How many megapixels would it take to equal a 35mm film maximum quality?

The short answer is that a modern 24-megapixel digital SLR offers around the same level of resolution as a good film scanned in a modern minilab. Ken Rockwell has tested the Nikon D3X to this effect; see related links below.

Now for the long answer.

In Theory

It's quite easy to work out the maximum theoretical resolution of film; manufacturers document this for their films. It's measured in lines per millimeter (L/mm); each one of these is roughly equivalent to two pixels in one direction. So all we need to do is multiply the width of the film (in millimeters) by the L/mm figure, multiply that by two. Do the same for the height, and there you'll have the maximum resolution at which a film, shot in perfect conditions, can be scanned, without interpolation (made-up pixels -- something we'll come back to later).

((lpm * 36 * 2) * (lpm * 24 * 2)) / 1000000

If we're going to make a comparison with a high-end digital camera, it's not fair to take a cheap consumer negative film for the comparison. For our comparisons, we'll use the professional Velvia film, at 160 lines per millimeter, we end up with a figure of about 88 megapixels. (If you wondered why we multiply by two, this is because one line = two pixels.)

Yet, this figure does not show you how many megapixels a digital camera would need to achieve the same resolution. Most of a digital camera's "pixels" simply do not exist; they do not capture both red, green, and blue on each pixel, only one colour each. They then use Bayer interpolation to make up the rest. Ken Rockwell claims this "lie factor" as being about two. You may prefer different figures, but we'll use this. So, the final "megapixel" count becomes about 176. You could even up this somewhat if you take into account the fact that digital cameras love nuking any fine details with noise reduction, but we'll leave that aside for now.

(Ken Rockwell ends up with a figure of 175 for the same film. Depends on how you round it.)

This comparison is unfair to digital, though. This assumes that you're shooting test charts specifically designed to extract the maximum resolution from a piece of film; moreover, it assumes that very fine amounts of barely-resolved detail are just as significant as coarser details resolved near-perfectly. The real world doesn't work like this. Some people look at the MTF curve of the film, and see at what point the curve drops below 50%. This would leave you resolving 50 lines per millimeter, or about 8 megapixels (16 if Ken Rockwell's "lie factor" is accurate). You may choose another, just as arbitrary, percentage, look at your film's MTF curve, and do the math yourself.

(An explanation of MTF curves is beyond the scope of a quick answer, and not particularly interesting anyway. Sorry.)

In Practice

This theoretical resolution does not translate into the real world very well.

For one, you're assuming that your lens is able to resolve this much detail. It might not be, in which case it's wasted. Moreover, if you're going to make a fair comparison of resolution, you will do so side-by-side on a computer screen. You have to get the picture off the film onto a computer. Yes, you could compare prints, but then you'll be comparing your optical printing system versus your computer printer, which isn't fair. That or comparing how well your printer prints certain things.

It isn't fair to shove a cheap print (itself made digitally!) into a cheap flatbed scanner, compare it to a digital file, and conclude thereafter that film sucks and digital rules. Yet, neither is it fair to assume that any photographer will be ready and willing to spend a small fortune on an 12,000 dpi drum scanner and then spend 30+ minutes on scanning one slide (or failing that, spending a hideous amount of money to have someone do it for them). More likely that they'll own a CCD film scanner, or drop it into a minilab to have it scanned for them.

In short, the moment you leave the laboratory and take the comparison into the real world of photography, too many things get in the way for a simple resolution comparison to be made. Here, then, are various "real world" figures that some people here have offered.

• 2-3 megapixels: Visual equivalent in a 4x6 print to 80-85% of the population, on a camera with a decent lens.
• 3 megapixels: Some people use the comparison with an A4 print. A 3 megapixel camera will do this happily.
• 6-8 megapixels: For cheap consumer negative film. People who own the 6-megapixel Nikon D70, or the Canon D60, say that a 20x30 poster print from one of these is roughly equivalent to that from 35mm film. Make of that what you will.
  

• 12 megapixels: Another arbitrary figure sometimes offered.
  

• 24 megapixels: Ken Rockwell's comparison of the 24-megapixel Nikon D3X showed that it resolved details about as well as Velvia 50 film scanned commercially. This is probably the answer that fits the real world of photography better than any others.
  

And finally...

There are varying advantages and disadvantages of film that cannot be captured in a megapixel figure, so barring some massive change in digital camera sensor technology, a fair comparison will never be possible.

• The size to which you want to print an image matters. Digital breaks down in a nasty and unusable fashion when it reaches its resolution limit. Film gradually gets crummier till you reach a blurry mess somewhere around 40X enlargement.
• Film has much greater dynamic range than digital. Another thing that doesn't fit into megapixel figures, but is very important.
• Digital has noise issues. Film has grain issues.
  
• There are good films and bad films, and good digital cameras and bad ones. As stated earlier, any comparison you make has to compare like and like. Don't compare a point-and-shoot digital camera (regardless of megapixels) with an expensive professional slide film. Don't compare a cheap negative film from Wal-Mart with an $8000 digital SLR.
• Megapixels are irrelevant. Nearly any digital camera can be used to take amazing shots. Nearly any film camera can do the same. This has nothing to do with megapixels. If you're not doing enormous enlargements, or only viewing your pictures on a computer screen, then don't worry about it.