Transmeta's original plan was to execute the x86 architecture in emulation faster than actual silicon.

To this end they created a very good emulation scheme and eliminated bottlenecks. For instance they eliminated several bottlenecks that result in excess heat generation.

The idea was sound, but it took longer to implement than they planned on. So they wound up with a slower chip. But it didn't produce much heat, so it was lower power. They then decided to market a low power x86 chip, and hoped (because of the simplicity of their design) to execute cycles of Moore faster than Intel, eventually getting back to their original goal. (Business note: a low power x86 chip is essentially only marketable to the mobile Windows/DOS market. Mobile so that you have tight power requirements, and Windows because other architectures let you switch to better low-power chips like ARM.)

However as Thane points out, Intel cut off their air supply in the low power market, and at least in part due to resource problems, Transmeta wound up executing Moore more slowly than Intel did.

Transmeta knew from the beginning that they were either going to completely dominate x86 or they would sink without a trace. They're going to do the latter.

But it was an interesting idea.

Cheers,
Ben