The First Half of 2026 Is Over. These 2 Spectacular Artificial Intelligence (AI) Stocks Can Soar in the Second Half.

The article argues Nvidia is undervalued despite continued AI-driven demand, citing forward P/E of 21.7x (vs ~34x average over two years) while analysts project ~96% YoY revenue growth next quarter. It highlights Meta as a near-term laggard (down ~12% YTD; 19.5x forward earnings) but claims upside from monetizing excess AI data-center capacity to external clients and either a successful AI ROI or an asset sell-down scenario. Overall, it frames this as a potential shift in the AI trade—moving from prior leaders to other beneficiaries across the AI supply chain.

Analysis

The market is still pricing AI as a single-factor trade, but the next phase is likely to be a dispersion trade between monetizers, infrastructure vendors, and over-owned beta. NVDA remains the cleanest scarcity asset because its earnings power is tied to multi-year compute replacement, not a one-quarter capex burst; if demand really extends into 2027, today’s multiple still looks too low relative to its growth durability.

META is the harder but potentially more asymmetric setup: the stock is being penalized for capex intensity because investors lack a third-party revenue proof point. If excess capacity sales become meaningful, the market could start valuing part of META like a hybrid cloud/infrastructure asset, which would mechanically reduce the discount on the whole equity story; if not, the ad business still provides a cash-flow backstop, so the downside is more about time than thesis failure.

The second-order winners are the pick-and-shovel names tied to sustained AI capex, especially memory. MU should stay supported as long as hyperscaler buildouts keep tightening HBM/DRAM supply, while AMD and INTC need share gains or product-cycle evidence to overcome richer/less defensible valuation optics. The main falsifier for the bullish AI complex is not sentiment fatigue but any sign that capex guides rise faster than revenue disclosure or that 2027 demand expectations get pushed out again.

Key Decisions for Investors

• Long NVDA vs short AMD for 1-3 months: NVDA has the better growth-to-valuation spread, while AMD is priced for more execution than it has proven; trim if NVDA forward multiple fails to re-rate above ~25x after the next earnings cycle.
• Buy META on pullbacks over the next 4-8 weeks or express via call spreads into the next print: the stock is discounting capex without giving credit for optionality; thesis breaks if external capacity revenue remains immaterial for two consecutive quarters.
• Add MU on weakness as a second-order beneficiary of sustained hyperscaler spend; this is a 6-12 month trade tied more to memory pricing and supply tightness than to day-to-day AI sentiment.
• Set an alert on NVDA and META guidance revisions: if either company signals a 2026-2027 demand or capex deceleration, de-risk immediately—those would be the first credible catalysts for multiple compression.
• Avoid chasing INTC/AMD purely on AI enthusiasm; use them only as relative-value shorts/hedges against the higher-quality AI enablers until they show sustained datacenter share gains.