We review the properties of the non-Gaussian cold spot found in the WMAP data. The spot, which was first found in the WMAP 1-year data at position (b=−57°, l=209°) and subtending ≈10° in the sky, has been now confirmed with the WMAP 3-year data. It is clearly detected with several different statistical methods acting on wavelet coefficients. The probability of finding such a spot in Gaussian simulations is around 1%. The frequency dependence of the spot is flat at a very high precision, rejecting the possibility of being due to the Sunyaev–Zeldovich effect or Galactic foregrounds. Finally, we discuss different possibilities which can help to explain its origin.