The biggest mistake CEOs make with AI has nothing to do with the technology

Key event: the Trump Administration issued an Executive Order and a follow-on national AI framework intended to create federal guardrails and avoid a patchwork of '50 different state rules'. The author argues AI will transform industries but is unlikely to destroy established software firms — leadership judgment, not consensus, will determine winners. For portfolios, regulatory clarity would be modestly positive for AI innovators, but impacts are sector- and company-specific; companies that integrate AI strategically and act early are more likely to outperform peers who simply copy the crowd.

Analysis

The AI wave will bifurcate economics across three dimensions: model providers (compute + IP), application owners with proprietary workflows/data, and low-moat integrators. Expect large-scale customers to internalize model inference for high-value, high-frequency workflows (finance, search, ops), shifting 20–40% of incumbent SaaS renewal dollars into platform customization and compute spend over 12–24 months; companies that monetize proprietary training data will capture the lion’s share of incremental margin.

Second-order supply-chain effects will show up in capital intensity and energy: data-center PUE and GPU inventory dynamics can swing vendor margins quickly. A sustained shift to on-prem or colo inference for latency-sensitive use cases could lift colo utilization by 5–15% and accelerate orders for accelerator hardware in discrete 3–9 month waves, producing boom/bust capex signals for semiconductor suppliers and real-estate landlords.

Regulatory outcomes and implementation friction are the largest path-dependent risks — either raising barriers (benefiting incumbents with compliance scale) or enabling competition via open-source standards that collapse model pricing. Probability-weighted planning should assume both states: allocate portfolio exposure to capture incumbents’ balance-sheet advantages over 12–36 months while keeping optionality for rapid de-risking if open-source commoditization accelerates within a 6–18 month window.

Key Decisions for Investors

• Overweight Microsoft (MSFT) 6–18 months: buy an outright position or a 0.75–1.25 year call spread (e.g., buy 12-month 10% OTM calls, sell 24-month 25% OTM calls) to express platform/enterprise AI monetization while capping carry. R/R: asymmetric — limited premium vs 15–30% upside if enterprise adoption and M365/LinkedIn monetization accelerate; tail risk is execution/consent setbacks.
• Long Nvidia (NVDA) in 3–9 months to capture discrete GPU demand surges from inference and retraining waves — prefer leap calls or buy-write to manage entry. R/R: high upside from continued accelerator shortages (>20–50%) but subject to inventory-driven retrenchment; cap position size to 3–5% of tech sleeve.
• Pair trade: long data-center/colo REITs (e.g., EQIX) and short low-margin IT services/integration names (select small-cap consultancies) over 12–24 months. Mechanism: rising on-prem/colo demand increases REIT cashflows while commoditization pressures service revenue and pricing; target spread capture of 10–15% annualized, hedge with IG duration exposure for macro risk.
• Hedge for open-source commoditization: buy out-of-the-money puts on the AI hardware/infra longs (NVDA, AMZN) with 6–12 month tenors sized to limit portfolio drawdown to <=6% if model pricing collapses rapidly. This preserves upside while protecting against a fast re-rating triggered by free/open models.