

• Stay long AVGO into the next 3-6 months, but size it as a compounding core rather than a momentum trade; the upside is in multiple durability, while the main risk is a 2-4 quarter delay in AI conversion.
• Pair trade: long AVGO / short a basket of semi names with higher training-exposure beta for the next 1-2 quarters; thesis is that inference/custom silicon mix shift benefits AVGO more than general-purpose compute peers.
• Use any post-earnings or AI-capex pullback to add AVGO on weakness rather than chase strength; expected reward is better entry for a multi-year re-rating, while downside is limited to execution slippage rather than thesis break.
• For hedged exposure to AI spend, consider long AVGO and short a smaller notional NVDA call spread if you expect incremental enterprise AI budgets to rotate from training to inference over the next 6-12 months.