In a white paper published this summer, Cisco Systems argued that, despite attention given to YouTube and its ilk, carriers need not lie awake at night dreading an onslaught of Internet video traffic. Though online video has grown to represent 18% of North American Internet traffic (up from 7% just two years ago), it is still just 11% of global consumer Internet traffic, Cisco said. And even as it quadruples in volume over the next four years, YouTube-type video will be eclipsed by Internet video that is viewed on TV sets (à la companies such as Veoh). And even that will be surpassed 10 years or so from now by what Cisco said is the true pipe-choker: real-time video communications and non-Internet on-demand IP video.

Those services, more than any other, will drive a spike through carrier networks such that, by 2011, Cisco said, consumer IP traffic will exceed 18 exabytes — that's 18 billion Gbytes — per month. Internet video, delivered to TV screens, will take up about 1 exabyte, while more than 10 exabytes per month will be non-Internet IP video: cable and IPTV video-on-demand. (To put that startling claim in context, researchers at the University of California at Berkeley estimated in 2000 that the world produced less than 2 exabytes of unique information per year.)

Odlyzko won't rule out such a scenario, but he's skeptical of some specifics of Cisco's projections. Though Cisco is now reporting 50% Internet traffic growth over the past few years, Odlyzko pointed out that the same company claimed 100% growth in 2005 and this summer predicted acceleration toward 300% or 500%.

According to Cisco, the deluge of non-Internet IP video will cause metro traffic to swell beyond even core traffic in the next decade. To manage the coming crush, the router giant recommends a mix of content delivery systems, capacity upgrades and compression technologies. But beyond the network engineering questions, carriers will have to solve the economic conundrum posed by video, which today fetches about $0.0001 per megabyte, Cisco said — an unacceptable return for such burdensome cargo.

Cisco imagines two basic scenarios for approaching consumer video transport economics. Roughly speaking, one way is to convince consumers to pay per megabit for some service guarantees. Another way is for carriers to find new revenue sources for consumer services — revenue from outside the consumers themselves, such as newer and more sophisticated forms of advertising, personalization and charging content providers for content delivery.

It is for this reason that Odlyzko urges carriers to find ways to accelerate traffic growth rather than merely stay ahead of it. If traffic keeps growing at about 50%, it will only offset declines in the price of optical networking technology, he said, leaving little room for profit.

“I'm sometimes amused or dismayed when telecom people talk about bandwidth prices stabilizing,” Odlyzko said. “Can you imagine Intel saying, ‘We don't need to produce more powerful chips; things are stabilizing’? You just expect that each year, Intel will produce a chip at the same price as before that has more processing power. And the gain comes from producing more chips. Something like that should also apply in telecom with transmission capacity.”