AGI Bust, AI Boom: Why AI's Winners And Losers Aren't Who You Think

Analyst argues the market is overvaluing AGI potential and expects LLM progress to plateau, leading to diminishing returns on AI infrastructure investment. Recommends strong buys in BRK.A/BRK.B and UNH and buy positions in TRI, SPGI, and MCO, citing these firms' ability to drive efficiency and productivity gains by adopting current AI tools; this is an actionable idea for stock selection rather than a market-wide trade.

Analysis

Incumbent, cash-rich, recurring-revenue businesses are set to capture the next wave of productivity gains because their bottlenecks are process integration and data plumbing — not raw model scale. Expect 100–300bps of operating-margin tailwind over 12–36 months for firms that automate high-frequency, rule-based workflows (claims, risk models, pricing engines), with most near-term delta coming from FTE redeployments and third-party spend compression rather than headline revenue lifts.

Second-order winners include rating and information providers (faster report generation, fewer contract renewals lost to price negotiation) and vertically integrated insurers/healthcare payors (claims adjudication, utilization review). Conversely, vendors that sell one-off AI infrastructure upgrades or rely on perpetual hardware refresh cycles face demand compression if model training economics plateau; capex profiles for hyperscalers and high-end GPU cycles are the key supply-chain lever to monitor over the next 6–18 months.

Primary reversal risks are technological (a >10x drop in training/inference cost from a new architecture) or regulatory (data-localization / liability rules that fragment training datasets), either of which would re-accelerate infrastructure spend and revalue the incumbents downward. Near-term sentiment shocks (earnings misses tied to integration timing) can create 5–15% drawdowns in names that have already priced in productivity gains, presenting tactical entry points.

Key Decisions for Investors

• Long BRK.B (or BRK.A) — 12–36 month holding. Thesis: optionality + dry powder to deploy into mispriced assets if AI capex stalls; target 20–30% total return vs S&P over 2 years. Size: 3–5% NAV; risk: operational underperformance or insurer losses; stop-loss at -20% from entry or trim into strength.
• Long UNH — 12–24 month holding. Thesis: 150–300bps margin expansion from automating claims and utilization management; expect 15–25% upside if execution on care-management initiatives stays on cadence. Risk management: buy 1.0–1.5 year ITM calls financed by selling short-dated OTM calls (buy-write) to improve carry.
• Long SPGI + MCO pair — equal notional, 6–18 month timeframe. Thesis: information and ratings franchises monetize higher throughput with minimal marginal cost; prefer call spreads (buy 12-month $X calls, sell higher strike) to cap cost. Hedge: pair against a small short position in an AI-infrastructure ETF or 1–2% notional put protection to guard against tech-capex resurgence.
• Tactical: use 3–6 week earnings windows as re-entry points — add on 5–10% post-earnings pullbacks that exceed 8–12% intraday, since these are often timing/guide misses not structural demand shifts.