2 Artificial Intelligence (AI) Stocks That Wall Street Loves but Most Investors Haven't Heard Of

Tower Semiconductor reported record Q4 revenue of $440M (+14% YoY) and its shares have risen >300% over the past 12 months, but a high P/E of ~85 indicates a stretched valuation. Hut 8 grew revenue to $235.1M in 2025 (from $162.4M in 2024) while posting a net loss of $248M vs prior-year net income of $331.4M; it has secured >1 GW of energy capacity across 15 facilities and plans to add ~8.5 GW, with a price-to-sales ratio of ~23 and a beta of ~6. Wall Street fund buying (e.g., Hood River, Rockingstone, Flight Deck, Oasis) underpins investor interest, but the article advises waiting for price dips given lofty multiples and Hut 8's volatility and recent loss.

Analysis

Photonics and analog specialists sit at an unusual inflection: adoption is governed less by short-term software hype and more by multi-quarter hardware lead times, test/pack capacity, and qualified customer design wins. That creates a two-speed market where vendors who lock multi-year supply agreements and co-design wins with hyperscalers can sustain >30% incremental gross margins while others face fierce pricing pressure as customers consolidate suppliers.

For power-centric compute providers the economics are dominated by how capacity is contracted and where transmission constraints bite; owning generation without secured long-haul transmission or firmed renewables exposes margins to merchant power spikes and capacity-market repricing. The optionality from being able to monetize curtailment windows and behind-the-meter renewables can transform unit economics, but it requires 12–36 month execution and regulatory navigation in multiple ISOs.

Key catalysts in the next 6–18 months are large-scale design win announcements, long-term capacity contracts with a top-3 cloud, and incremental foundry/test capacity investments; conversely, the principal reversal vectors are rapid vertical integration by hyperscalers and a spike in wholesale power that forces passthrough limits. Monitor customer concentration metrics and remaining contracted power years as high-leverage early-warning indicators.

Consensus appears to underweight execution and capital-risk: hardware wins convert into durable cash only if supply-chain, packaging, and power are simultaneously solved. That argues for asymmetric exposure — concentrated directional bets via defined-loss option structures or short-duration pair trades rather than outright multi-quarter equity holds.

Key Decisions for Investors

• Tactical long (TSEM): Buy 12–18 month 25–35% OTM call LEAPS on a >20% pullback, sizing at 1–2% of NAV. Rationale: captures asymmetric upside from design-win cadence while limiting downside to premium; target 2.5–4x on successful customer ramps, cut if option premium falls 50%.
• Hedge/short (HUT): Purchase a 6–9 month put spread sized 0.5–1% of NAV (protective, limited-risk) or initiate a small outright short if paper liquidity permits. Rationale: capital-intensity and power exposure create binary execution risk; unwind if company secures multi-year take-or-pay contracts with top-3 cloud providers.
• Pair trade (TSEM/HUT): Go long TSEM equity or calls and short HUT via puts in 1:1 notional, 6–18 month horizon, 1–2% net exposure. Rationale: expresses view that differentiated hardware moat will outperform capital-intensive power-resale model; expected asymmetric payoff if supply wins materialize.
• Event hedge: Buy 3–6 month call spreads on a broad AI infra proxy (e.g., NVDA or market ETF) sized to offset short-tail volatility from sector re-ratings. Rationale: protects against sector-wide rallies that could re-rate both names before fundamentals change.