Company Builder

But RFI isn't exactly swimming in a small pond. Many of the domains that Raza is pursuing -- optical-networking systems and components, routers, traffic-engineering systems, wireless broadband -- are practically industries unto themselves. Each field is ultra competitive. Still, it's not as if RFI is taking on a score of pet-food portals. Raza figures that by competing exclusively within the high-barrier-to-entry domain of broadband, RFI can relentlessly innovate during the other critical phase of a startup's evolution: product development.

Bill Hunt and Waqar Shah are the key engineering executives for getting products on a fast-track growth cycle. Shah, 46, who does product-development work at RFI, held top engineering and management posts at AMD, National Semiconductor, and Trilogy Systems Corp. Hunt, 33, who oversees all things silicon, worked in similar key positions at Avanti Technolgies Inc., IBM, and National Semiconductor. Part of Hunt and Shah's mission at RFI is to give partner companies the freedom to put their time and energy into real product-design work.

Because RFI focuses exclusively on broadband, it has found a way to reuse many of the core designs of enabling technologies. "One thing that we've all got to do is build boxes and systems," says Shah. "It takes a minimum of three to six months for a startup to develop the foundation for these systems. Our goal is to install this foundation technology, such as a best-of-breed CAD system for designing complex asics, in each of our companies. When I was at Trilogy, we had to develop all of this stuff from scratch. At RFI, there's no need to spend an infinite amount of time inventing stuff. We improve it, and then we reuse it."

Hunt and Shah are fanatical about cutting time to market, and they apply a time metric to every step of the product-development phase. First, they find out when the customer needs the product. From that deadline, they create a work-back schedule, starting at the ship date and ending at day one. That master schedule typically covers more than 800 items.

Startups fall behind schedule, says Shah, because they never put together a real-world schedule in the first place. "Even smart entrepreneurs don't make good program managers," he says. "They've never run large programs before. Most of the entrepreneurs I've seen are very good technically in a particular field. They may have deep knowledge of a particular algorithm. But they're clueless about issues like testing for manufacture. So we invest the time to develop a scheduling template, and in the long run, we save them time."

As this issue of Fast Company went to press, none of RFI's partner companies had shipped product. But several of them say that they are on pace to hit their release date before the year is out. Now ensconced at AMCC, YuniNetworks's first chip was taped out in July; its switching fabric is due out by December. Sharma says that Pacific Broadband's first two products will hit the market by year's end.

"The window of opportunity doesn't stay open for very long," concedes Hunt. "That's why as soon as we've defined a compelling product, we devote all of our energies to blasting through that window and getting the product to market. We put a minimum of 10 engineers on the project. We hold weekly check-ins to ensure that we're hitting our internal milestones. Get the product to market, and you can live to innovate another day."

Are You Scalable?

Raza never intended to build a 21st-century business foundry. He had a different plan in mind when he left AMD: He would launch a small fund and invest in broadband startups. But he was too successful. His goal was to raise $5 million. The fund closed at $100 million. Though a small figure when compared with the billion-dollar funds that some big venture capitalists are raising, that $100 million exceeded even Raza's loftiest expectations.

Still aglow with his success, one morning Raza had breakfast with Bruce Dunlevie, general partner at Benchmark Capital. "I told Bruce that the fund was going incredibly well, and he asked me who my partners would be," Raza recalls. "I gave him the names of two people." Dunlevie proceeded to rate the two, using Raza as a baseline. By Dunlevie's reckoning, both candidates rated below Raza. (Presumably, the unnamed candidates would have had a different take, but they never got the chance to voice it.)

Then, says Raza, Dunlevie made his point. "He said, 'Collectively, the three of you are less attractive than you alone. So the problem that you have, Atiq, is that you are not scalable.' And I must admit, he stumped me."

During the next few weeks, as Raza wrestled with Dunlevie's challenge, he thought about his own journey through the world of work. He had come a long way since 1979, when he arrived in the United States from his native Pakistan. He had spent seven years in various engineering and management positions at high-tech firms. Then in 1988, he became VP of engineering at NexGen Inc., a microprocessor company. Eight years later, then president, CEO, and chairman of the company, he sold it to AMD for $850 million.