I've ranged from working 90 hours a week when we were in Y Combinator to doing around 40-50 hours a week now. I find the current amount to be much more sustainable and productive. Of course there are weeks when stuff needs to get done and 40-50 hours turns into 70-80, but I think it I am more productive when that is the exception rather than the rule.
The key point, in my opinion, is that it's a marathon, not a sprint. You might have some hard deadlines (investor meetings, etc.), and for those you might have to approach 24/7. But in general you need to be aware of your physical and mental limitations and optimize for medium-term productivity.
For me, that means trying to get at least 6 hours sleep, eating well, exercising regularly, and taking at least a half day a week off to avoid burnout/staleness. I end up working 60-70 real hours (i.e., not including screwing around on the computer) and then a bunch of additional time mulling things over.
Like most quantitative metrics, hours worked are the means, not the end. Getting the right stuff done and creating value for customers are the goals, so I try not to obsess about the wrong metrics.
Sense of ownership makes a difference. If you're Founder/CEO, working is more than an obligatory punching of the clock and is honestly not work - it's a some combo of obsession and passion. If the project is in development, every minute you're not spending on it can seem wasted. This can be unhealthy and create an overly-competitive work environment however.
Your employees should have vested interest but not to the point of exhaustion. Agree with Jamie Quint, less can be more after early stages!
I agree w/'It's a marathon, not a sprint' thinking. Working smart is the way to go, especially because you're probably being asked to take on many different tasks/projects - in and out of your comfort zone. Getting in daily time to 'zen out' and refresh during a day is needed as you finish shift 1 of the day and go onto shift 2. Defining success is an individualistic thing, so make sure you figure that out ahead of time. Just because you clock 80-100 hours doesn't mean that time was well spent or made you successful.
Having done both "normal" and "insane" hours I think the optimal is actually 40-50 hours/week but REALLY working productively during those hours (not goofing off) and spending the rest of time to rest and recharge so you can keep up 40-50 productive hours/week for 5+ years. Startups are marathons not sprints.