It currently costs $40+ for a comparable fiat service (wire transfers).
[1] https://freedom-to-tinker.com/blog/randomwalker/the-low-tran...
Transferring value in the form of maintaining electronic accounts isn't comparable to wire transfers of fiat where the funds are made available in local currency. Banks don't just track numbers in accounts: providing liquidity and ensuring those numbers actually mean something is not trivial. If bitcoins were globally accepted it would be a more competitive service, but even then there would be the issues of fraud protection and compliance with money laundering legislation that bitcoin doesn't handle well.
I mean we can argue semantics, but if you want to send $100 from the US to the Philippines, bitcoin is generally a lot faster than your bank, even when you have to buy bitcoin and sell it in the Philippines because of international transactions taking much longer than local ones.
As for capacity, bitcoin's 'Chief Scientist' just wrote a blog post on scalability: http://gavintech.blogspot.de/2015/01/twenty-megabytes-testin...
To make a long story short, thousands of transactions per second (Visa averages about 2k) is certainly possible. I'd be more worried about bitcoin adoption (from an interest/tech/growth point of view) as a limit than technical scalability issues as a limit, especially with things like sidechains. (Look at Blockstream's Sidechains for example as a solution, just got a $21m investment.)
It's generally held that it can scale, whether it can grow and become popular and well used remains to be seen (I think it's likely but it's far from certain).
The block size limit is in the process of being changed, but there isn't a rush since blocks are not being filled. If they do get filled, transaction costs will go up and tiny (0.01 USD) micro-transactions won't be viable until the block limit is increased. So the only thing that might happen is transactions of a few cents are temporarily not viable (the transaction cost is larger than the amount sent).
