They’re not directly comparable spec-for-spec, but the key tradeoff is this: image intensifiers amplify a very weak optical signal, while digital sensors try to measure it with very low electronic noise.

From the answers, intensifier tubes generally have lower quantum efficiency than solid-state sensors, so although they produce a bright output image, they also amplify shot noise. Digital sensors can be more efficient at detecting incoming photons, but their performance is limited by read noise and other analog noise before A/D conversion.

High-end solid-state sensors can get read noise below about one electron per read, especially with longer integration times, which makes them extremely capable in low light. In principle, that means digital sensors can rival or exceed intensifiers in some situations, particularly when longer exposures are allowed.

The practical difference is that intensifiers are designed to give a real-time bright view at extremely low light levels. Digital sensors can approach this, but for true passive, real-time night vision they are still constrained by photon counts, read noise, and exposure time. So there’s no simple theoretical barrier that says CMOS/CCD can’t compete, but in practice the use case matters: real-time viewing favors intensifiers, while imaging with some integration time increasingly favors digital sensors.