After Hirzebruch's generalization of the classical Riemann-Roch formula, Grothendieck extended this result to a relative version. Then Atiyah and Hirzebruch gave a "differentiable analogue" of Grothendieck's theorem, which is called Atiyah-Hirzebruch's Riemann-Roch theorem. In spite of the original topological proof, I will present another one, in a more differential-geometric flavor, using Mathai-Quillen's construction of the Thom form.