AI Stocks: The HALO Trade As A Hedge Against The AI Sell-Off

• HALO trade stocks like Newmont, Exxon, Valero, Deere, and Comfort Systems have gained 25-30% year-to-date, significantly outperforming major AI stocks.
• The five U.S. AI hyperscalers (Alphabet, Meta, Oracle, Microsoft, Amazon) are projected to spend over $650 billion in capex in 2026 alone, exceeding their total pre-2022 investment.
• Major AI stocks have underperformed in 2025, with Oracle and Microsoft down 13-15%, Amazon down 6%, and Alphabet down 10% from its January breakout despite a 70% gain in 2024.

Summary: The HALO Trade as AI Stock Hedge

Key Concept: The "HALO" trade (Heavy Asset, Low Obsolescence) is emerging as a Wall Street hedge against AI disruption, focusing on companies with substantial physical infrastructure that serves as competitive moats resistant to AI replacement.

Featured Companies and Performance (YTD 2026):

• Energy: Exxon and Newmont up ~25%; Valero up ~30%
• Industrials: Deere and Comfort Systems both up 25%+
• Consumer Staples: McDonald's up 3%

AI Stock Underperformance:

Major AI stocks have struggled in 2026:

• Oracle and Microsoft: worst performers among hyperscalers
• Amazon: down 6%
• Alphabet: down 4% (retreated 10% after January breakout)
• Meta Platforms: down 0.4% YTD (versus 86% gain 2023-2025)

Capital Expenditure Surge:

Goldman Sachs reports the five U.S. AI hyperscalers (Alphabet, Meta, Oracle, Microsoft, Amazon) have invested $1.5 trillion in capex since ChatGPT's 2022 launch—more than double their $600 billion total pre-2022 investment. 2026 alone projects over $650 billion in capex.

Market Implications:

Goldman Sachs analysts note a paradox: the tech sector driving AI innovation faces the greatest obsolescence risk. Markets are now rewarding "capacity, networks, infrastructure and engineering complexity" over digital assets. The Wall Street consensus predicts capital-intensive HALO companies will deliver faster EPS growth and higher ROE than previous forecasts.

The shift reflects investor concern that AI can easily replace software but not sprawling physical infrastructure.