Thirteen lines of the CO band near 4.7 μm have been observed on a jovian hot spot at a resolution of 0.045 cm -1. The measured line profiles indicate that the CO mole fraction is 1.0±0.2 ppb around the 6-bar level and is larger in the upper troposphere and/or stratosphere. An external source of CO providing an abundance of 4 +3-2×10 16 molecules cm -2 is implied by the observations in addition to the amount deposited at high altitude by the Shoemaker-Levy 9 collision. From a simple diffusion model, we estimate that the CO production rate is (1.5-10)×10 6 molecules cm -2 s -1 assuming an eddy diffusion coefficient around the tropopause between 300 and 1500 cm 2 s -1. Precipitation of oxygen atoms from the jovian magnetosphere or photochemistry of water vapor from meteoroidal material can only provide a negligible contribution to this amount. A significant fraction of the CO in Jupiter's upper atmosphere may be formed by shock chemistry due to the infall of kilometer- to subkilometer-size Jupiter family comets. Using the impact rate from Levison et al. (2000, Icarus143, 415-420) rescaled by Bottke et al. (2002, Icarus156, 399-433), this source can provide the observed stratospheric CO only if the eddy diffusion coefficient around the tropopause is 100-300 cm 2 s -1. Higher values, ∼700 cm 2 s -1, would require an impact rate larger by a factor of 5-10, which cannot be excluded considering uncertainties in the distribution of Jupiter family comets. Such a large rate is indeed consistent with the observed cratering record of the Galilean satellites (Zahnle et al. 1998, Icarus136, 202-222). On the other hand, the ∼1 ppb concentration in the lower troposphere requires an internal source. Revisiting the disequilibrium chemistry of CO in Jupiter, we conclude that rapid vertical mixing can provide the required amount of CO at ∼6 bar for a global oxygen abundance of 0.2-9 times the solar value considering the uncertainties in the convective mixing rate and in the chemical constants.