Chip Startups Process Profitability
Right now, network processors remain the domain of public companies, with Intel Corp. (Nasdaq: INTC) and Applied Micro Circuits Corp. (AMCC) (Nasdaq: AMCC) leading the pack in terms of revenues. But a few of the private firms appear to be doing quite well. And one, Wintegra Inc., is starting to talk up an IPO.
"As long as we leave '04 at a profitable level, we're talking '05 for an IPO," says Cam Witt, Wintegra's director of technical marketing.
Of course, the market will decide whether Wintegra's got the IPO mojo. But trash talk is refreshing in an industry that hasn't heard such exuberance in years.
Profitability for Wintegra would require a little more than $15 million in revenues in 2004, according to Witt, who contends it's feasible even if the company adds no new customers. "If the current customers stick to their schedules, we're profitable in '04."
The sector overall is starting to perk up, offering some stability for the survivors. "We saw a big pickup in revenues in the second half of last year," says Bob Wheeler, an analyst with The Linley Group. "The market is finally expanding [in revenues], and it's consolidated down to a relatively small number of vendors."
Another startup spouting good news is EZchip Technologies, which announced its NP-2 family of processors last month and claims 33 customers, four of which have begun shipping systems in production volumes (see EZchip Intros NP-2 Network Processors ). That means 2004 will be the first year EZchip gets real money, as opposed to the loose change collected from samples. "Revenues this year will be based mostly on production," says Eli Fruchter, EZchip CEO.
And Bay Microsystems Inc. is claiming it could become profitable this year, having found steady work with the U.S. government's next-generation networking plans (see Startups Spout Bravado at NPC West).
It's too early to call these startups successful, but their performance looks good next to the parade of players that have recently exited the business, including Clearwater, Cognigine, Fast-Chip, and, most recently, Internet Machines, now called IMC Semiconductor Inc. (see Internet Machines Starts Anew.)
For the survivors, one key to success has been price. Originally intended for core routers, network processors have fared better as access/edge players. That's meant higher volumes but lower margins; EZchip in particular had to discount its product, a concession to the importance of low-priced markets (see Net Processors Aim for Access).
Wintegra claims to be the only vendor that planned for the access market from the start, crafting chips closer to the Motorola Inc. (NYSE: MOT) PowerQuicc microprocessors than, say, Intel's IXP2400. A late start helped. The company launched in 2000 when other network processor firms were already underway, but that meant Wintegra already knew core routers were a dead end. "We kind of took advantage of where the market was," Witt says.
It also helped that Wintegra went multiservice, supporting Asynchronous Transfer Mode (ATM) and Frame Relay at a time when other network processors obsessed over Internet Protocol (IP) and Ethernet. "The comparable ATM products being sold were fixed-function devices from PMC-Sierra and so on, Wheeler says. "Wintegra did the same thing with a programmability that other ATM products didn't give you. They hit the multiservice thing right on the head."
Wintegra also stayed frugal, a reflection of its focus on access networks rather than core routers. The company raised just $39 million in four years, despite building a sizeable staff -- 80 employees, including a design team of 60 in Israel.
At this point, Wintegra claims 100 design wins, including some with Fujitsu Ltd. (OTC: FJTSY; Tokyo: 6702), Lucent Technologies Inc. (NYSE: LU), Texas Instruments Inc. (NYSE: TXN), and Motorola.
— Craig Matsumoto, Senior Editor, Light Reading