The high‐frequency spectral noise voltage appearing across a metallic conductor electrically heated to high temperatures is experimentally found to be in agreement with the Nyquist theorem even though the conductor is far from thermodynamic equilibrium. Furthermore, because of the thermal origin of the nonlinearities in the current‐voltage characteristic the proper value of resistance to use in the Nyquist expression corresponds to the voltage/current ratio, rather than the differential resistance at the operating point. At frequencies below that corresponding to the thermal time constant of the conductor, an additional noise voltage caused by temperature fluctuations is detected. The observed magnitude and spectrum of this noise voltage is in very good agreement with a simple calculation of temperature fluctuations due to heat conduction.