Up 23%, Should You Buy Broadcom Stock?

Broadcom’s AI revenue was $20 billion last year, and CEO Hock Tan has suggested it could exceed $100 billion by next year as XPU deals ramp up. The company is gaining share in AI inference/custom silicon with major customers including Alphabet, Anthropic, OpenAI, and Meta Platforms, supporting a projected 41% annualized earnings growth rate over the next 3-5 years. Despite the stock’s sharp run-up and a P/E above 80, the article argues the growth outlook still justifies holding or buying shares.

Analysis

The real implication is not that Broadcom is “another AI winner,” but that the AI supply chain is fragmenting into a two-layer market: premium training silicon remains concentrated at Nvidia, while inference-optimized custom ASICs become the margin pool for hyperscalers trying to lower cost per token. That shifts bargaining power toward the largest buyers, but it also raises switching costs for Broadcom once a design wins and software/tooling are embedded. The second-order beneficiary is the broader data-center ecosystem: networking, memory, and foundry capacity should see sustained pull-through even if unit growth in GPUs slows.

The market is likely underestimating how much of this story is duration rather than near-term EPS. Custom silicon ramps are lumpy, but once deployed they tend to have multi-year refresh cycles, which makes revenue visibility better than a typical semiconductor cycle. The consensus appears focused on headline valuation, but the more relevant question is whether Broadcom’s AI mix can compress dependence on slower-growth legacy franchises enough to keep earnings compounding above 30% for multiple years; if yes, the multiple can stay elevated without immediate mean reversion.

Main risk is not demand disappearing; it is timing slippage. If hyperscaler capex gets reprioritized toward power, data-center buildout, or a renewed GPU cycle, the inference transition could prove slower than expected and the near-term AI revenue trajectory would look too aggressive. Another risk is concentration: a handful of customers can pressure pricing once Broadcom is viewed as strategic infrastructure rather than scarce innovation, so the upside case depends on preserving custom design economics, not just landing logos.

Key Decisions for Investors

• Maintain/accumulate AVGO on 1-3 month weakness, but size for valuation risk; best risk/reward is on 6-12 month horizon if AI revenue mix keeps inflecting and earnings revisions continue upward.
• Pair trade: long AVGO / short a basket of lower-quality AI “enablers” with weaker pricing power; thesis is that custom silicon winners deserve premium multiples while commoditized exposure should de-rate as inference spend matures.
• For relative value within semis, favor AVGO over NVDA for the next 6-12 months if the market continues rewarding inference economics and multi-customer optionality; NVDA remains structurally stronger, but AVGO has more upside to estimate revisions from custom wins.
• Use call spreads rather than outright stock for new entries: buy AVGO 6-12 month call spreads to express upside to continued AI revenue re-rating while limiting drawdown from a P/E compression event.
• Set a catalyst watch on hyperscaler capex commentary over the next 1-2 earnings cycles; if cloud spending shifts back toward general-purpose compute or power constraints slow deployments, trim AVGO exposure aggressively.