Excess Returns
You’re Getting AI Wrong | GMO’s Tom Hancock on Finding Conviction Amid the Hype
Most Important Insight
The market is mistakenly valuing AI as a standalone sector rather than a productivity tool that will primarily benefit established 'Quality' companies with proprietary data and existing distribution moats.
Most Original Insight
AI disruption is more likely to compress margins for pure-play software providers through code commoditization than it is to unseat legacy incumbents who own the customer relationship and specialized datasets.
Key Points
- Quality stocks, defined by high return on equity and low leverage, provide a superior risk-adjusted way to play AI compared to speculative hardware or model-builder plays.
- The 'AI Tax'—the capital expenditure required to stay competitive—will be a significant drag on the cash flows of second-tier tech firms through 2027.
- Proprietary data is the only durable moat in an era where large language models have democratized sophisticated reasoning and coding capabilities.
- Valuation spreads between the highest-quality compounders and the rest of the market remain attractive despite the 2025-2026 rally in mega-cap tech.
- Investors should distinguish between 'AI enablers' like semiconductor firms and 'AI adopters' who will see long-term margin expansion from operational efficiencies.
- The risk of 'over-earning' in the semiconductor space is high as the initial build-out phase of AI infrastructure reaches a plateau by late 2026.
- Quality companies in non-tech sectors, such as healthcare and professional services, are the most undervalued beneficiaries of AI-driven cost reductions.
Investment Implications
| Asset / Sector / Instrument | Action | Source | Notes |
|---|---|---|---|
| Global Quality Equities | BUY | explicit | Hancock argues these firms have the balance sheets to survive the 'AI transition' and the scale to implement it profitably. |
| Healthcare Information Services | BUY | implicit | Identified as a sector with high-quality characteristics and proprietary data that AI can leverage to significantly lower COGS. |
| NVIDIA (NVDA) | HOLD | implicit | While acknowledging its leadership, the warning about a plateau in infrastructure build-out by late 2026 suggests caution on further upside. |
| High-Leverage Growth Stocks | SELL | explicit | Hancock emphasizes that in a 'higher for longer' rate environment, the lack of self-funding capability is a terminal risk during AI disruption. |
| Legacy SaaS Providers | SELL | implicit | Hancock suggests that AI commoditizes code, potentially eroding the high margins and high switching costs of traditional software seats. |
Hang on a sec…
- Hancock claims that proprietary data is an 'unassailable moat,' yet ignores the rising efficacy of synthetic data which could allow competitors to train models without legacy datasets.
- The assertion that Quality stocks are 'attractively valued' overlooks that many Quality indices are currently trading at their highest P/E premiums relative to Value in a decade.
- He suggests AI will primarily benefit incumbents, but historical technological shifts (like the internet) frequently saw incumbents burdened by 'technical debt' while nimble startups captured the new value.