AI investors face two fundamentally different types of risk — one that shrinks the total value created by AI, and another that merely reallocates it among winners and losers — according to a new Goldman Sachs framework that warns current AI stock gains already approach the upper bound of even the most optimistic scenarios.
AI investors face two fundamentally different types of risk — one that shrinks the total value created by AI, and another that merely reallocates it among winners and losers — according to a new Goldman Sachs framework that warns current AI stock gains already approach the upper bound of even the most optimistic scenarios.
"Size shocks threaten the total value AI creates; distribution shocks change who captures it, leaving the aggregate unchanged," the bank's global markets team wrote in a July 23 commentary.
AI-related stocks, including private companies, have added roughly $26 trillion in market capitalization since November 2022, or about $23 trillion after deducting baseline returns. Under Goldman's baseline assumptions, the present discounted value of US corporate profits from AI productivity gains totals only about $9 trillion. Even under the most optimistic combination of assumptions — faster productivity growth, quicker adoption, and a larger capital share — that figure reaches only about $28 trillion, meaning current valuations have already priced in near-best-case outcomes.
The distinction matters because the two risk types travel through markets along completely different paths — and require completely different hedging tools. Size shocks — triggered by slower adoption, tighter financial conditions, or macro disruptions — push down broad equity indices, spike volatility, widen credit spreads, and typically lower Treasury yields. Distribution shocks, by contrast, leave macro assets flat while driving violent sector rotations beneath the surface.
Five market events over the past 18 months illustrate the pattern clearly. Three size shocks — the DeepSeek selloff in January 2025, broad AI valuation concerns in February 2026, and a nonfarm payroll-driven rate spike in June 2026 — each pushed the S&P 500 down 1.5 percent to 2.6 percent, sent the VIX surging 20 percent to 40 percent, and widened credit spreads. Two distribution shocks — Google's June 2 announcement of an equity offering to fund AI capital expenditure, and Meta's July 1 plan to build a cloud business selling excess compute capacity — left the S&P 500 and Treasury yields essentially unchanged. But beneath the surface, semiconductors surged 5.8 percent on the Google news while hyperscalers fell 2.4 percent; the Meta announcement flipped the script entirely, with hyperscalers gaining 2.5 percent and semiconductors dropping 6.4 percent.
The Market Is Already Signaling Which Risk Dominates
Current options market data tells the same story. S&P 500 single-stock implied volatility sits at the 99th percentile of its 15-year range, while implied correlation between stocks has fallen to the lowest level in 15 years. That combination — extreme individual stock uncertainty paired with near-zero co-movement — is precisely what distribution-shock dominance looks like. Individual names are pricing in dramatic potential moves, but the market as a whole sees no reason to reprice AI's aggregate value.
The DeepSeek event, which some observers classified as a distribution shock because it lowered AI innovation costs, actually belongs in the size-shock category from a US equity perspective, Goldman argued. While cheaper AI models redistribute value to consumers globally, they also reduce the share of AI profits captured by US corporations — effectively shrinking the cake for US-listed stocks.
Hedging One Risk Is Far Harder Than the Other
For broad equity investors holding the S&P 500, size shocks are the primary threat — and they are relatively hedgeable. Macro assets respond predictably: stocks fall, bonds rally (unless rates themselves are the shock), and cross-asset diversification works. Distribution shocks, by contrast, are self-canceling at the index level, meaning traditional macro hedges offer little protection.
For investors with concentrated AI exposure or overweight positions, the situation is more difficult. Both risk types hit concentrated portfolios, but distribution shocks are far harder to hedge because the assets that would serve as natural offsets — competing AI stocks, sector ETFs, or macro instruments — do not reliably move in the opposite direction when the reallocation occurs.
"The question every AI investor should ask is not just 'what could go wrong?' but 'what kind of wrong?'" Goldman's team wrote. "The answer determines whether a simple S&P put works — or whether nothing in your hedging toolkit does."
This article is for informational purposes only and does not constitute investment advice.
