The U.S. economy was stagnant in 2025 — with one exception

U.S. economic growth in 2025 was effectively stagnant except for a concentrated AI investment boom: Harvard analysts attribute 92% of GDP growth this year to California tech firms’ AI infrastructure spending while another study credits AI data-center spending with half of growth from Q2 2024–Q2 2025. The Trump administration has backed more than $1 trillion in AI-related investments, including a $500 billion data-center buildout (Stargate with OpenAI and Oracle) coming online 2026–2028, even as AI-driven layoffs (48,000 nationwide) and California job losses (158,734 through October) coincide with weak broad productivity and wage growth barely above inflation. Markets have been narrowly led by top tech names — the top 10 S&P 500 stocks drove 60% of the yearlong rally — raising bubble concerns and concentrating consumption gains among the wealthiest households.

Analysis

Market structure: The immediate winners are AI infrastructure owners (ORCL, AMZN, GOOGL, MSFT, NVDA and data‑center REITs such as EQIX/DLR) capturing heavy capex; losers are mid‑sized tech, entertainment and consumer discretionary sectors where wage income and broad consumption are weak. Concentration risk is acute — top 10 S&P names drove ~60% of the rally — so index performance is hostage to a handful of compute/capex chains. Supply/demand for compute will swing from undersupply now to potential overcapacity by 2026–28 as Stargate and other hyperscale builds come online, pressuring GPU/colocation pricing and driving incremental power/copper demand in the near term.

Risk assessment: Tail risks include a 30–50% drawdown in mega‑cap AI names if private/public capex funding stalls, or a regulatory shock that restricts cross‑border data flows or cloud services. Time horizons: days–weeks = sentiment and index dispersion; 3–9 months = earnings/capex revisions and guidance changes; 2026–2028 = binary realization of productivity gains vs. stranded asset risk. Hidden dependencies: grid capacity, long‑dated cloud contracts, and utility pricing; key catalysts to watch are Fed moves, CPI surprises >0.5% m/m and any federal/state AI regulation in the next 30–60 days.

Trade implications: Direct actionable plays are infrastructure and real‑asset exposure rather than pure application names. Establish a tactical 2–3% long in ORCL (infrastructure revenue) sized to capture a 25–35% upside by end‑2026 while using 3‑month QQQ 5% OTM put spreads (~0.75% portfolio notional) as tail protection against a 15–25% tech shock. Trim 1–2% absolute exposure to top‑10 concentrated S&P/QQQ over 30 days and redeploy into EQIX/DLR and regulated utilities for 12–24 month holds; buy 9–12 month ORCL call spreads to play upside while capping premium.

Contrarian angles: The market underprices the durability of multi‑year infrastructure contracts — winners with locked colocation/cloud contracts (ORCL, EQIX) may show steadier cashflow than headline AI app winners. The consensus that AI will immediately lift broad productivity is likely premature; history (late‑90s tech concentration) shows infrastructure winners emerge despite many failures. Unintended winners include grid modernization and energy storage suppliers if data‑centre builds proceed; options skew currently underprices a tail-event in concentrated tech, presenting cheap hedging and long‑dated asymmetric long infrastructure opportunities.

Key Decisions for Investors

• Establish a 2.5% long position in ORCL (Oracle) over the next 30 days, target 25–35% upside by 31‑Dec‑2026 if Stargate ramps; initial stop‑loss at 18% below entry and scale half the position on any >10% pullback.
• Trim 1–2% absolute weight from top‑10 concentrated equity exposure (reduce QQQ/SPY top‑10 overweight) within 30 days and redeploy proceeds into data‑center REITs: allocate 1% to EQIX and 1% to DLR for a 12–24 month hold to capture lease/capex pass‑through.
• Purchase a 3‑month QQQ 5% OTM put spread sized to ~0.75% of portfolio notional as a tail hedge against a 15–25% tech drawdown; if VIX spikes >40% convert into longer‑dated VIX calls.
• Set a conditional short trigger: if Stargate/partners or ORCL announce >12‑month revenue delay or aggregate corporate AI capex guidance is cut >20% y/y in the next two quarters, initiate a 1.5–2% short via NVDA 6–9 month 10% OTM puts (or equivalent) to play re‑rating risk.