I'd like preface my answer with a note that a tripod is not only useful in conjunction with ND filters — it also improves the results from image stacking as well. By fixing the position of the camera, the tripod eliminates changes in perspective, which can occur through minor motion while hand-held shooting a sequence for image stacking.
Aside from fixing the camera's position, the tripod can also eliminate low-frequency motion blur, when correct tripod technique is employed. This allows for long-duration shutter exposures. Note that the two concepts, fixing the camera position with a tripod, and elimination of low-frequency motion by use of a tripod, are not necessarily entirely the same thing.
"Dumb" (but otherwise still very good) image stacking can't account for changes in perspective, which in the most fine-grained terms is the precise location of the camera's entrance pupil. So hand-holding a camera and taking, say, 50 or 100 images, while slightly moving because of body motion, may result in more images in the stack being thrown out. Worse, those images might not be thrown out, and they get averaged into the result, which contributes slightly to lower sharpness.
Conversely some of the latest smartphones, that have 3D depth mapping capabilities in conjunction with really good computational photography algorithms, can theoretically deal with small shifts in perspective (or if not now, then they will be able to soon, at cheaper and cheaper price points). That will certainly improve hand-held image stacks to simulate/recreate long-shutter ND filter shots. But in my opinion, that's no reason to rely on technology to make up for a simple problem in technique. Instead, why not augment the power computational photography algorithms with good tripod technique. Everything you do to help the camera produce the best results it can create will only improve the achievement of your artistic vision. If the weight and bulk of a tripod is a concern, you can look for new tripod technologies, such as Peak Design's recent compact travel tripod
Regarding ND filters vs image stacking, I'm a fan of of the slow tripod-and-ND filter route. I like the process of taking my time, setting up the tripod and filters, computing the exposure, selecting my filters, etc. I enjoy the challenge of trying to balance the exposure as correctly as possible in camera.
When it comes to scenes that also have to balance dynamic range in different zones by using graduated ND filters, image stacking may or may not yield the same results. I mean, it can yield the desired results, but you'd have to take a stack of images exposed for the lightest parts of the scene, and then take a stack of image exposed for the darkest parts of the scene, and mix those two stacks in post. It absolutely can be done (and has), but it's not the route I prefer to go, when I can just use graduated ND filters to mix the dynamic range in camera.
Now, there are strong arguments for doing away with graduated ND filters, and just mixing exposures in post. This is really just a type of HDR or exposure fusion. And this can also apply to long-exposure ND filter type shots as well. But again, without physical filters, you're essentially performing 3 post-processing steps: image-stacking the high exposures; image-stacking the low exposures; and HDR/exposure-fusion blending the stacked results together.