This Might Be the Most Underrated Artificial Intelligence Stock to Own in 2026

Amazon's heavy capex buildout for AI and data centers is compressing free cash flow (trailing 12 months free cash flow fell to $14.78B by Q3 2025) even as capex rose to $115.90B (ttm) in Q3 2025. In Q3 2025 AWS produced $33.0B in revenue (up 20% YoY) with a 34.6% margin, while North America e-commerce revenue was $106.27B (4.5% margin) and international $40.9B (2.9% margin), illustrating the services-business profitability gap. Management is pushing custom Trainium3 chips (announced savings up to ~40% vs. Nvidia) to lower future run costs and create a new revenue stream, which supports a thesis that current capex-driven cash flow pressure could lead to materially stronger results and re-rating in 2026 if adoption and margin realization occur.

Analysis

Market structure: Amazon’s capex spike (TTM capex up from ~$54bn to ~$116bn between Q2‑24 and Q3‑25) and FCF decline (TTM FCF down to ~$14.8bn) shifts near‑term cash dynamics but strengthens AWS’ scale and proprietary stack. Winners are cloud infrastructure suppliers, data‑center power and automation vendors, and Amazon (if Trainium adoption hits AWS customers); near‑term losers include low‑margin retail peers and any vendor losing share to Amazon’s vertically integrated stack.

Competitive dynamics: If Trainium3 delivers ~40% cost savings claimed and meaningful software compatibility, AWS could capture incremental share vs MSFT/GOOGL and reduce Nvidia’s TAM for certain workloads—pressure on NVDA unit pricing and ASPs could follow within 12–24 months. But adoption depends on ecosystem (compilers, frameworks); Nvidia’s CUDA moat remains a material headwind to rapid migration.

Risk assessment: Tail risks include Trainium technical failure, large capex overruns forcing asset sales/debt issuance, regulatory crackdown on Amazon’s cross‑business practices, or a macro slowdown that deflates enterprise AI spend—each could materialize within 6–18 months. Catalysts to watch: Trainium benchmarks, AWS pricing/instance launches (next 3–9 months), quarterly AWS revenue growth >25% or <10% as triggers.

Trade implications: Near term (days–months) expect elevated volatility in AMZN/NVDA and modest spread widening on AMZN credit; bonds may see marginal widening if FCF stays < $15bn. Commodities impact: higher power demand supports utility/midstream names over 1–3 years. Options markets will reprice around Trainium news and earnings windows.

Key Decisions for Investors

• Establish a staged 3% long position in AMZN equity over the next 3 months (buy 1% per month) and concurrently buy Jan 2027 AMZN LEAP calls equal to ~2% notional (ATM) to capture 2026 upside; trim or stop‑loss if trailing 12‑month FCF falls below $10bn or TTM capex exceeds $140bn.
• Implement a relative value pair: long AMZN (2%) vs short AAPL (1.5%) to express cloud/AI monetization over consumer hardware; unwind if AWS revenue growth drops below 10% YoY or AAPL outperforms AMZN by >20% in 3 months.
• Use defined‑risk options: buy 12‑month protective puts 10% OTM on 50% of AMZN holdings ahead of the next two earnings/releases; alternatively sell 3‑month covered calls to partially finance LEAP exposure if portfolio already holds AMZN.
• Rotate +1–2% into data‑center power/utilities and select semiconductor exposure (NVDA/AMD) to play AI demand, and reduce pure e‑commerce retail exposure by 2%; add to semiconductor shorts only if Trainium independent benchmarks show ≥30% performance/cost parity against Nvidia within 9–12 months.