Walter Rudin's Real and Complex Analysis has long been one of my favorites. Like Serre, Rudin seems to strike a nice balance for detail, and his proofs are always slick and fun to read; I became heavily interested in analysis after reading that one.
Silverman and Tate's The Arithmetic of Elliptic Curves got me interested in that area for some time, too. The exposition is fun to read, with both motivation and rigorous proofs.
I wish I had seen Volume 1 of Spivak's A Comprehensive Introduction to Differential Geometry a bit sooner; I think I would have had a much easier time with things like tensor fields, Lie derivatives, differential forms, etc. But for some reason in the next volume everything becomes extremely messy and non-invariant.
Not sure if this qualifies as a math book, but Sipser's Introduction to the Theory of Computation has an elegant, well-motivated but rigorous style; I wish I had seen this way sooner, since the prerequisites are minimal, and it is fun to read. The exercises are also fun.
