Alpha channels have to be explicitely specified to be used. To apply one, a lot of extra work is required and if you're not going to use the alpha channel, you may as well leave it switched off or call functions that don't use it.
The value is still 32 bits instead of 24 bits in memory for one simple reason: your computer has instructions to work with 8, 16, and 32 bit values. To work with a 24 bit value (ie, just RGB without the alpha channel), it would have to do at least two operations (grab 8 bits from memory, then grab the next 16 bits) per pixel. By keeping the (useless) alpha channel, the image wastes a little bit of space but gets processed a lot faster (just one operation: grab 32 bits from memory. It's comparable to completing a shipment in one single trip with a big truck instead of in two trips with a smaller one.)
Note that on disk, you can store a pixel under any format. Disk access can be slow since it usually has to occure just once and then the image stays in memory.