I wouldn't bother with a bitcoin startup, but for a real ecash startup, US for technology development, and then Hong Kong or maybe Singapore for server ops, deployment, and financial backing of currency issues.
The key IMO to all of this is separating out software development from server operations from financially backing a currency from applications of the system from auditors from users. Banks suck at software development; developers suck at finance. By separating the concerns, you can do regulatory arbitrage as well as focus on your strengths. (this isn't my idea; I think Ian Grigg of Systemics came up with it in the early 1990s, if not earlier). Fully separate arms-length entities, different people, not the e-gold "offshore, really" approach. You could create a fully anonymous bearer blinded token infrastructure as software, and then have servers for it, where some of the currencies were entirely non-financial and thus non-regulated (distributed locking mechanism for distributed software systems, resource allocation, anti-spam credibility systems, game currencies), and others which were financial currencies issued by real, established financial institutions with KYC (a dollar currency, shares, futures, ...). Same software, so the developers have no legal risks themselves.
Luxembourg also works, if you have $3-5mm or so to meet the requirements to issue (at least, that's what it would have cost in the late 1990s when I last looked; you need some connections, but they come with having money and hiring people from the industry). Switzerland and possibly Liechtenstein may work. I'd skip dodgy ee or south pacific places (e.g. Vanuatu, Narau) since correspondent banking relationships would be trivial to sever.
If you were totally focused on gambling ("gaming"), some crown colonies (IoM, Gibraltar) might work, or Panama.
You could do it in the USA if you met KYC and also treated it as a commodities trading system, regulated under CFTC. You can become a CFTC exchange (like Chicago) for $5-10mm. I'd probably actually do the operations for that in Chicago just to be totally known to all the local regulators, although NYC or SF would also be possible.
The thing I absolutely wouldn't do is try to run a payment system which is at all anonymous in the US, and if I did a payment system in the US, it would meet all KYC and MTB requirements. That puts your capitalization required at a few million dollars more than otherwise, but it's not a big deal.
KYC among a group of commercial customers, and then getting all the advantages of easy settlement, instant settlement, etc. You can layer identity and audit on top of the actual settlement system, since that part doesn't need to be in real time.
The idea is to use the same infrastructure for that, for non-financial mass-market anonymous stuff, and then if anyone wants to run it, for anonymous financial stuff.
End user retail purchasing on the Internet (which is what PayPal largely does now), and user to user small value transfer ($1-3k) which is also paypal, is kind of uninteresting in general. Credit cards are fine for purchasing, and user to user payments are a pretty niche thing.
The key IMO to all of this is separating out software development from server operations from financially backing a currency from applications of the system from auditors from users. Banks suck at software development; developers suck at finance. By separating the concerns, you can do regulatory arbitrage as well as focus on your strengths. (this isn't my idea; I think Ian Grigg of Systemics came up with it in the early 1990s, if not earlier). Fully separate arms-length entities, different people, not the e-gold "offshore, really" approach. You could create a fully anonymous bearer blinded token infrastructure as software, and then have servers for it, where some of the currencies were entirely non-financial and thus non-regulated (distributed locking mechanism for distributed software systems, resource allocation, anti-spam credibility systems, game currencies), and others which were financial currencies issued by real, established financial institutions with KYC (a dollar currency, shares, futures, ...). Same software, so the developers have no legal risks themselves.
Luxembourg also works, if you have $3-5mm or so to meet the requirements to issue (at least, that's what it would have cost in the late 1990s when I last looked; you need some connections, but they come with having money and hiring people from the industry). Switzerland and possibly Liechtenstein may work. I'd skip dodgy ee or south pacific places (e.g. Vanuatu, Narau) since correspondent banking relationships would be trivial to sever.
If you were totally focused on gambling ("gaming"), some crown colonies (IoM, Gibraltar) might work, or Panama.
You could do it in the USA if you met KYC and also treated it as a commodities trading system, regulated under CFTC. You can become a CFTC exchange (like Chicago) for $5-10mm. I'd probably actually do the operations for that in Chicago just to be totally known to all the local regulators, although NYC or SF would also be possible.
The thing I absolutely wouldn't do is try to run a payment system which is at all anonymous in the US, and if I did a payment system in the US, it would meet all KYC and MTB requirements. That puts your capitalization required at a few million dollars more than otherwise, but it's not a big deal.