# The AI Put That Wasn't: A Time Traveller's Report from January 2026 *Gold Coin Club · Gold Rush · 2026-10-03 · by James Hurst (https://www.hurst.world)* ## The Setup Picture this: a hoarder in late summer 2025, staring at a portfolio where equity exposure carries an implicit bet that AI capital expenditure by hyperscalers will continue to justify stretched valuations across the entire market. The Bank of England had just published a working paper flagging AI-linked credit risk as a transmission channel for correlated stress. KKR's credit team had issued a memo warning that leveraged AI bets were being financed in ways that could amplify shocks. The consensus held that AI spending was durable, that any correction would be shallow, that the 'AI put' under equities was a structural floor. A hoarder runs a thought experiment: what if the thing everyone missed in August 2025 was that the AI put was actually a fragility amplifier? What if rotating fifteen percent of equity exposure into gold—the Play labeled gr-1791021705828 in the Hoard Engine—was worth simulating as one branch of the distribution tree? This is that simulation, narrated from a vantage point eighteen months forward, clearly framed as one possible path, never a prediction. ## The Consensus The consensus read in mid-2025 was that AI capital expenditure had become the backstop for corporate earnings and equity valuations. As long as Microsoft, Amazon, Google, and Meta kept building data centers and buying chips, the spending would flow through to semiconductor firms, utilities, construction, and a widening circle of suppliers. The 'AI put' was the idea that any equity wobble would be shallow because the capex cycle was durable and broadly stimulative. Analysts pointed to multi-year build-out timelines, to the strategic imperative of AI leadership, to the fact that hyperscalers had balance sheets strong enough to sustain spending even if monetization lagged. The narrative was that AI investment was different from past tech bubbles because the buyers were profitable giants, not venture-backed startups. Corrections would be buying opportunities because the fundamental driver—AI buildout—was structural, not speculative. ## The Crazy Read The contrarian read was that the 'AI put' under equities was actually a call-away waiting to happen. When a single narrative becomes the consensus justification for holding stocks, any disappointment in that narrative triggers correlated selling across the entire market, not just the sector in question. The Bank of England working paper and the KKR credit memo were early signals that the transmission mechanism was already being constructed: leveraged bets on AI-adjacent credits, covenant-lite loans to data center developers, structured products linked to semiconductor earnings. If AI monetization timelines slipped—if the revenue growth from AI products lagged the capex by even a few quarters—the unwind would start in credit markets and cascade into equities. The thing everyone missed in August 2025, in this branch of the tree, was that the AI put was not a floor but a fragility amplifier. The more the market relied on one narrative, the worse the risk-reward for broad equity exposure became. A hoarder looking back from January 2026 might have seen a drawdown that felt worse than many had expected, because the positioning was more crowded and the leverage was in credit markets where forced selling is less forgiving than equity redemptions. The hypothetical Play was to rotate fifteen percent of equity exposure into gold, not because gold was certain to rally, but because the correlation structure in a narrative-driven equity drawdown could favor uncorrelated hard assets over stocks that all shared the same implicit AI bet. ## The Machine Says Across 10,000 bootstrap paths (seed 229667), this play moves the reference hoard's 15-year median from $549,529 to $545,732 ($-3,797), and its goal probability from 6.3% to 4.4%. The median terminal outcome falls by a few thousand dollars, a rounding error in a fifteen-year horizon. The goal probability drops by nearly two percentage points, which in the Engine's logic means the reallocation shifts probability mass away from the upper tail. The fifth percentile, however, moves from $283,293 to $320,439, an improvement of more than thirty-seven thousand dollars in the worst-case scenarios. The ninety-fifth percentile max drawdown shrinks from 38.8 percent to 31.5 percent, a reduction of more than seven percentage points in peak pain. The distribution the machine draws is one where the hoarder gives up some upside in exchange for a fatter left tail and shallower drawdowns. The Play is a bet that the paths where equities suffer a sharp, correlated selloff are non-trivial in probability, and that gold's behavior in those paths is different enough to cushion the blow. The Engine does not say this will happen; the Engine says that if one believes the AI call-away scenario is a meaningful branch of the tree, this is what the distribution looks like when that branch gets weighted in. > Across 10,000 bootstrap paths (seed 229667), this play moves the reference hoard's 15-year median from $549,529 to $545,732 ($-3,797), and its goal probability from 6.3% to 4.4%. ## The Ways This Is Wrong First way this is wrong: the AI capex cycle could prove more durable than the contrarian read assumes. Hyperscalers might tolerate years of capex leading revenue if the strategic stakes are high enough, and the credit markets might finance that patience without stress. The 'AI put' could turn out to be real, and the hoarder who rotated into gold would have given up equity upside for nothing. Second way: gold and equities could fall together if the AI call-away coincides with sticky inflation or a regime where central banks tighten into weakness. The bootstrap paths that generate the fifth-percentile improvement from $283,293 to $320,439 assume gold behaves as a crisis hedge, but that correlation is itself a scenario bet, not a law of nature. If real rates rise or if gold faces its own selling pressure, the hedge fails and the hoarder is left with lower median outcomes and no drawdown protection. Third way: the Bank of England working paper and the KKR memo might be directional signals without the magnitude to justify claims of systemic fragility. The comparison to dot-com-era concentration is rhetorical unless someone can quantify that current AI-linked credit exposure approaches or exceeds the equity concentration that drove the 2000-2002 drawdown. The data cited does not yet earn that comparison. Fourth way: fifteen percent might be too much or too little. The Play is a point estimate in a space that is itself a distribution. A hoarder running this in the Engine could test five percent, ten percent, twenty percent, and find that the drawdown reduction saturates or that the median cost grows non-linearly. The single Play does not explore the sensitivity surface. ## What Would Change Our Mind Evidence that would weaken the contrarian read: AI revenue growth at hyperscalers accelerating faster than capex, so that monetization timelines compress and the narrative bet pays off within quarters, not years. Credit spreads on AI-adjacent debt tightening rather than widening, indicating that lenders see durable cash flows, not speculative positioning. Data showing that AI capex is being financed out of operating cash flow with minimal leverage, so that the credit transmission channel the Bank of England flagged does not build. A broadening of equity leadership away from AI-dependent names, so that the market no longer relies on a single narrative and correlated selling risk falls. Evidence that would strengthen the contrarian read: a hyperscaler cutting or pausing AI capex guidance, even modestly, and the equity market reaction being sharp and broad rather than contained to tech. Credit rating downgrades or covenant breaches in AI-linked structured products, signaling that the leverage is real and the unwind has started. Historical analysis showing that when a single narrative dominates equity valuations, the subsequent drawdowns are deeper and longer than periods of mixed leadership. A quantification of AI-linked credit exposure that approaches or exceeds the equity concentration seen in prior bubble peaks, giving the dot-com comparison empirical weight it currently lacks. --- Gold Coin Club is an idea machine, not an adviser. Nothing here is financial advice. Every projection is a distribution, not a promise.