"open source tends to be winner-take-all"
This makes no sense, and flies in the face of, well, quite a lot of business Out There. Not because it never happens, but because it is by no means a given. By that logic, why hasn't, say, redmond just opened all their source and used their market position to cement being the winner plus yanking the rug from under the whole linux fanboi cult in one fell swoop?
Because it doesn't work that way. It seems nokia were thinking something along those lines with symbian, and, er, it caused them no end of grief. First in the shoddy execution, when delivering essentially useless source to quickly disappointed masses failed to build that community, and then losing massive face by closing up again and saying "we only ment we're open for business". Syeah right. You're never going to get more of my business, that's for sure.
The thing is of course that open source is not so much a business model but rather orthogonal to that. Making money isn't the goal of open source at all, merely a nice side benefit for many. For comparison, take PostgreSQL, and observe its lack of a winner who took it all. Instead there's quite a few businesses that can deliver pgsql consultancy competing with each other.
While MySQL has its own company attached, or had before it got bought and ended up in the hands of a company not well-known for its open source love, so there's not much point in trying to consult for it. In fact, that take-over has already shown to be a weak point for continued enterprise reliance.
Or take videolan. Again, no big winner but a bunch of companies doing amazing things using its software and contributing back to the community. It's very cost effective, and the secret isn't in the source but in the delivery.
So I say that it may well be possible to try and become the redhat of hadoop, but since redhat is an outlier, just one way of doing business in a much larger palette, there's no hard & fast reason why it must work that way. Nice idea but by no means a law of nature, making the bets that much shakier.
Worse yet, big data is a big hype as of yet, a current buzzword, but it is again by no means a given that hadoop is the right approach. Or the only one, or even will remain a significant player in five or ten years. So of course, if you can cash in big or at least pay yourself a princely salary from venture capital for as long as the hype lasts, then by all means. But that again still doesn't guarantee success. That makes this piece a glimpse into silly valley thinking, but not exactly an incentive to go out and bet on hadoop yourself.