The bandwidth debate rages on. In an April research note, Think Equity analyst Eric Kainer argued that AT&T will likely scale back its current fiber-to-the-node deployment over the next two to three years in view of the realization that it simply doesn't provide enough bandwidth to the home to effectively compete against cable and satellite video providers.

Here's Kainer's math: Assume that technological advancement can compress high-definition television signals from their current size of about 20 Mb/s down to 10 Mb/s. The typical American home will soon have three HDTV sets, Kainer said. And in many cases, each set will have a dual-tuner digital video recorder (DVR), doubling the bandwidth it consumes. Three HDTVs with dual-tuner DVRs comes to 60 Mb/s — and that's just the video service, not voice or data. In that picture, AT&T deploying 25 Mb/s to the home, Kainer said, “is like going to war with pitchforks when the enemy has laser-guided robots.”

Of course, the speed of the VDSL technology AT&T is deploying varies, depending on distance. And at close distances, it can likely offer speeds approaching 100 Mb/s, which will probably make it fine for most apartment buildings, Kainer said. But for the bulk of its footprint, he said, “[AT&T] needs to plan for 100 Mb/s or more to the home today, with a path to Gigabit Ethernet to every home in the next five to ten years at the most.”

The question of how much bandwidth to deliver to residential homes has been raging for years. And because overhauling access networks is costly and time-consuming, carriers are forced to imagine not just what consumers' bandwidth needs are today but what they will be a decade or half-decade hence.

Popular applications can catch fire in much less time, which is why participants in the bandwidth debate (particularly equipment vendors acting as arms dealers) often hold out the possibility of as-yet uninvented bandwidth-hogging applications. To prove how unknowable future bandwidth needs might be, sometimes they even cite Microsoft Founder Bill Gates' infamous prediction that personal computers would never need more than 640K of memory. But he actually never said it.

In last year's second half, when equipment vendors began to aim their marketing machines at the demand for Gigabit passive optical networking (GPON) equipment, Tellabs Chief Executive Krish Prabhu tried to temper the hype by assuring investors that GPON would not be deployed significantly until 2007 because it wasn't necessary until carriers perceived a demand for 100 Mb/s to the home rather than the 30 Mb/s delivered by existing PON gear.

To unleash the first wave of attack (or defense) against cable triple-play offerings, carriers in general don't need more than about 25 Mb/s to the home, said Derek Kuhn, senior director of strategic solutions for Alcatel. “That's the kind of number we've thrown about. It seems to be a number everybody has embraced.”

To relieve the pain points of cable and satellite providers, much work is being done to compress the amount of bandwidth consumed by HDTV, often pegged as the gating factor in residential bandwidth debates. (Notably, telcos aren't feeling this pain but will benefit from moves to remedy it.) Though today's HDTV content (based on MPEG2) typically takes up nearly 20 Mb/s, encryption experts are moving increasingly toward the International Telecommunications Union's H.264 standard, which promises the ability to send quality HDTV at no more than 10 Mb/s. According to Kuhn, vendors already are doing better than that.

“We've seen stuff well below 10 Mb/s now that looks spectacular,” Kuhn said. “Seven or eight [megabits per second] remains a very conservative number. We'll see some vendors do significantly better than that in the coming months.”

If that's true — if HDTV could be offered at, say, 5 Mb/s — the three-TV home that Kainer warned would need 60 Mb/s just for video can now get the same thing with only 30 Mb/s.

So if Kuhn is right, what will consumers do with the 100 Mb/s they will get from the GPON gear that Alcatel, for example, introduced a year ago?

“I have no idea,” Kuhn said. “There's no service that requires that at this stage. Having said that, somebody's going to figure out how to use it.”

One key open question in the debate, of course, is how many HDTV streams consumers will want simultaneously. At the end of last year, nearly 13% of U.S. households had an HDTV, according to Parks Associates. By the end of 2010, it will be up to 65%. However, the number of households with multiple HDTVs is much smaller. Although nearly 17% of U.S. homes have one HDTV, less than 2% have two HDTVs and less than 3% have three or more sets, Parks said.

It's also important to note that in order to deliver services such as video, carriers don't just need fat pipes but also quality transmission: low latency and jitter, for example. And in fact, the more quality carriers have in their networks, the less important the size of those pipes becomes. Carriers with plans to offer video services are often deploying twice the capacity they need to prevent bandwidth constraints from getting in the way of a clear picture. But over time, they may add network management capabilities that allow more efficient use of that bandwidth (for example, software that assigns a hierarchy to different traffic types, protecting video but not, say, peer-to-peer traffic when networks are congested). This would make those existing pipes seem even fatter because carriers would no longer be throwing excess capacity at their quality concerns.

And while the debate over how much bandwidth is enough generally focuses exclusively on downstream bandwidth today, in the future, carriers may have to re-evaluate how much upstream bandwidth is necessary. Although today's video games usually don't require more than 256 kb/s upstream bandwidth, some of the newest games showcased at the E3 Expo video game show earlier this year reportedly require upstream speeds of 1 Mb/s and 1.2 Mb/s.

If coming applications include more interactive video (such as HD gaming, file trading or videoconferencing), those requirements could get much fatter. And if incumbent carriers want to make good on their expressed desire to turn bandwidth-rich applications into revenue streams by charging content providers for network quality, those carriers will want to provide all the capacity those new application providers can eat.

“Incumbents today are more concerned about their core services, making sure they have a high-bandwidth pipe but not concerned about what's going over that pipe,” said Deepa Iyer, Parks Associates analyst. “But that's going to change.

BANDWIDTH REQUIREMENTS FOR BROADBAND SERVICES