You could mine, even if it was unprofitable. Spending 10 million in mining costs to produce $5million in bitcoin is still a win for a lot of people trying to move money out of China.

Does that work - I think the mining reward is 25 BTC, about 75,000 USD. So for 7.5million thats being the winning miner 100 times.

I honestly don't know what setup it takes to be reasonably certain to be the winning miner these days - does anyone have stats on how many miners and how often one might be expected to "win"?

The stats don't really matter, because the market is designed to equilibrate the price: it costs 7.5 million to mine 7.5 million's worth (in expectation, net present value and all that), and always will. If you're worried about reducing your variance you can join a pool rather than mining directly.

But over what time period?

At a guess we'd be talking 6 months - 2 years - my vague memory/impression is that's roughly the timescale on which mining hardware goes from new to worthless. If you're more vertically integrated in terms of buying datacenters/chip fabs/what-have-you then I'd imagine longer.

There was an article posted a few weeks ago about a large Chinese mine that was doing $200K/day.

Sure, and that's a lot, I'm just wondering if there's anything approaching a investment->return timescale.

Not that it really matters! It's not like I have a few mil hanging around looking for something to do...

You can mine in a pool that distributes the pool's winnings proportionally to hashpower to its members. You don't need to actually mine any blocks on your own.

The largest mining pools were based on China.