Better Artificial Intelligence Stock: Arm Holdings vs. SK Hynix

SK Hynix is identified as the preferred AI-semiconductor investment, supported by 46.8% FY2025 revenue growth to KRW97.2 trillion, a 44.2% net margin, more than 50% of the HBM market, and a 5.4x forward P/E versus Arm's 125.0x. Arm delivered FY2026 revenue growth of 22.8% to $4.9 billion and guided for fiscal Q3 revenue of $1.4 billion as it expands from licensing into fabless data-center CPU products. The comparison favors SK Hynix's lower valuation and stronger near-term AI-memory growth, while highlighting Arm's China/customer-concentration risks and SK Hynix's memory-cycle, competition, and capex exposure.

Analysis

The relevant distinction is not simply asset-light versus manufacturing: ARM is attempting to move down the stack into merchant server silicon, creating potential channel conflict with its largest licensees (notably NVDA, QCOM, and hyperscaler custom-silicon teams). That strategy can lift revenue per deployed CPU if execution works, but it also changes ARM’s earnings profile from recurring royalty economics toward product-cycle, inventory, and customer-concentration risk. At an extreme revenue multiple, even modest evidence that royalty growth is being displaced rather than supplemented by chip sales would drive multiple compression over the next 1-3 quarters.

SKHY’s US listing can broaden the marginal buyer base, but the key earnings variable remains HBM qualification and mix, not headline AI demand. HBM supply is capacity-constrained and qualification-intensive, so incremental volume through 2027 should favor the incumbent supplier; the second-order beneficiary is MU, which has the clearest upside if it converts next-generation HBM qualifications into share gains. The bear case for SKHY is that aggressive capacity additions by Samsung and Micron turn a high-margin HBM shortage into a conventional DRAM downcycle faster than consensus expects, likely visible first in contract-price commentary and capex plans.

Consensus may be treating ARM’s valuation as a pure AI-IP scarcity premium and SKHY’s as a normal memory-cycle discount. A more balanced view is that ARM needs sustained royalty acceleration plus credible server-chip economics to justify its premium, while SKHY needs only HBM margins to normalize materially less than feared to rerate. Near-term, this is more likely a relative-value opportunity than a directional semiconductor call given shared exposure to AI capex expectations.

Key Decisions for Investors

• Initiate a 3-6 month pair: long SKHY / short ARM, sized beta-neutral. The setup captures HBM scarcity and US-listing liquidity support against ARM’s asymmetric multiple-risk; target 15-20% relative outperformance, with a stop if ARM raises royalty or server-related guidance by more than 10% or SKHY signals HBM pricing/mix deterioration.
• Maintain MU as a watch-to-buy rather than chase: add on evidence of material HBM3E/HBM4 customer qualification or a confirmed share-gain outlook. MU offers the cleaner upside optionality to HBM share convergence, but lacks sufficient data here to underwrite a full position immediately.
• For ARM holders, reduce exposure ahead of the next earnings print unless management quantifies data-center chip gross margin, customer commitments, and incremental R&D/capex. Falsification of the bearish valuation view is royalty growth above 30% with stable operating leverage and no sign of licensee pushback.
• Monitor Samsung HBM qualification progress, SKHY’s 2027 capex trajectory, and DRAM contract pricing monthly. A broad HBM capacity ramp or falling contract prices for two consecutive months would weaken the SKHY thesis and favor exiting the pair rather than relying on a low headline P/E.