Most jobs, even technical ones, truly do not require a Ph. D. This is just the market sorting things out. Why have zero (or negative) income to get post baccalaureate education if you can go work somewhere (for money) and learn on the job?
Well, ok, so here's the thing: If you apply to grad school in one of the physical or life sciences and they don't offer you a TAship for your first year[1], then they are politely telling you to f*ck off. It's "generally understood" (which means that nobody knows this unless someone tells them) that any serious offer comes with a TAship that will just barely pay your rent in a crappy one-bedroom apartment with a little left over for food.
[1] And make wild promises about how everyone has a full RAship with ponies by his second year
1) A recent PhD with $40,000 to his name
2) A Google engineer who graduated from undergrad at the same time and has $400,000 to her name
How many MTV or NYC Google engineers have $400k to their name after 5 years, as opposed to $4,000 and a bunch of money torched on rent? Google pays well, but not as well as you seem to think.
I'd say that (2) wins, if only because someone who sticks with Google for 5 years is probably one who made Real Googler and now has a fair amount of freedom. Google is pretty nice once you're above the Real Googler Line.
But that's aside from the actual point. It's not even about what the Googler does after they've done 5 years. That kind of money is enough to do whatever the hell you want to kind of money. You want to read papers all day? You can do it full time and pay yourself a graduate student's stipend for life, without having to worry about pleasing the dictates of your advisor, academic fads, TAing, and internal politics. Even the lucky students who become professors still have to apply for funding and teach students, on top of producing popular research to get tenure: with route 2), you never have to apply for funding, and you can choose how much time you want to devote to teaching and what kind of teaching you want to do.
It's fair to quibble with the actual numbers, but the point is that money buys you a hell of a lot. The fact that someone might not save, or pathologically continue slaving away at the corporate behemoth they hate once they've saved enough, doesn't change that possibility.
Now admittedly, $20k/year is a hell of a lot to be able to save for normal Americans! But at the end of five years it leaves you with $100k of savings, not $250k. In the Boston area where I was, you couldn't even put a down-payment on a house with that little money. It was basically just lots and lots of beer money.
On the other hand, if you know a magical company in a magical land, perhaps adorned with pastel ponies, at which I can make enough money with a low enough cost of living to retire after five or ten years to become a self-funded gentleman scientist, I'll certainly take that over graduate school. I just don't think it actually exists, as you can probably tell by the sarcasm.
1) Most engineers do not work at Google, and 2) I have plenty of friends who do/did, and, guess what?: They still write code all day! Most of which does not do neat things like drive cars, but rather: increases AdWord click-through rates! Or: manages address book contacts. Fun!
I don't care how much free pizza and foosball you offer me, nothing can match the freedom of pursuing my own interests.
Seriously. IMO someone reading through, say, the TAOCP, Rudin's Real and Complex Analysis, Horowitz & Hill's TAOE, or any selection of good, difficult books like these, will have acquired undoubtedly more intellectual freedom afterwards than any PhD graduate. For starters, it's self-directed study, which is orders of magnitude harder, and more important to self-growth, than study directed by an advisor, or any bit of structure. It's also faster. It doesn't sink you in any significant debt, which is a plus on the order of physical freedom. It's much more eclectic, which is incredibly important, both in terms of the monopoly power you'll bear, and the scientific value your unique analogies between concepts will create. Hopefully.
Sure, there's no paper afterwards to show. My take is that, credentials or not, you'll still have to show your competence to a potential employer or investor, your exposition tailored to the particular needs you suspect they have.
So be careful when you say "[reading these books will give you] more intellectual freedom afterwards than any PhD graduate." That's simply not true. And I'm not even touching on the benefits of having a good mentor...
You do have a point, but the way I see true "intellectual freedom" (lolcraft may have meant somth similar) is that let's say you spend your late 20s-early 30s slaving away at a post-doc only to have a chance to access a "petaflop supercomputer" and then you suddenly decide that you actually are really interested in classical Persian poetry and you'd like to devote some time to learning Persian, only to realize you do not have time for that because hey, somebody needs to write those papers for the grant money to come in and your chances for tenure to remain intact.
Replace "Persian poetry" with the study of Mathematical logic, the reading of the pre-Socratics or trying to make sense of the early-medieval migration patterns, as things stand right now both the people following PhDs and those too immersed in industry are not "intellectual free" because they have no time for these sorts of intellectual pursuits. We do need to find that middle-ground between extreme-science and industry again, we do need universal people like Leonardo and Democritus back.