Because the latter barely ever happens in 'twice the time' right from the start. Even assuming the flat 2x, that's a lot for a project that takes 6 months the 'dumb way' for any company nowadays. The pay-off for 'smart way', is unlikely to be there before 2 years from project start. You're lucky to have a company look ahead more than 2 quarters, let alone 2 years.
That's not to mention all the other problems as commenters on a different article about speed pointed out yesterday. Dumb way at least gives you something. 'Smart' way goes in and makes a hundred assumptions only to realize they forgot something a prototype would've pointed out to them. Not the dev's fault management doesn't listen to "this is a throwaway prototype".