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

Nvidia stock is still ~2.6% below its 52-week high, despite strong momentum: fiscal Q2 revenue of $96.2B (+106% YoY) and data center revenue of $89B (+117%), with management guiding fiscal Q3 to ~$108B. The article argues the peer discount is opportunity, citing Broadcom’s AI revenue of $16.7B (+221% YoY) and an AI target of $115B in fiscal 2027 (then $230B in fiscal 2028), alongside Marvell’s 46% data center growth and raised outlooks. While AMD and Micron also trade meaningfully below highs (about 18% and 19% below, respectively), valuation/cycle concerns are framed as manageable, supporting a risk-managed “buy the deeper-discount names” view.

Analysis

The market is starting to price AI exposure by controllability of cash flows, not by top-line growth alone. That favors AVGO and MRVL because custom silicon and networking are the “picks-and-shovels” layer that monetizes hyperscaler capex even if GPU unit growth slows; the flip side is that these names are more exposed to a handful of deployment schedules, so sentiment can swing hard on one customer slip. NVDA remains the highest-quality expression of the theme, but the easy multiple re-rating has likely already happened unless supply expansion surprises again.

AMD is the clearest relative loser on a 3-6 month horizon: it needs both product execution and software adoption to converge while its bill of materials rises if memory tightens. That creates margin leverage in the wrong direction if accelerator pricing gets competitive before ecosystem lock-in improves. MU is the hidden beneficiary of higher AI intensity, but the market will keep treating it like a cycle peak until pricing evidence shows up in spot and contract data; that makes it tradable, but not yet a high-conviction compounder.

The contrarian risk is that the “discounts” are partly justified by concentration. AVGO/MRVL are more dependent on a narrow customer set than the market is comfortable admitting, and the path from announced AI revenue targets to delivered earnings can be punctuated by delays. If hyperscaler capex growth decelerates from the current pace or a major cloud customer pushes out AI spend, the cheap names can cheapen further before re-rating. I’d watch for any guide-down in custom ASIC ramps, gross margin compression beyond ~100 bps, or AMD Instinct take-rate data as the thesis breakers.

Key Decisions for Investors

• Long AVGO / short AMD for 1-3 months: better risk/reward if AI spend keeps shifting toward custom silicon and networking; target 10-15% relative outperformance, stop if AMD’s next guide implies faster MI ramp or AVGO customer concentration becomes a headline risk.
• Buy MRVL on weakness only, with a 6-12 month horizon: the stock is discounting the right kind of problem (mix-driven margin pressure) rather than demand failure; add if management reiterates bookings strength and no material customer delay appears.
• Hold MU as a structural hedge, but do not add aggressively until memory pricing data confirms AI-related tightness; if DRAM/NAND spot stays firm into the next earnings cycle, MU can re-rate from ‘cycle peak’ to ‘under-earning’ in 3-6 months.
• Avoid chasing NVDA here unless there is a pullback or a supply-capacity upside surprise; at current levels the trade is more about maintaining core exposure than seeking incremental alpha.