Markets
BTC $55,921 -8.92% ETH $1,549 -9.95% BNB $512 -9.72% XRP $0.8704 -8.93% SOL $65.37 -9.13% LINK $8.08 -7.67% BTC $55,921 -8.92% ETH $1,549 -9.95% BNB $512 -9.72% XRP $0.8704 -8.93% SOL $65.37 -9.13% LINK $8.08 -7.67% TOTAL $1.95T BTC.D 57.2%

Global Economy

Bitcoin Sat Out the AI Rally. Its Record on Nasdaq Down Days Still Matters

DWF Labs found Bitcoin fell on 77% of days when the Nasdaq-100 dropped more than 1%, a pattern that predates the spot ETFs by four years.

Bitcoin has spent the past year failing to participate in the artificial intelligence trade and still moving with it on the way down. DWF Labs reported on Nov. 19 that bitcoin fell on 77% of US trading days when an ETF tracking the Nasdaq-100 lost at least 1%, measured from the start of US spot bitcoin ETF trading on Jan. 11, 2024 through Nov. 13, 2026. Its average decline on those days was about 2%.

The asymmetry is stark over the same stretch. From September 2025 to the end of the data set, the iShares Semiconductor ETF gained 85% while bitcoin fell 39%. Crypto captured almost none of the upside from the companies driving the equity market and appears to have kept the downside beta.

That matters because the Federal Reserve has started discussing the scenario explicitly. In minutes of the Oct. 27-28 meeting released Nov. 18, almost all participants discussed what could happen if returns on AI investments disappointed. Most pointed to rapid growth in AI-related debt and the complex, opaque structures used to finance it. A few noted that exposure was concentrating among a small number of firms, raising questions about how the buildout is funded and what the projects will earn.

Fed staff flagged two mechanisms that could amplify losses. One was off-balance-sheet leases, which leave firms committed to future payments that do not appear on their balance sheets. The other was circular financing among AI-related companies, which ties their funding to one another. In a stress scenario, investors who borrowed to buy assets would have to sell to reduce debt, and fund managers would have to sell to meet withdrawals. Sales by hedge funds, bond mutual funds and target-date funds could make it harder for markets to absorb trades. Elevated valuations and hedge fund leverage were the staff’s biggest listed vulnerabilities, while strong bank and household balance sheets were the buffer.

Some participants questioned whether supervisors had enough data on banks’ indirect exposure through nonbank institutions, such as credit lines to private credit funds. That gap is the kind of thing that turns a repricing into a disorderly one, and it is the channel through which an AI selloff would reach crypto: not through token fundamentals, but through the same liquidity that has been funding the trade.

The historical record is consistent. From 2020 until the ETFs launched, bitcoin fell on 74% of days when the Nasdaq-100 proxy dropped at least 1%, against 77% afterwards, so the correlation was not created by institutional products. Using adjusted closes, bitcoin declined on 91 of 118 such days in the later period, with an average log return of -1.93% across all 118, including days it rose or held flat.

Two caveats belong with those numbers. Closing prices do not capture the order of moves within a session, and bitcoin’s daily candle ends at midnight UTC, three to four hours after the US close, so some of the measured reaction is to news that arrived later. The larger point survives both: if the AI complex unwinds and drags the index with it, bitcoin’s record on those days suggests it will not be a hedge.

Share this article

Viber uses viber://forward?text= when the app is installed.