Some great advice in these answers. Here are some specific questions to ask:
1. Will you be selling to consumers or businesses? This will impact your choice of stack
2. What level of sophistication do you need in web functionality? If you simply need to display information, such as in a web site, then you will make a very different choice than if you are building an application with interactivity. If, in addition, you'll need to store data, then you'll need to make choices about the data back-end. If, in addition, you'll need to do a lot of processing on the server, then you're going to make a different choice than if your performance needs mean you can stick to client-side processing.
3. Have you validated the concept? If not, then consider starting with "pretotyping": building a landing page and an adwords campaign to test interest in your offering, and then doing A/B testing to pivot until you've found a point of resonance. This is particularly helpful for consumer plays. You can build this in less than a day.
5. If you're B2B then consider a B2B interview campaign. Even a short study, with a dozen or so participants, can go a very long way. Ask people who might ultimately be users to give you advice. There are some tricks to getting decision makers on the phone, too, in such a way that everybody wins.
6. What ultimate level of performance will your application need? Cloud- and web-hosted apps will eventually hit a point of diminishing returns. But you might not hit that for a very long time.
7. Will there be particular security or privacy requirements? This can have an impact on your cloud choices, and also on the software you'll need to build.
Finally, I'd like to reiterate a point made above: get it out fast, get some feedback in whatever way makes sense to you. Technology is hard, but finding a fit of technology to a point of pain that really solves an important problem, and that isn't solved elsewhere, is hard to the power of hard. It's easy to get distracted by technology and technologists. Market/technology fit should always be your primary focus.