> I'm putting the number closer to 40 (the "NFL cities") right now, and
> 150 by the end of the decade, and ultimately any "metro" with population
> greater than 50K in a 100 sq Km area will need a neutral exchange point
> (even if it's 1500 sqft in the bottom of a bank building.)

What application will require this dense peering?

today, we need this dense peering to keep local traffic local just with
apps that folks already run. kazaa and gaming are examples where isochrony
or high volume bidirectional peer-to-peer traffic are already present, but
the fact is that "hub & spoke" is a better topology for a metro than for
a region, even where http/smtp/ftp are still the primary applications.

going forward, movies on demand and other things that we currently use
satellites or cable TV systems for. voip. internet-delivered radio, using
things like 802.11 and bluetooth as the "last mile." i want a Dick Tracy
wristwatch and i know that thousands of other people will want it too and
i can do the arithmetic to see that there will be more than one (probably
more than several) providers per metro, even in small metros, and that if
their closest exchange point is in some other metro, it can't take off.

someone mentioned SIX. but a peering switch does not an exchange point make.
without a PNI upgrade path, which means a certain amount of hard colo, the
ceiling is too low. (that's one reason why ATM-based metro exchanges are not
growing very fast, and why nobody is building new ones any more.)

Its possible/likely that what Paul is saying may happen, but it requires
a lot of locality-specific high-bandwidth applications (none exist now or
in demand now) and technologies that make it possible (cost-effective) to
manage such complex peering network for a very large network

Maybe 10 years from now some of this maybe true but I see a more likely
scenario that large ISPs will be also application/broadcast providers
(i.e. like AOL, MSN) or otherwise affiliate with one of large broadcast
networks. All these large providers more then likely will see a need to
put local application serving & broadcasting stations in each city (like
mirrors or broadcast<->unicast trancievers, etc) to make it worth while
for their subscribers in that city to make most advantage of those
services, however same companies may NOT want the subscribers to access
another network's system (direct competitors) and this unfortunetly will
result in them not wanting to provide 0-latency interconnect to another
network, so more then likely there would not be a peering exchange in
every city.

Of course, I'm just speculating on something that is far, far ahead ...