Are you saying there are 7+ million BTC stored on exchanges? I disagree based on two data points: Bistamp's reserves were 180K in November 2013 and MtGox's were 850K (with 650K of them missing) at the time of its death.
Even if you add BTC-E, Chinese exchanges, Coinbase and LocalBitcoins, it most likely won't exceed 2-3 million which means the remaining 80% is owned by end users directly.
So even if it's a minority at 20%, that's a pretty substantial number, and I can't imagine why it won't grow. Bitcoin's success depends on mass adoption, and I can't imagine a scenario where Jane Public would rather learn the technical skills to secure a local wallet versus trusting a company like Coinbase or Circle.
America is a free country, but not everyone in America is free (some are in prison, for instance).
Wrong.
And this is why these Bitcoin debates are so annoying and repetitive.
For some reason everyone feels entitled to their little doomsday speech without bothering to research even their most basic premises.
edit: Did you totally change your reply? From the time I started my comment to hitting submit, your comment completely changed.
But to your point, you're presuming I have no clue, that I own no Bitcoin, that I'm not a regular participant in communities like Bitcointalk and /r/bitcoin. You believe there's a world where everyone has paper wallets and Trezors and ASICs. No: the early adopters are becoming a smaller percentage of the Bitcoin world. People are buying via Coinbase and Circle, and if they're really advanced, they sweep to a hosted wallet on Blockchain. Even many miners aren't immediately sweeping their coins, instead, leaving them on the pools. Do you really think it's going the opposite direction, towards less centralization?
...
See the comment by qnr.
I don't have the numbers, but guess what, you don't either. I'm going on mostly anecdotal evidence: Bitpay's growth, Circle's growth, Coinbase's growth, etc. I may be wrong about "most", but in the face of hard numbers (which many companies like Coinbase don't provide), how could any observation of the community not lean towards a conclusion of increasing centralization?