We use highly accurate density functional calculations to study the band structure and Fermi surfaces of NbSe2. We calculate the real part of the non-interacting susceptibility, ℜχ0(q), which is the relevant quantity for a charge density wave (CDW) instability and the imaginary part,ℑχ0(q), which directly shows Fermi surface (FS) nesting. We show that there are very weak peaks in ℜχ0(q) near the CDW wave vector, but that no such peaks are visible in ℑχ0(q), definitively eliminating FS nesting as a factor in CDW formation. Because the peak in ℜχ0(q) is broad and shallow, it is unlikely to be the direct cause of the CDW instability. We briefly address the possibility that electronelectron interactions (local field effects) produce additional structure in the total (renormalized) susceptibility, and we discuss the role of electron-ion matrix elements.In 1964 V.L. Ginzburg proposed excitonic superconductivity in quasi-2D structures 1 composed of metal layers sandwiched between insulating layers. By that time, only two layered materials were known to be superconducting, 2,3 PdTe 2 and NbSe 2 . In the four decades since then, NbSe 2 and isostructural selenides have been intensively investigated (close to 1.5 thousand papers published to date). However, the main interest in this compound has shifted from the fact that it is a layered material (hundreds of layered superconductors are now known), to the existence of a nearly-commensurate charge density wave (CDW) 4 instability and it's possible interplay with the superconductivity that sets in at a lower temperature. We mainly forego discussion of the interesting issues surrounding the superconducting state, its origin, and its relationship to the CDW state, and instead concentrate on the mechanism behind the CDW transition itself.The first electronic structure calculations for NbSe 2 were presented by Mattheiss in 1973 5 . Using a non-selfconsistent potential he was able to produce a band structure with basic features in reasonable agreement with more recent self-consistent calculations 6,7,8 , but which showed only two bands crossing the Fermi energy (it is now known that there are three), and underestimated the energy depth of a saddle point at ∼ 1 2 ΓK. Fermi surfaces based on this band structure led to early suggestions that the CDW transition was driven by nesting 4 , an assumption that has carried through to the present time. The nearness of the saddle point to the Fermi energy (E F ) led Rice and Scott 9 to argue that CDW formation was driven, not by Fermi surface (FS) nesting in the conventional sense, but rather by saddle points lying within k B T CDW of E F and separated by the CDW wavevector, Q CDW = ( 1 3 , 0, 0). A significant amount of effort has been spent on resolving the 'controversy' between the nesting and saddle point theories for NbSe 2 and for related CDW compounds such as 2H-TaSe 2 , 1T-TaSe 2 , TaS 2 and others, but no specific feature that would give rise to an instability at Q CDW has been convincingly isolated. As early as 1978, Doran et al 10...