There are two hypotheses about what a crash in artificial intelligence (AI) shares does to Indian equities. The first is exposure. Investors sell everything risky, the rupee weakens and Indian shares fall with the rest. The second is shelter. India has almost no listed company that makes AI chips, memory or servers, so money leaving such shares comes here.
I tested both with daily prices from January 2023 to November 13, 2026. I isolated the part of the Philadelphia Semiconductor Index that the broad American market does not explain. I then asked how each country’s shares behave on that measure’s worst 10 per cent of days, after allowing for oil, the dollar and American bond yields.
Korea’s sensitivity is 0.78. Taiwan’s is 0.54. India’s is 0.03, statistically no different from zero. Between October 6 and October 30, semiconductor shares fell 19.1 per cent and Brent crude rose 17.4 per cent. The Nifty 50 gained 1.2 per cent. Shelter wins.
Is that reassuring? Partly, and two further tests strengthen the case. The first separates good days from bad. On days the AI measure falls, India’s sensitivity is 0.02. On days it rises, the sensitivity is 0.09, small but statistically significant. Korea is the reverse, 0.73 on the way down and 0.57 on the way up. India takes a little of the upside and none of the downside. An options trader would call that convexity, and would pay for it.
The second test separates a chip rout from a general rout. Since 2023, there were 72 sessions that followed a worst-decile AI day on which the broad American market held. The Nifty rose 0.19 per cent on average. There were 17 on which the American market also fell by more than 1 per cent. The Nifty fell 0.46 per cent, much as it does on any bad American day.