Nvidia Is Near Its High While Its Biggest Chip Peers Sit 18% to 32% Below Theirs. These Are the Chip Stocks to Buy.

AI chip peers remain below their 52-week highs (AMD ~18%, Micron ~19%, Broadcom ~28%, Marvell ~32%, Nvidia ~2.6%), despite heavy hyperscaler capex expected near $800B in 2026 and $1.3T by 2027. Nvidia reported fiscal Q2’27 revenue of $96.2B (+106% YoY) with data center revenue of $89B (+117%) and guided fiscal Q3 to $108B. Broadcom aims to double AI revenue to $115B next fiscal year and double again to $230B in fiscal 2028, while Marvell logged record $2.7B revenue (+37% YoY) with data center growth of +46% and raised outlook for both FY27 and FY28; the article favors Broadcom and Marvell at their deeper discount to earnings.

Analysis

The market is still pricing AI as if the only scarce asset is GPU units, but the next leg of returns likely comes from content per rack: custom silicon, interconnect, and networking attach. That structurally favors AVGO and MRVL over the more obvious GPU beta, while downstream beneficiaries include optical, power, and cooling vendors; the risk is that investors keep paying up for the cleanest narrative while ignoring where hyperscaler capex actually compounds. Over 1-3 months, the key driver is not demand rhetoric but whether these companies keep converting capex into visible backlog and guide-up cycles.

AVGO and MRVL are both facing a familiar tradeoff: concentration and margin pressure buy visibility. In this phase, that is usually a feature, not a bug, unless customer deployment schedules slip or gross margin falls more than management telegraphed; that would be the first sign the market was right to discount them. If hyperscaler spending stays on plan, the deeper discounts should close faster than NVDA’s, because the base is lower and the growth surprise can be incremental rather than heroic.

AMD looks like the highest execution hurdle because it needs AI share gains to outrun valuation, while MU remains the classic cycle versus secular debate. The contrarian view is that the market may be over-discounting AVGO/MRVL for customer concentration while underestimating how sticky multi-year design wins can be; conversely, MU is cheap for a reason if memory normalizes faster than HBM scarcity persists. Falsify the bullish semis view if next-quarter AI capex commentary rolls over, if AVGO/MRVL order visibility weakens, or if AMD’s ramp materially outpaces expectations with no corresponding multiple compression.

Key Decisions for Investors

• Buy AVGO on any 3-5% pullback and hold into the next earnings cycle; target 15-20% upside from multiple expansion if AI revenue conversion stays on track. Stop if customer-specific deployment timing slips or gross margin guide cuts by >100 bps.
• Initiate a long MRVL / short SOXX pair for 1-3 months to express idiosyncratic upside from custom AI content versus generic semiconductor beta. Risk/reward improves if the stock remains priced for margin fears while backlog keeps rising.
• Relative-value short AMD versus long AVGO as a valuation/execution pair trade. Thesis breaks if AMD shows clear evidence of faster Instinct monetization and the market rewards it with a lower forward multiple.
• Hold MU, do not add, until the next HBM pricing and margin readthrough confirms whether memory is entering a sustained tight regime. Reassess if forward gross margin trends below the mid-80s or if spot/contract pricing softens materially.