Up Nearly 40% Since April Began, Is Broadcom Stock Still a Buy?

Broadcom's custom AI chip business is gaining traction with hyperscalers, and CEO Hock Tan now projects more than $100 billion in custom AI chip revenue by 2027. Wall Street expects revenue to rise from $63.9 billion in fiscal 2025 to $158 billion in fiscal 2027, while the stock already trades at 38x 2026 earnings and 24x 2027 earnings. The article remains constructive on Broadcom despite the recent ~40% rally since early April.

Analysis

The market is starting to price AVGO less like a semicap supplier and more like a toll road on hyperscaler capex. That matters because custom silicon shifts bargaining power: if hyperscalers standardize on a narrower set of architectures, Broadcom can capture more design wins while GPU vendors lose some pricing elasticity at the margin. The second-order beneficiary is GOOGL, whose TPU stack becomes a proof point for external demand; if outside adoption scales, it validates a broader “build-vs-buy” trend across cloud AI budgets.

The key question is not whether AI spend continues, but whether it migrates from experimental compute to infrastructure efficiency. Custom chips improve cost-per-token economics, which can extend AI deployment even if aggregate capex growth slows, making AVGO’s revenue stream more durable than a simple cyclical hardware trade. The flip side is concentration risk: if even one or two hyperscalers delay node transitions or internalize more of the silicon stack, AVGO’s forward estimates can de-rate quickly because the stock is already discounting a strong 2026-27 ramp.

Consensus is likely underestimating how much of AVGO’s upside now depends on execution cadence rather than demand alone. At ~24x 2027 earnings, the market is effectively paying for near-flawless rollout plus continued AI budget expansion; that is achievable, but it leaves less room for supply-chain hiccups, yield issues, or a pause in cloud spend. NVDA is a relative loser only at the margin here: not because demand disappears, but because every dollar shifted into custom inference/training silicon is a dollar with lower attach to the full GPU stack.

Key Decisions for Investors

• Stay long AVGO, but treat it as a 6-12 month compounder rather than a fresh momentum chase; add only on pullbacks of 5-8% or after broad tech risk-off days, since upside from here is more about estimate revisions than multiple expansion.
• Pair trade: long AVGO / short NVDA in a modest size basket over 3-6 months. Thesis: custom silicon wins a growing share of hyperscaler AI budgets, compressing the relative multiple of the more obvious beneficiary while AVGO’s cash flows broaden.
• Buy GOOGL on weakness as a second-order beneficiary over 6-18 months. If TPU externalization accelerates, it creates an underappreciated monetization path for cloud AI and validates Broadcom’s custom-chip thesis.
• Use AVGO call spreads rather than outright equity if initiating now: the stock has already rerated, so capped upside with limited premium outlay offers better risk/reward than chasing spot into earnings-guided optimism.
• Avoid shorting AVGO outright unless there is evidence of order delays or capex rollovers from hyperscalers; the better bearish expression is a relative-value short against a slower-growth AI hardware basket, not a standalone directional bet.