The historic telecommunications crash of 2001 is burned indelibly into Ciena CEO Gary Smith's memory, yet he is cautious about viewing rising anxiety over today's AI boom as an echo of that past.
Smith, 66 years old, has led the optical networking pioneer since May 2001. He led the company through the telecom nuclear winter of the early 2000s, which was more economically devastating than the preceding dot-com bust. It was marked by a glut of unused fiber, a collapse in prices, frozen venture markets, massive bankruptcies, widespread job losses and a recession.
Ciena experienced the wild swings of that era. It went public in the late '90s in a record IPO that valued the company at $3.4 billion. Its market capitalization peaked at more than $40 billion a few years later, but fell more than 90% during the bust.
The company survived that test and now has a market cap of about $45.75 billion. It ranks fifth on the WSJ Leadership Institute's ranking of best boards. You can read the full board report here.
The telecom era and today's AI boom frame Smith's 25-year tenure as Ciena CEO. And while they share obvious similarities as innovation-driven, capital-intensive infrastructure build-outs, Smith argues that their differences are far more pronounced. He doesn't view the telecom cycle as an indication of what comes next for AI. The AI era is bound to produce its share of losing bets, but he says it needs to be understood on its own terms.
"Having lived through the pain of that, I'm suitably paranoid around all of this, right? As you'd expect. But I honestly believe that there's a whole set of different contextual things here in many ways," says Smith.
Compounding waves of demand
Smith says the AI era won't follow the familiar boom-and-bust pattern that has defined many investment cycles such as telecom and the dot-com era. He expects this period to set its own course and have even greater impact.
"This is a long ramp of compounding waves of application and demand," he said.
The first of those waves, the connection of multiple data centers, is under way and still ramping up. The next wave, known as inference computing, is just beginning, according to Smith. The concept of inference refers to end users such as businesses, organizations and individuals putting to use previously trained models to solve actual problems.
The subsequent wave is based on the rise of AI agents that complete tasks with varying degrees of autonomy, complexity and duration. The next wave in sight will be driven by robotics, or what is known as physical AI.
"I think the cascading hierarchy or critical path is now very much the network. And I think people have grossly underestimated the amount of infrastructure that's going to be required," he said.
To meet that demand, Smith said "Ciena is pushing the boundaries of physics." The company has developed proprietary 3-nanometer versions of its WaveLogic chips capable of transmitting data at 1.6 terabits per second over a single fiber strand and is developing a 2-nanometer version as well. A nanometer, or 1 billionth of a meter, refers to the size of a transistor, and the smaller the size, the greater the number of transistors that can be placed on a chip. The 2- and 3-nanometer standards are considered state of the art.
A structural difference
About 94% of the fiber laid during the telecom boom was dark, or unused, according to Heath Terry, global head of technology and communications research at Citi. "There are no dark GPUs," Terry, who leads AI research as part of his role, told me in a prior conversation.
Smith said that is only one way in which today's AI market is structurally different from the '90s telecom boom. The AI market also benefits from a generally much sounder customer base, including highly profitable cloud-computing giants.
"Today, we haven't been able to keep up with demand," Smith said.
Ciena expects to have a compound annual growth rate of 30% for the fiscal years 2026 through 2029, and is bringing more capacity and capability online, according to Smith. The company expects to come out of fiscal 2026 with an order backlog of about $10 billion.
In Smith's view, AI isn't simply another tech cycle. He views it as a true industrial revolution.
"We don't fully appreciate it yet, but it's going to change how we live," Smith said.
Comments