In a white paper 1 published in October 2026, Orbis Investment Management's Simon Skinner, Head of Global Investment Team, and John Christy, Senior Investment Specialist, advance a thesis that cuts sharply against the prevailing consensus on AI and capital markets: rather than correcting the irrational herd behaviour that has defined investing for generations, artificial intelligence is more likely to amplify it.
Narrative, Not Truth
The paper opens with the instructive case of Zoom and ExxonMobil. In 2020, lockdowns and the collapse in oil demand briefly made Zoom worth more than one of the world's largest energy companies. Zoom shares subsequently shed roughly 90% of their value while ExxonMobil more than quadrupled. Skinner and Christy invoke Nobel laureate Robert Shiller's Narrative Economics as the explanatory framework: stories spread not because they are right, but because they are vivid and repeatable. The investors who bid up Zoom had more data and better technology at their fingertips than any previous generation. None of it mattered.
Turbo Sheep
Drawing on Yuval Noah Harari's Nexus, the paper argues that information technology amplifies prevailing narratives rather than correcting them. The first runaway bestseller after Gutenberg was a witch-hunting manual; Copernicus had to wait nearly a century. Skinner and Christy apply the same logic directly to AI. "AI may simply create herds of 'turbo sheep' that form faster and run harder than ever in the same direction," they write. The core risk is not that AI processes information badly. "The real danger is that it makes us all think less independently, move faster, and act with greater conviction." If investors outsource their thinking to AI, the reflexive feedback loop that drives every bubble -- price drives narrative, narrative drives price, confirmation bias floods capital in -- simply spins faster and with greater momentum.
Markets Are Not Getting More Efficient
If decades of better technology had made markets more efficient, the historical record should show narrowing mispricings by now. It shows the opposite. Valuation dispersion between the most and least expensive stocks in the FTSE World Index remains near its widest since 1990. The pandemic-era Everything Bubble -- in which meme stocks, NFTs and negatively yielding government bonds inflated simultaneously -- arrived at the very peak of technological sophistication. "Despite having the most technologically sophisticated investors in history," Skinner and Christy write, "we are only a few years removed from one of the most dramatic bubbles in history." The widest valuation spreads in the historical record have clustered around the Japan bubble, the Dot-com era and the Everything Bubble -- all later revealed as periods of narrative excess rather than rational pricing.
The Passive Problem
Passive vehicles now hold 56% of global equity fund assets, up from roughly 16% two decades ago. Even the active bucket is misleading: a 2025 study found that 38% of U.S. active fund assets sat with closet indexers by end of 2019 -- portfolios built to not lose clients and portfolio management jobs in the near term, not to win. Markets have narrowed to their most concentrated levels since the TMT bubble, with gains compressing into a handful of mega-caps. AI applied to this already herd-driven environment may do what every previous market technology has done: speed up herding, not slow it down.
The SPIVA Case Revisited
The paper challenges the SPIVA scorecard, the most widely cited evidence in favour of passive investing. A 2026 working paper by Cremers, Fulkerson and Riley identifies three methodological distortions that each tilt SPIVA's results against active managers: survivorship bias, equal fund weighting regardless of asset size, and comparison against a fee-free theoretical benchmark rather than the real investable passive funds an allocator would actually buy. Correcting these distortions roughly doubles reported active outperformance rates -- from 16% to 30% in the U.S., from 10% to 31% globally, from 15% to 45% in international markets and from 13% to 28% in emerging markets over the ten years to December 2024. Four characteristics recur among managers most likely to add genuine value: meaningful differentiation from the benchmark (high active share), concentration in highest-conviction ideas, idiosyncratic stock-selection skill as distinct from factor exposure, and genuine alignment through manager co-investment.
The Durable Edge
At Orbis, more than 35 years of proprietary paper portfolio decision-making data -- trade rationale, earnings forecasts, investment meeting votes and transcripts -- is being mined with AI for insights that were previously cumbersome or impossible to extract at scale. General AI productivity gains are, in the authors' view, table stakes: necessary to keep playing the game but insufficient to create a winning edge. "A truly durable edge has to come from something that the new technology can amplify but competitors cannot replicate," Skinner and Christy write. The conclusion follows directly: "Independent human judgement will remain the most enduring source of alpha and will only become more valuable in the age of AI." The payoff to skill and independent thinking is, the paper argues, often largest precisely where the fewest investors are exercising it.
5 Key Takeaways for Advisors and Investors
- AI is more likely to accelerate market bubbles than to prevent them. The reflexive feedback loop that drove Zoom's inflation in 2020 -- price moves narrative, narrative drives price -- will spin faster and with greater conviction when investors outsource their thinking to AI.
- Valuation dispersion remains near historical extremes. Decades of increasingly sophisticated technology have not made markets more efficient; the pandemic-era Everything Bubble arrived at peak technological sophistication.
- The SPIVA case against active management is weaker than the headlines suggest. After correcting for survivorship bias, fund weighting and benchmark methodology, active outperformance rates roughly double across most categories.
- Genuine independent price discovery is scarcer than it appears. With 56% of equity assets in passive vehicles and a further 38% of "active" assets in closet indexers, the pool of truly independent thinking in markets is narrow -- and narrowing.
- Proprietary data and independent judgement are the scarce resources of the AI age. Advisors and allocators should prioritize managers whose edge stems from proprietary insights and cultures of independent thinking -- advantages that AI cannot commoditize and competitors cannot replicate.
Footnote:
1 Skinner, Simon, and John Christy. "New Tools, Old Habits: Investing in the Age of AI." Orbis Investment Management, Oct. 2026, https://www.orbis.com/documents/ai-whitepaper-us.pdf.
