How AT&T Got the Internet

No one's exactly sure when AT&T Corp. got Internet religion. The idea that the Internet will be the driving force in 21st-century telecommunications has been percolating for a while, and as far as anyone can recall, there was no single moment at AT&T when the light bulb went on. But somehow the company has changed its entire mind-set -- identity, even -- from a tired long-distance phone company to one of the most bullish Internet communications forces in the world. The obsessive network acquisitions of CEO C. Michael Armstrong have attracted the spotlight, but the transformation seems to have happened before his arrival at the company. And it seems to have happened far away from the executive suite, in the enclave now known as AT&T Labs, where small groups of Internet researchers quietly mapped out the revolution to come.

1996 was the year AT&T should've dropped the ball. Robert Allen was CEO, presiding over the "trivestiture" that split the monolith into three smaller companies. AT&T focused on long-distance service; Lucent, on networking hardware; and NCR became a high-end computer company. There were staff purges and forced reorganizations, but somehow in the midst of the confusion, a small miracle took place. A group of researchers within the Labs emerged with a clear sense of the company's future.

Under Allen, AT&T had hired two key people from Silicon Valley. The first to come aboard, in March 1995, was Dado Vrsalovic, hired to head the new IP Platform Group, working on the project that became AT&T WorldNet, the ISP service. Under Vrsalovic's direction, AT&T engineers and scientists first came to understand the power, potential, pitfalls and promise of Internet platforms.

The engineers realized that AT&T needed to design a new, unified network built around the Internet Protocol. IP describes the way data travels on the Internet, and they realized that it could be the best means for transporting all communications, voice and data.

AT&T is even convinced IP will replace its century-old circuit phone networks. Circuit switches are the huge devices phone companies have used for over a century to route traditional analog voice calls. The industry expects circuits to be replaced by packet switches, which turn voice traffic into digital packets that can be let loose across the Net. Circuits are expensive, because every phone call requires its own circuit. By breaking calls into packets, Internet telephony vastly increases the capacity of fiber-optic lines. AT&T was not alone in this realization; competitors like Qwest, MCI WorldCom and Sprint are all jumping on IP.

And while AT&T is busy trying to get its old-world networks to fit the new IP world, nimbler startups are introducing IP telephony products years before AT&T will be able to do so. "While they try to do it internally, innovation has gone on all around them," says Jonathan Heiliger, a Silicon Valley telecom VC and the former CTO of Frontier Global Center.

Vrsalovic, formerly the chief scientist at Sun Microsystems, is an imposingly tall man with a thick Croatian accent and a powerful voice. Vrsalovic was able to convince some great minds that AT&T was going to change the world. People didn't just agree -- they moved to New Jersey. In about eight months, these scientists built the Internet platform that would become WorldNet.

But in 1996, after much of the work on WorldNet started, things began to get a little weird within the Labs. The trivestiture had begun. The legendary Bell Labs, which set the standard for innovation by producing the transistor, the laser and sundry other technological achievements, split from AT&T -- taking two-thirds of the staff -- to become part of Lucent. What was left was a smaller, leaner R&D operation, simply known as the AT&T Labs, which now had to forge a new identity.

Meanwhile, another even more influential Internet proponent had come aboard. In April 1996, AT&T hired Dave Nagel, then chief scientist at Apple, to be its chief scientist and CTO. Nagel insisted that he be allowed to spend half his time in California, half in New Jersey. It was a lonely first few months. He spent six months as the only California employee, first in a small office in Los Altos, Calif., then in San Jose, where the AT&T IP Platform Group finally settled.

"AT&T really did need more than just an outpost here," says Nagel. "All the development in our industry is happening here. The deal flow, the opportunity flow, is all enhanced here compared to New Jersey."

Vrsalovic, too, knew that he needed to move his group closer to the Internet action. He also wanted to shield his engineers from the confusion in New Jersey. In late 1996, he moved the group to San Jose.

"A lot of us had just sold a house in Silicon Valley to move to New Jersey," says George Vanecek, chief scientist with AT&T Labs IP Technologies. "And then we get out there, and the breakup was nothing but confusion. We wanted to get far away from that, so two months after I got there, I had to pack up and move back to San Jose."

Thanks to Vrsalovic's decision, Nagel got his wish to have his network architects in San Jose. What looked like a hasty retreat turned into a significant advance: AT&T finally got a foothold in Silicon Valley.

Together, the 400 researchers in New Jersey and San Jose -- through hiring and internal development -- have slowly put together a small army of the most experienced people in the business. Focused on building a next-generation Internet platform, they tend to have the requisite missionary zeal. Hamid Ahmadi, for example, who heads the networking and distributed-systems group, is downright giddy about the challenge before him. As he quickly sketches his ideas for the AT&T network architecture on paper -- great big clouds representing the Internet, little boxes illustrating different AT&T hardware components, and lines crisscrossing the page to connect it all -- his eyes take on a fervid gleam.

"There are two answers to how we can win the war. Mike Armstrong's answer would simply be execution. My answer is much, much more complex," he says. "There's a battle between us and the computer companies, and I believe companies like AT&T could win this battle, because not every device will be a computer. We'll build intelligence into the networks, and the computers will just be dumb devices."

But IP traffic has a long way to go before it overtakes AT&T's core business. Traditional circuit-switched voice traffic still accounts for 45 percent of AT&T's network traffic, and it will continue to be the company's bread and butter for some time.

The amount of IP traffic, however, is changing dramatically. "The rate of growth for IP traffic on our networks has been 400 percent per year," says Inder Gopal, VP of Internet architecture at AT&T Labs. "I don't believe we'll see sustained growth of 400 percent per year, but growing at that rate for three to four years, IP will easily overwhelm everything else on the network. In a few years, all our investments will be on Internet Protocol."

So after working on the WorldNet program, Vrsalovic and company began working on an ambitious new project, code-named GeoPlex, which laid the groundwork for the emerging IP platform. Vrsalovic and the IP Platforms Group were beginning to realize that IP could be the underpinning of a common platform for everything within AT&T.

When Vrsalovic came aboard, most engineers' understanding of IP was immature at best. "This is a caricature, but when I first started interviewing telecom engineers, all they wanted to do was to throw a bunch of servers into one big building and connect it to the world with one fat pipe," says Vrsalovic. "After WorldNet, we realized that the solution was much more distributed. We understood the complexities of how to run a service on the Web."

Moving to IP will change the AT&T infrastructure from many separate networks into one integrated network. And it will use any medium possible to reach the consumer, from copper wire and cable to wireless.

The engineers now envision an open platform, modeled on the PC revolution of 10 years ago. "The development of software and applications has been refined quite a bit compared to 10 years ago, thanks to the PC revolution," says Nagel. "We've just now discovered the concept of open platforms, reusable code, and open standards and specifications."

Software developers will be able to repurpose an e-mail program, for example, for cable, cell phones or the Internet. Once standards are set, code and applications can be reused in different environments, all running on the same IP backbone.

IP may be the most efficient network platform created, but for AT&T engineers, who obsess about the "five nines" of network reliability (that is, that a network must work to 99.999 percent of the time) moving to IP is a scary proposition. If you consider how reliable your telephone is compared to your Internet access, you can understand why AT&T engineers are leery of the Web. As Frank Field, VP of IP platform deployment planning at AT&T Labs, puts it, "IP is said to be 99.9 percent reliable, which may sound good to some people, but to me, that means 1,000 defects per million."

Going to IP, then, will be difficult. "We are going to have to cross a chasm," says Field, echoing Geoffrey A. Moore's high-tech marketing best-sellers. "And once we get there, we'll only find ourselves in the eye of a tornado."

If AT&T under Mike Armstrong has a mantra, it's "own it, don't rent it." AT&T aims to buy or build networks, rather than being forced to lease, borrow or rent access, especially from the regional Bell companies, which charge exorbitant rates to use local phone loops. Armstrong has gone on a $137 billion shopping spree, buying disparate networks like the IBM Global Network, cable giants TCI and MediaOne, Vanguard Cellular and the TCG local phone business.

"If you look at these networks in isolation, it's hard to make a business case for us to be running them all," says Vrsalovic. "But if you look at the overlap -- like places without a fiber loop, but you can find a cable head end -- you can economically get the same services rolled out practically everywhere."

When then-CEO Robert Allen hired him, Nagel's first project was to work on ways to get data and phone services to a consumer using technologies other than the traditional copper wire, such as cable, satellite or wireless. At that time, the company primarily owned phone networks.

Then Armstrong came along. He had been in charge of development at IBM when the National Science Foundation came by and proposed collaborating on the Darpanet project, which evolved into today's Internet. That background has certainly helped Armstrong understand the medium's growth potential.

When Armstrong began suggesting that AT&T buy cable networks, Nagel and his scientists were ready and eager.

"When a bunch of us sat down with him and he started talking about acquiring cable companies, a lot of heads started nodding," says Nagel. "Oh, God, it was a huge difference. It was a wonderful feeling. Mike came from the computer industry, and at Hughes Electronics (Armstrong's previous job), he was competing against the cable industry. I have very effective conversations with him. He has the same ideas. It's very different than the old AT&T. Once he knows what he wants, it's 'Katie bar the door,' because he's not afraid to put the money where his mouth is."

But Armstrong's aggressiveness can be overwhelming. Nagel has suddenly found himself working with $137 billion in new networks. "I never thought about this stuff in orders of magnitude of billions," he says. "It really presents a high hurdle for new development. If I go to a finance guy and ask for $10 million to build a new service, he says, 'Great. How many billions of dollars are we going to make off this?' If you don't come up with a lot of revenue, it's just not interesting to AT&T."

Nagel thinks AT&T should pursue smaller projects that generate little or no revenue, like its investment in a2b, the Internet music-streaming technology. "It makes it hard to start an entirely new thing, to stake a claim to an emerging technology. Somebody saying, 'Gee, that sounds great. How many billions will it make?' makes it a lot harder to get something started." The fact is, while the Labs make up a great engineering environment, they comprise a very different animal from the fabled Bell Labs. It's not necessarily an environment that will produce radical innovations.

Armstrong has decreed that all of these diverse networks should work in concert. He has asked his engineers to assemble a global, Internet platform that can pipe every basic communications service to any consumer or business. But because AT&T has acquired decades of legacy data and circuit-switch infrastructure, integrating all these old data repositories with a new network will be a monumental task. "Circuit networks have evolved for so many years that they are pretty confusing," says AT&T Labs' Gopal. "They are not modularized and don't have clean interfaces, and customer data is stored in, like, 16 different places."

The AT&T platform is also designed to prove that AT&T is not just a consumer company, but a serious business-to-business provider. The first test of the platform, designed for multinational companies that are trying to support disparate offices, will be a joint venture with British Telecom. The expectations are pretty high. The operation is expected to have revenue of more than $10 billion, growing more than 15 percent a year. But only time will tell if these companies are ready to abandon the applications and older networks on which they have been relying.

The kinds of applications the AT&T and British Telecom joint venture will offer include IP voice calling, roaming international Internet access, Web hosting, e-mail and virtual private networks. Virtual private networks, which enable company offices to run corporate applications over the Internet securely without paying for a dedicated line or connection among remote offices, are expected to be the killer app that will attract new business customers.

Mike Armstrong has certainly brought new energy. but that doesn't mean AT&T can overcome the burden of its own weight. It is still a behemoth, and Armstrong continues to order draconian budget cuts to boost margins. The company absorbed nearly 18,000 layoffs in 1998, and has another 3,000 network-engineering layoffs to go as it tries to integrate the disparate networks it's buying. And the company has already struggled with its first major acquisition -- TCG, a leading local phone company. Robert Annunziata, the former chairman and CEO of TCG, became president of AT&T's $22 billion business services group last September, but quit soon thereafter to join telecom startup Global Crossing. Meanwhile, the TCG group has failed to hit its revenue growth targets, falling short of the $330 million to $450 million in new 1999 revenue that AT&T had predicted.

Not to mention that the technology AT&T is banking on -- cable -- remains unproven for the traffic AT&T plans to dump on it, including phone, TV and the Internet. The engineers are confident it will be possible to use the cable as Armstrong envisions. "It's not a trivial task, but it's not insurmountable," says Gopal. However, analysts say the $1.8 billion the company estimates it will take to upgrade the networks may be far too low. Some predict that such changes will cost tens of billions of dollars.

And while AT&T is gearing up to deal with new competitive realities, the Labs themselves have had to undergo a significant transformation. They are now under constraints that keep them from being the hotbeds of innovation that their predecessor, Bell Labs, once was. Since AT&T's breakup in 1984, the company is no longer an overcapitalized monolith. In an effort to remain competitive, the company is pouring all of its resources into shorter-term applications.

Scientists who once could've disappeared into the basement to work on the kind of blue-sky projects that led to the invention of the laser are now expected to meet strict revenue expectations and to show real-world success. A physicist at IBM Labs says, "Quite simply, AT&T Labs is not Bell Labs. They don't do any of the pure science that made Bell Labs a great institution." While the AT&T research park is a safe distance from the interference of corporate suits, researchers still chafe at their constraints.

But AT&T's engineers are proving efficient at laying the groundwork for the next generation of networks. Employees seem to be optimistic about the company's progress. Al Hanson, VP of AT&T local services, was with AT&T 20 years ago, but quit because "AT&T at that time was always downsizing," he says. "That was a reality I didn't want to deal with." So he left, eventually becoming an executive with TCG, the nation's second-largest competitive local exchange carrier -- until AT&T bought it last year for $11 billion.

Hanson says that while the atmosphere is much improved from the dark days he experienced 20 years ago, the cultural differences between AT&T and startups is dramatic. "TCG was a lot more entrepreneurial, more 'can do,' and had a lot less bureaucracy than AT&T," he says. "When Mike Armstrong acquired us, he said one of the reasons was to bring some of that entrepreneurial attitude here. They're getting there, but it's a work in progress."

Of course, there's still one great advantage to working at AT&T. "The money's good," says Hanson. "My capital budget in 1999 is twice what it was in '98. AT&T has strong financial wherewithal to build out networks."

AT&T has invested an estimated $1 trillion in its networks over the years, and it can't just give up on the technology that has kept it in business since 1885. But a long, slow, painful transition is underway. "God made the world in six days and rested on the seventh day, but that was only because he didn't have an installed base," says Dado Vrsalovic, reciting the old engineering joke. "I cannot stop the world and start from the beginning. I would like to, but I'd be without revenue for the year."

Join the newsletter!

Error: Please check your email address.

More about AT&TBell LabsBT AustralasiaGlobal CrossingHughes ElectronicsIBM AustraliaLucentMCIMCI WorldComMediaOneNCR AustraliaQwestSprintSun MicrosystemsSystems GroupThe Chief ScientistWorldCom

Show Comments

Market Place