2 Artificial Intelligence (AI) Stocks Poised to Run in 2026 and Beyond

Rapidly rising demand for AI compute and data-center capacity underpins bullish case for Advanced Micro Devices and Hut 8: McKinsey estimates $7 trillion of data-center investment by 2030 while Goldman Sachs forecasts >$500 billion capex in 2026 and Morgan Stanley flags a 47 GW power shortfall. AMD reported Q3 revenue up 36% YoY and adjusted net income up 31%, has Oracle deploying 50,000 MI450 GPUs starting in Q3 and will supply MI450s to OpenAI in H2 2026, and projects ~35% revenue CAGR with adjusted EPS >$20 within 3–5 years (management pegs related deal revenues at a cumulative $100 billion). Hut 8 struck a 15-year, $7 billion (potentially up to $17.7 billion) deal to supply 245 MW to Anthropic scaling toward 2.3 GW, has Q3 revenue up 91% YoY to $83 million, 13,696 BTC (~$1.6bn) vs $387m debt, and an 8.6 GW long-term development pipeline, positioning it to monetize constrained data-center power capacity.

Analysis

Market structure: Winners are GPU/AI chip suppliers (AMD, ORCL as a channel partner), data‑center power landlords (HUT) and select cloud providers; losers include legacy CPU vendors and commoditized colo REITs as watts and accelerator rack designs become the scarce resource. McKinsey’s $7T by 2030 and Goldman’s >$500B capex in 2026 imply multi‑year pricing power for high‑density rack suppliers and for sites with grid access — expect ASPs for hyperscaler GPU deployments to rise 10–30% vs. commodity colo.

Risk assessment: Tail risks include export controls/antitrust (U.S./EU/China), a delayed OpenAI/H2‑2026 ramp, Anthropic contingencies, or a BTC crash that undermines Hut 8’s balance sheet (BTC < $30k would cut its cushion materially). Immediate (days) risk = headline/quarterly volatility; short term (weeks–months) = financing and permitting outcomes; long term (years) = interest‑rate driven capex cost and TSMC capacity allocation.

Trade implications: Direct plays are selective long AMD (capture OpenAI/Oracle ramps) and long HUT for power scarcity, size modestly (2–3% and 1.5–2% respectively) and use defined‑risk option structures near key catalysts (Oracle Q3‑2026, OpenAI H2‑2026). Pair trades: long HUT vs short DLR/EQIX to express premium for power over floor space for 9–18 months. Rotate into semiconductor equipment and power infrastructure names, trim legacy colo/REIT exposure.

Contrarian angles: Consensus underprices execution/financing risk at Hut 8 and overprices AMD’s “certainty” — AMD at ~11x the optimistic long‑term target assumes flawless ramp and no margin squeeze from Nvidia/TSMC dynamics. Historical parallels (data‑center capex cycles) show big mean reversion when demand shifts from speculative AI pilots to measured enterprise adoption; stress test positions for a 30% drawdown and require key milestones (OpenAI delivery, Anthropic financing) before adding size.

Key Decisions for Investors

• Establish a 2–3% long position in AMD (ticker AMD) via a 12‑month defined‑risk call spread (buy Jan‑2027 $200 call, sell Jan‑2027 $320 call) to capture OpenAI/Oracle ramps; trim half if next two quarters’ revenue growth <25% or if OpenAI supply slips beyond H2‑2026.
• Establish a 1.5–2% position in Hut 8 (HUT) by buying shares and hedging with a 9–12 month protective put ~20% OTM; add up to +1% if company confirms >50% project financing for Anthropic or BTC > $45k, and liquidate if BTC < $30k or equity dilution >15%.
• Implement a 9–18 month pair trade: long HUT (1.5%) / short Digital Realty (DLR) (1.5%) to capture re‑rating of power‑rich sites vs legacy colo; unwind if DLR outperforms HUT by >15% or if DLR announces a major power‑focused strategic shift.
• Overweight semiconductor equipment and power infrastructure by reallocating 3–5% from legacy colo/REITs into names exposed to capacity expansion (e.g., LRCX/AMAT or utilities with transmission upgrades) and re‑evaluate after Oracle’s Q3‑2026 GPU rollout and OpenAI H2‑2026 milestone announcements.