Is the AI boom a bubble waiting to pop? Here’s what history says

The AI-driven rally has pushed major indices sharply higher—S&P 500 up 16% in 2025 and 79% since end-2022, with the Nasdaq 100 up ~130%—concentrated in names such as Nvidia, Microsoft, Alphabet, Amazon, Broadcom and Meta, which together represent roughly 30% of the S&P 500 (the top 10 stocks are ~40%). Big Tech capex from Microsoft, Alphabet, Amazon and Meta is projected to rise ~34% to about $440 billion next year, while OpenAI has signaled infrastructure spending in excess of $1 trillion, raising concerns about overinvestment and circular funding arrangements. Credit risk and debt issuance are emerging sensitivities—Oracle’s $18bn bond sale preceded a sharp stock drop and Meta, Alphabet and Oracle may need to raise ~$86bn in 2026—leaving investors weighing robust AI-driven earnings versus valuation concentration and bubble risk.

Analysis

Market structure: AI is concentrating real economic and market share into a handful of hardware (NVDA, AVGO) and cloud/software owners (MSFT, GOOGL, AMZN, META). Expect pricing power for high-end GPUs, NICs and AI-optimized ASICs for 12–24 months; datacenter-related power, real estate and networking will see 20–40% incremental demand growth vs. pre-AI baselines in 2025–26 while commodity and electricity demand edges higher regionally.

Risk assessment: Tail risks include a liquidity-driven re-pricing if credit markets tighten (corporates face ~$86bn maturities in 2026 for META/GOOGL/ORCL) or if a major AI vendor (e.g., OpenAI) falters operationally; these could trigger >25% drawdowns in crowded names within weeks. Near term (days–months) watch flows and IV; medium term (3–12 months) watch capex cadence that could create oversupply; long term (2+ years) fundamentals should separate durable winners from overbuilt suppliers.

Trade implications: Favor concentrated exposure to NVDA/AVGO/MSFT for 6–18 months but size positions modestly (2–4% each) and buy downside protection; short structurally weaker issuers reliant on bond markets (ORCL) or over-levered datacenter builders. Rotate 3–5% of risk budget into energy/utilities and industrials (power infra, copper) as indirect plays on AI infrastructure build.

Contrarian angles: Consensus underestimates circular funding and private-public feedback loops (OpenAI ↔ large cloud providers) that amplify credit risk; also underprices the durable demand for power/industrial capex even if compute capacity overbuilds. Historical parallel: railroads/internet — initial overbuild then durable platform rents; position for mean reversion in crowded mega-cap positioning rather than outright avoidance.

Key Decisions for Investors

• Establish a 3% notional long position in NVDA via a 6–9 month call spread (defined-risk) sized to equal 3% portfolio exposure; hedge with a 0.5% notional SPY 3-month put spread to protect against sudden de-risking events.
• Build 2–3% combined long in AVGO and MSFT (1–1.5% each) on any pullback >8% within the next 90 days; if either rallies >15% off current levels within 30 days, trim 50% to lock gains.
• Initiate a 1–2% short ORCL position via a 6–12 month put spread (defined risk) to express credit/fundamental stress; enter immediately or upon any new debt issuance or downgrade headline — add size if ORCL underperforms NVDA by >10% over 30 days.
• Rotate 5% of passive S&P cap-weighted exposure into value/cyclicals over the next 60 days: allocate 3% to a UK large-cap ETF and 2% to energy (XLE or equivalent) to hedge concentration risk and capture commodity upside if datacenter power demand rises.
• If IG corporate spreads widen by ≥50 bps or a top-6 tech stock sees an intra-month outflow-driven drop ≥20%, increase cash/hedge allocation by 2–4% and lengthen put protection on concentrated tech holdings (buy 3–6 month put spreads sized to cover 50% of tech beta).