This is comparing apples to oranges. The CD is a medium for storing bits; MP3 is a way of representing sound as bits.
The way of representing sound as bits that most CD players use — stereo 16-bit 44 100-sample-per-second PCM — which I’ll call CD-DA for convenience here, even though that can also refer to the CDs that have that format of audio on them — requires many more bits to encode one second of sound. Specifically, it needs 2 * 16 * 44100 = 1 411 200 bits per second.
MP3 allows you to pick your bit rate; I’ve heard of encoding MP3 as low as 96 000 bits per second, but more typical is 128 000 bits per second (with 256 000 bits per second for extra quality). There are even MP3 encoders that do variable-bit-rate coding so that parts of the sound that have more detail encode into more bits.
MP3 lets you record more sound in less bits. That’s all it does.
You can record MP3s on a CD. A CD holds about 5 734 400 000 bits; if those bits are a CD-DA recording, they represent about 4060 seconds, or about 68 minutes. If, on the other hand, they are 128kbps MP3 recordings, they represent about 44800 seconds, or about 12 hours.
There are MP3 players that can read CDs with MP3s on them.
But the ability to record more sound in fewer bits means that it becomes reasonable to try to record sound in solid-state memory like flash RAM. Flash is much, much more expensive per bit than CDs are. At http://www.pricewatch.com/, I see sixteen-megabyte flash cards at $30 to $100; that’s 134 217 728 bits of solid-state, no-moving-part memory. If you stick 128kbps MP3s in that, that will hold 1048 seconds of sound, but if you stick CD-DA in it, it will hold a minute and a half.
So the lower bit rate enables the use of higher cost-per-bit media like flash RAM to hold reasonable amounts of music. But it also enables the use of low-cost-per-bit media like CDs to hold unbelievable amounts of music — or just sound in general.
Think of things like IBM’s microdrive: http://slashdot.org/articles/99/02/02/102255.shtml and http://www.wired.com/news/technology/0,1282,18938,00.html. 340 megabytes — half the size of a CD — and one inch square. And dual-photon optical storage.
The bottom line is this: sooner or later, most recorded audio will be compressed with something like MP3 or Ogg’s Vorbis (see http://www.xiph.org/ogg/vorbis/index.html — it has the advantage of being free, unlike MP3.) Right now, decompressing these formats takes some fairly pricey electronics (a hundred dollars or so), but that is changing rapidly. But the jury is still out on the storage medium.
There’s the possibility that the music industry might be able to clamp down on our rights to free speech to the point where we have no choice. But, barring that, lossy compression formats like Vorbis (and MP3) are going to win over CD-DA.
This offers the possibility, actually, of recording sounds more accurately, and we’ll finally be able to donate all our LPs to museums.
Why? Well, we can record at higher sample rates (88,200 samples per second, for instance), but if we have to distribute our music in CD-DA, there’s no point. But if we have other formats — capable of handling the better-sampled music and even reducing it to a more manageable size — we have an incentive to do better recording.