By any measure, Netflix (NFLX) is having a really good year. Its subscriber base jumped by 52 percent in the third quarter, and its stock price has doubled since July 1. Analysts and customers are bullish about the Los Gatos (Calif.) company as it moves from a DVD-delivery service to an on-demand entertainment provider and de facto rival to cable TV. Netflix's 16 million subscribers are so eager to stream Sandra Bullock movies—Crash and The Blind Side are currently the No. 1 and No. 5 most-streamed movies—that the company now accounts for 20 percent of all Internet traffic during the typical American evening, according to Sandvine, which makes network monitoring equipment. At the Web 2.0 conference in mid-November, an onstage interviewer asked Netflix Chief Executive Officer Reed Hastings whether the Internet's infrastructure can withstand the strain as his streaming business grows. "If there's anything you'd want to bet on," said Hastings, "it's that technology will make bandwidth faster and cheaper."
That bet may not be as safe as it seems. It's true that history is reassuring, and the steady progression from the dial-up modem to fiber-optic cable has led to bandwidth that easily meets demand. Yet there has been nothing like the double whammy of video and mobile that's under way, say industry executives and analysts. A high-definition movie is magnitudes larger than an e-mail or a Web page, the kinds of content the Net was built to transmit. And there are now more than 50 million smartphone owners in the U.S., many of whom want to catch up on Glee while in line at the supermarket.
The most widely cited estimate of Internet traffic, from networking giant Cisco Systems (CSCO), suggests it will triple by 2014, to 64 exabytes a month. (Monthly traffic in 2006 was 5 exabytes, enough to store every word ever spoken.) By then, more than 90 percent of the traffic will be video. "This is the most dramatic change in the network that has ever occurred," says Michael Hatfield, a serial entrepreneur whose latest venture, Cyan Optics, makes gear to manage the data flood. Michael Howard, co-founder of market researcher Infonetics, says Cisco's numbers may be conservative. "It's only a matter of time before the train wreck happens," he warns. His worst-case scenario: Carriers halt upgrades, leaving consumers with slow connections and hindering Internet innovation.
The issue is as much about economics as technology. For the same $40 monthly broadband fee, consumers can send 1-kilobyte e-mails—or watch the 30-gigabyte director's cut of a Hollywood thriller on their large-screen PC. Unlike with power or water bills, there's no meter to keep gorgers in check. A study from Juniper Networks (JNPR) highlights this "revenue-per-bit" problem. The report predicts that carriers such as AT&T (T) and Comcast (CMCSA) will see Internet revenues grow by 5 percent a year through 2020. Meanwhile, traffic will surge by 27 percent annually, and carriers will need to increase their investments by 20 percent a year to keep up with demand. By this math, the carriers' business models break down in 2014, when the total investment needed exceeds revenue growth. Juniper CEO Kevin Johnson presented these findings at a company event attended by 227 carrier executives. Few of them flinched. "At least half of them said it wouldn't happen in 2014—because it was already happening," he says.
Juniper stands to gain by selling the networking equipment that eases the carriers' pain, so take the company's analysis with a grain of salt. While few experts expect carriers to stop investing in new capacity, there's widespread agreement that a financial crunch is coming. Sanford C. Bernstein (AB) analyst Craig Moffett has studied the issue from the perspective of the wireless carriers. As traffic soars, he expects the revenue per megabit to fall from 43 cents today to just 2 cents in 2014. That means a far lower return on investment, a key measure for telecom companies. "The carriers are faced with an incredible deflationary spiral," says Moffett. Stuart Elby, vice-president for network architecture at Verizon Communications (VZ), says that simply building bigger pipes "feels like a losing battle. I can find ways to build the capacity, but it's hard to justify it if I'm not reaping any benefits in terms of revenue."
The late-November tussle between Comcast and Level 3 Communications (LVLT) shows how the issue can quickly become electric. Level 3, which operates backbone networks that quickly ship bits between cities, recently struck a deal with Netflix to help speed delivery of its streaming videos. The result was a sudden surge in Level 3's traffic, which eventually goes through Comcast's cables to reach the company's subscribers. On Nov. 29, Level 3 accused Comcast of charging exorbitant rates to carry the additional traffic. Comcast shot back that it had no obligation to bear the load for free. The exchange is a sign of the times: Even if the technology is up to the task of shipping huge data packets, no one is sure how to pay for it.
Ultimately, most experts expect that the heaviest data consumers will have to start paying more, most likely in the form of tiered pricing plans. These are already common in Europe and Asia, but Americans are used to no limits. The wireless networks have already moved in this direction: In June, AT&T discontinued its all-you-can-eat $30 a month data plan, forcing mobile consumers to choose between two plans that cap usage at 0.2 or 2 gigabytes per month. A Sanford Bernstein survey found that a third of AT&T's customers felt "more negative" about the company as a result of the switch—even though the average subscriber stays well below the 2-gigabyte cap.
On Dec. 1, Federal Communications Commission Chairman Julius Genachowski gave his approval to caps for fixed-line networks that carry data to home or businesses and said carriers should have "meaningful flexibility...to address the effects of congestion." That might sound like common sense, but it's a blow to advocates of "net neutrality," who believe the carriers shouldn't be able to discriminate against particular kinds of content, such as HD movies.
Such business-model changes are new enough that the big data senders like Netflix haven't yet adapted to them. But on an otherwise triumphant earnings call on October 20, CEO Hastings did concede that AT&T's data cap might limit demand for watching movies on mobiles devices. To Bernstein's Moffett, it was a striking admission. "That was the first time I've heard one of those tech CEOs admit what should be obvious: that you can't simply bet on continued bandwidth availability."
The bottom line: Video may make up 90 percent of all Internet traffic by 2014—a fact that's changing the economics of delivering data.