AT&T's Group President John Stankey spoke at Merrill Lynch's Communications Forum today in New York, in the process taking a shot at cable broadband networks (webcast here). Stankey says that AT&T went into one of their cities, purchased cable broadband service for some 150 users (from the speed tiers cited, they appear to be Comcast connections), and then tested network performance.

The company poked and prodded the network for a period of several months. AT&T claims they found that peak downstream speeds were between 3-4Mbps, while average downstream speeds for the users ranged between 300kbps and 400kbps, significantly less than the advertised rate of six to eight megabits per second. Some Stankey comments we transcribed from the webcast (hat tip to IP Democracy):AT&T certainly has a point that your connection speed depends on the entire path of the network, as well as local congestion and network management. That said, AT&T's fastest 10Mbps U-Verse tier is going to struggle to compete with Comcast DOCSIS 3.0 deployment when it comes to promised end throughput.

The FTTN architecture also pales in comparison to FTTH projects like FiOS, so obviously they're going to downplay advertised connection speed. AT&T made the decision to try and placate investors by spending less on initial deployments, though it remains very likely they'll ultimately embrace fiber to the home anyway.

Comcast's network management PR disaster and congestion could give AT&T a temporary edge, and AT&T is supposed to announce faster speeds (for users closer to the CO and via pair bonding) sometime this year. So cable users, given you've tested and tweaked your connections extensively, do you agree with AT&T's findings? Most of you seem to be getting what you pay for.