US AI Complex vs Nifty IT: Correlation Study — Preliminary Findings
Prepared 4 August 2026.
Question and method
Does the US AI/semiconductor complex show negative correlation with Nifty IT — the substitution thesis being that AI capex displaces spend on outsourced IT services? Tested on daily log returns, 1 Aug 2023 to 4 Aug 2026. US basket is an equal-weighted average of NVIDIA, Micron, AMD, SOXX and KOSPI. India side is the Nifty IT Total Returns Index, used directly. Pearson correlation on matched trading days, no forward-fill across holidays.
Headline result
Full sample does not support a stable negative correlation. The reading is weak and positive: common tech-sector risk has dominated any AI-substitution effect over the three-year window.
Same-day correlation: 0.105
T+1 correlation (US day t, India day t+1): 0.063
R²: 1.1%
p-value: 0.006
n (paired trading days): 690
The relationship is regime-dependent
A single full-sample number hides the pattern. Broken into sub-periods, correlation stayed positive through most of the sample and only turned negative in the most recent stretch.
Pre-AI narrative (Aug ’23 – Dec ’24): n=314, correlation 0.149
Nifty IT topping (Dec ’24 – Feb ’26): n=266, correlation 0.260
AI shock (Feb – Jun ’26): n=87, correlation 0.073
AI trade falters (Jul – Aug ’26): n=24, correlation −0.433
60-day rolling correlation
Across the full sample, the 60-day rolling correlation ranged from −0.19 to +0.48. It stayed in a broadly positive 0.1–0.4 band for most of the three years, spiked to its highest level in February 2026, then fell sharply from around April 2026 onward. The current reading, −0.19, is the lowest of the entire sample.
How recent is the shift
Correlation from a fixed start date to today, at three cutoffs. Strengthens and only clears statistical significance in the most recent month; n shrinks fast as the window narrows, so the July figure should be read as indicative, not conclusive.
Jan 1, 2026: n=130, correlation −0.068, p=0.445, not significant at 95%
Apr 1, 2026: n=77, correlation −0.186, p=0.105, not significant at 95%
Jul 1, 2026: n=24, correlation −0.433, p=0.035, significant at 95%
Event study
Cumulative return in the five trading days before and after each catalyst date. 4 of 6 events show the two baskets moving in opposite directions; the 22 July divergence is the largest in the sample.
4 Feb 2026 — AI model launch, automation-fear selling begins. US AI +6.3%, Nifty IT −3.5%. Diverged: yes
12 Feb 2026 — Infosys, Wipro, TCS from 4 Feb 2026 — AI model launch, automation-fear selling begins. US AI +6.3%, Nifty IT −3.5%. Diverged: yes
12 Feb 2026 — Infosys, Wipro, TCS fall on AI disruption fears. US AI +1.4%, Nifty IT −3.5%. Diverged: yes
30 Jun 2026 — Nifty IT lowest since Apr 2023, −43% from peak. US AI −9.8%, Nifty IT +6.0%. Diverged: yes
2 Jul 2026 — Sharp IT sector rebound. US AI +3.8%, Nifty IT +1.9%. Diverged: no
22 Jul 2026 — Alphabet raises AI capex guidance, US tech sells off. US AI −19.7%, Nifty IT +8.7%. Diverged: yes
29 Jul 2026 — Indian IT rallies as global AI trade falters. US AI +8.8%, Nifty IT +1.9%. Diverged: no
Caveats
• Returns are in local currency, USD and INR. Not FX-adjusted; rupee moves are embedded in the India series.
• Same-day correlation is not contemporaneous. Indian markets close before the US opens. The T+1 test is the more defensible one, and it reads weaker, not stronger, than same-day.
• Correlation is not causation. A negative reading is consistent with the substitution thesis; it does not confirm it.
• Event dates were chosen because the sector moved on them. That guarantees a reaction and proves nothing about an ordinary day.
• Nifty IT is float-weighted and dominated by TCS and Infosys. The US basket is equal-weighted. The two sides are not symmetric.
Bottom line
Three years of data do not support a stable negative correlation between the US AI complex and Nifty IT. The relationship was mildly positive for most of the period. A negative correlation has emerged since roughly April 2026 and has strengthened since, reaching significance only in the past month. Treat this as an emerging regime to monitor, not a proven hedge relationship.
Data: NVIDIA, Micron, AMD, SOXX, KOSPI adjusted close (EODHD); Nifty IT Total Returns Index (NSE), as supplied.