I think rather than worrying about hours, entrepreneurs and other people who want to be highly effective should worry about how to improve their skills and focus when they are at work. There's a way in which a 100-hour work week can become a scapegoat... if you fail to produce at least no one can say you weren't committed. However, if you are truly only interested in results, you need to step back and ask yourself objectively are you making the best choices or are you just trying to cover your ass? I realize that outside perceptions actually do matter a tremendous amount to business success, but the bottom line is that the best person to build something is not simply the person who is willing to work the longest hours.
Most of the blowout successes I'm aware of have had a long period of protection from competition while the founders became world-class experts in whatever it was they were doing. That could mean a period in academia, open source, or a long period of learning at some other company in the industry.
There are a few cases I know of where the founders were not experts, but in those cases it was because they stumbled onto a new market, so nobody else knew more than they did. And even then they were well prepared to exploit it, through related work.
Also: 40 hours a week is a LOT. If you don't fritter it away with meetings, office hours, coordinating with vendors, and so on.
For example, if you look at Lars Bak's resume, he started working on the Beta programming language in 1988. (For HN readers that aren't familiar with it, Beta was an exceptionally innovative language, and I'd highly recommend learning a bit about it.) Then he joined the Self team in 1991, where he did the first dynamic-optimizing JIT compiler. Then he worked on StrongTalk at Animorphic/Longview, and then was the engineering manager for HotSpot at Sun. He's spent 20 years working on basically nothing but optimizing VMs for dynamic languages - it's no wonder that he could kick out a dynamic JIT for JavaScript in a year.
Paul Buchheit once had a thread on FriendFeed about how long it takes to build a really cool project. It took him about 3 years for GMail, and Google was kicking around in Larry & Sergey's head for 3 years before they incorporated. That seems fairly typical for most Google products.
So while most successful people will say "Work reasonable hours and get enough sleep," I wonder how many of them are able to say that because they've already put in the hours necessary to attain mastery in a subject. Perhaps the solution is to optimize for rate of learning+retention: work as long as necessary to fully understand the subject, but quit when you feel you can't concentrate and remember things. Sometimes you can work for 15 hours and learn nothing, while sometimes you'll put in 3 mind-bending hours a day and learn tons.