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

Tower Semiconductor shares are up >300% over the past 12 months through Mar 19; Q4 revenue was $440.0M (+14% YoY) and the company says it is the majority supplier of 1.6T silicon photonic integrated circuits, but its P/E is ~85 indicating elevated valuation. Hut 8 reported 2025 sales of $235.1M versus $162.4M in 2024 (~45% YoY growth) but posted a $248M net loss in 2025 (vs $331.4M net income prior year); it has secured >1 GW of energy capacity with plans to add ~8.5 GW, holds total assets of $2.8B and liabilities of $1.1B, and trades at a price-to-sales of ~23 with a beta of ~6. Wall Street interest (e.g., Oasis >$100M, Flight Deck ~$10M, other institutional buys) has driven strong share gains, but high multiples and Hut 8's lack of profitability argue for cautious entry or waiting for price weakness.

Analysis

Tower’s silicon photonics exposure creates a concentrated technology moat that is as much about ecosystem control (packaging, thermal management, co-integration with electrical ICs) as it is about raw IP. That implies downstream margin capture opportunities for suppliers of advanced substrates and test/assembly services, and conversely rising working-capital and capital expenditure intensity as customers scale from design wins to multi‑rack deployments. For Hut 8, guaranteed access to grid-scale power is a structural advantage versus compute peers that must negotiate transmission or compete in spot markets, but that advantage is contingent on contract tenor and unit economics of delivered power — not nominal gigawatts alone. If long-term contracted pricing is >$X/MWh (replace with internal threshold), margin expansion is achievable; if it’s indexed to volatile wholesale prices, upside evaporates quickly and capex converts to stranded cost.

Market positioning is now bifurcated: names with visible, multi-year contracted demand (or long-term customer commitments) deserve longer investment horizons; names reliant on near-term spot demand or a single large customer are tradeable event bets. Key catalysts in the next 3–12 months that will re-rate both groups are: public hyperscaler/accelerator capex cadence (NVDA/INTC announcements), announced long-term PPAs or hosting contracts, and wafer/test capacity lead times reported by downstream assemblers. Tail risks that could reverse the current momentum include a sudden technology pivot away from 1.6T photonics to alternative interconnect standards, a material flattening in AI infrastructure growth vs. consensus, or a sharp rise in regional electricity prices that compresses Hut 8’s unit economics.

Key Decisions for Investors

• TSEM relative-value entry: Add on a 25–35% pullback from current levels or if 12‑month forward P/E reverts toward the semiconductor equipment/analog median — target 12‑18 month hold, position size 2–3% NAV, stop-loss 20%. Risk/reward: asymmetry if photonics adoption continues (2–3x upside) vs limited downside if broader semiconductor multiple compresses.
• Market‑neutral pair: Long TSEM / Short INTC equal notional to neutralize semiconductor beta — horizon 6–12 months around NVDA/Intel capex announcements. Expect this to capture share-shift to specialist photonics fabs; risk: Intel execution surprises could widen losses—trim at 15% adverse move.
• HUT volatility & event trade: Buy 9–12 month HUT call spread (long-dated, moderately OTM) funded by selling 2–3 month calls to harvest premium; enter after a >30% drawdown or on confirmation of multi-year PPAs covering >50% of incremental capacity. Reward: convex upside if capacity is monetized; risk: high implied vol and execution delays could expire premium worthless—limit allocation to 1–2% NAV.
• Tactical hedge: Monitor NVDA and INTC capital-expenditure statements as primary triggers; if NVDA/INTC capex guidance is trimmed, immediately reduce long TSEM/HUT exposure by 40% and rotate into defensive tech (NDAQ or other low-volatility names) until visibility returns.