The Economist has an interesting discussion of several academic papers addressing aspects of the question.
Shane Greenstein of Northwestern University and Ryan McDevitt of the University of Rochester calculated the consumer surplus generated by the spread of broadband access (which ought to include the surplus generated by internet services, since that is why consumers pay for broadband). They did so by constructing a demand curve. Say that in 1999 a person pays $20 a month for internet access. By 2006 the spread of broadband has lowered the real price to $17. That subscriber now enjoys consumer surplus of $3 per year, even as the lower price lures more subscribers. The authors reckon that by 2006 broadband was generating $39 billion in revenue and $5 billion-$7 billion in consumer surplus a year. Based on its share of online viewing, Mr Greenstein thinks Wikipedia accounted for up to $50m of that surplus.
Such numbers probably understate things. The authors’ calculations assume internet access meant the same thing in 2006 as it did in 1999. But the advent of new services such as Google and Facebook meant internet access in 2006 was worth much more than in 1999. So the surplus would have been bigger, too.