Investors Keen on High-Flying Chip Firm.

Even as Internet content companies drop like flies, the race to deliver data, voice and video traffic ever faster is spurring investments in new firms convinced they can do just that.

One such company is Agoura Hills-based Internet Machines Inc., which despite being less than a year old has garnered over $40 million in venture capital and has grown to more than 100 employees.

Internet Machines is one of several companies working to build ever-faster network processors -- the semiconductor chips embedded in large Internet infrastructure equipment such as routers and switchers that send massive amounts of digital data through cyberspace.

The next-generation "wafers" are distinguished from previous generations by two very important features: adaptability and speed. Until now, companies have developed application-specific integrated components, or ASICs, that work on one particular product within the network. Now, a few startups are developing chips that are programmable and can be upgraded as the network infrastructure adapts.

In addition, these silicon chips are up to four times faster than the current generation, moving 10 gigabytes of data per second as opposed to the current top speed of around 2.5 gigabytes.

As a result, the company that delivers these flexible, high-speed silicon chips first is likely to make a lot of money.

"The Internet is predicated on information moving through it at very high speeds," said Internet Machines President Chris Hoogenboom. "The Internet is evolving very quickly with many different protocols used. Right now, a chip for an Internet router is different from a chip for a modem. Because there are so many protocols, it's useful to have a programmable chip to read all of them. It allows products to evolve as Internet protocols evolve."

With the Internet flowing through modems, cable, DSL and cell phones, the programmable network market is poised to explode, Hoogenboom says. Telecommunications companies of all kinds will need these chips to survive, and research firm Dataquest predicts sales will jump from $816 million next year to $2.8 billion in 2004.

"Network processors have already and will continue to change the way the industry does business," said John Metz, vice-president at Sterling Research, a Massachusetts-based technology research firm. "Until now, there's always been a compromise between how much you do in hardware and how much you do in software in building these networks. With software, you can go in and program (systems), change things. It gives you flexibility. With hardware, you get extremely high speeds. Network processors are fully programmable. There's no forklift upgrade necessary."

That's going to be of paramount importance to companies that design the infrastructure systems behind the Internet such as Lucent Technologies Inc. and Nortel Networks Corp.. Network processors will make the job of such companies easier by saving time and money replacing ASICs as the system evolves.

Hoogenboom says the company has several potential customers and partners but declined to give details.

"There are so many new systems, so many of these large (network) boxes that aren't out for more than a year or two," Metz said. "If you have network processors, if something changes, you can program it and keep the system in place for up to five years. From the guys who make the chips to the Ciscos and Nortels, to the start-ups, everybody wins along the value chain."

Internet Machines was started by four partners. Three are engineers who came out of network design company Xylan Corp.

Internet Machines' next challenge is getting out a product. The chips take months to design and manufacture, and Hoogenboom predicts production will start before the end of 2001.

As Metz points out, venture capitalists are standing in line to give network processing companies money, and Internet Machines has a few competitors, one an Israeli company and another in Silicon Valley. But Hoogenboom's outfit just raised more than $31 million in its second round of financing from investors, including Morgan Stanley Dean Witter, Redpoint Ventures and Meritech Capital Partners. That money should be more than enough to get the company through next year and deliver its product to market.

"These guys were a very successful team at Xylan, and they're targeting what we think will be a very high-growth period in technology," said Meritech Capital Managing Partner Mike Gordon. "Starting (to get product out) by late 2001 we think is fast enough, given the power and performance of Internet Machines' processors. There's nothing out there that can compare with what these guys can deliver."