You know, I like to think that I'm a nice person, but I'm going to just have to say this; you are an idiot if you think you can compare what's essentially opcode code translation with complex interpretation.
Have you ever found a CPU that was designed to run programs slowly? x86, along with every other ISA, is designed to run general applications as fast as possible. So they designed their CISC-philosophy chip to do instructions that would, if used, make programs run faster. But when they developed the RISC philosophy, and figured out how to implement superscalar design, of course you're going to get more performance out of the RISC core - it's implementing every part of that 5-cycle instruction simultaneously in one or two cycles. Besides, with the plethora of extensions x86 has tacked on to itself, do you really think that those instructions are the most important for running your computer?
However, the web is not machine code. There is no known design for a hardware processor that will 'execute' a web page. And that's because web pages are more like a page description language than anything else. Thinking metaphorically, it's more like source code for some high-level compiled language. Every time you render a web page, you have to wait for it to compile the code. And even after that, you have more complexity with javascript, AJAX applications, server pushes, flash, java, and more. And if you think about it, you'll realize that none of the web technologies were originally designed to do even a quarter of the things we make them do today. Do you really think that we'll be able to do everything that we can do with our relatively behemoth desktop machines on a web browser?
I'm not saying it's impossible to do everything we can do with computers in a fancified web browser. But it's definitely going to take a lot of work, and for it to work, the technologies involved in the internet are going to have to evolve dramatically.