NVDA: factor decomposition (Technology)

How much of NVDA's move over the window is systematic (market, XLK sector-excess, rate-excess via TLT) vs. idiosyncratic (stock-specific).

Return decomposition (log-return points, additive)

Market+38.41ppSector (excess)+36.88ppRate (excess)+0.78ppIdiosyncratic-50.87pp

Factor betas

Beta · Market (SPY)+1.90
Beta · Sector-excess (XLK)+0.89
Beta · Rate-excess (TLT)-0.22
R²0.53
Window2025-08-22 → 2026-09-24 (273 trading days)

Read

R² is 0.53: a reasonably reliable fit; the betas and the split below are worth reading at face value.

Note: the net move here (+25.2pp) is small relative to the systematic and idiosyncratic swings underneath it; they're largely offsetting each other, which is why the idiosyncratic share below exceeds 100% (it's dividing by a small net, not an error). Read the pp values themselves as the meaningful numbers here, not the ratio.

Of NVDA's +25.2pp move over this window, -50.9pp (202% of the total magnitude) is idiosyncratic, not explained by riding market, sector, or rate moves. The rest (+76.1pp) came from systematic exposure: market +38.4pp, sector +36.9pp, rate +0.8pp.