If you go about looking at energy density without evaluating the feedstock materials and embodied energy, your maths won't achieve the desired result.
I was doing the math on the back of a napkin, but there's quite a few papers on the topic.
www.vaclavsmil.com/wp-content/uploads/15.WINDTURBINE.pdf
I might be off by 25M-50M tons, but I think the overall assertion is valid.
My belief about how to fix the problems with energy grid: http://engineersf.com/things-humanity-could-do-right-now-to-...
Any sort of libertarian or free-market approach to this problem will likely fail. This is one of those mission critically deal breakers where human behavior can't be modulated in a desired direction.
"the aggregate installed wind power of about 2.5 terawatts would require roughly 450 million metric tons [of steel]"
That's terawatts (nameplate power capacity) NOT terawatt-hours! A very important distinction since the steel is a onetime cost for the life of the turbine and can potentially be recycled at EOL.
Yes, you need currently available usually fossil energy to make renewable sources, including PV and nuclear. What is the alternative? Yes, embodied energy / EROEI is important, but it has to be considered fairly across all technologies. Including battery storage and nuclear plants, and the replacement of petrol vehicles with EVs.
The turbines tend to have composite blades and steel towers. Actual studies put the amount of energy returned by wind turbines at ~18 times the energy invested: http://www.theoildrum.com/node/1863
And if not wind turbines, then what is your preferred low-carbon energy source?