This year, the revenue from Gigabit Ethernet chips will exceed that of Fast Ethernet for the first time.
This year, the revenue from Gigabit Ethernet chips will exceed that of Fast Ethernet for the first time. This transition, the likes of which happens only once every seven years, is both creating and taking away opportunities for chip vendors.
On the plus side, revenue is growing as corporations replace older equipment with new gigabit devices. IDC expects the market for Gigabit Ethernet chips to skyrocket, from $450 million in 2002 to $1.1 billion in 2004. Much of this growth is at the expense of Fast Ethernet chips.
Whereas Broadcom and Marvell built most of the early Gigabit Ethernet products, Intel, Vitesse and Agere are making strong plays to get a share of this revenue. Intel has long been the leader in Gigabit Ethernet controller chip sales, but until recently, its controllers were manufactured by Marvell. Last year, Intel began shipping a controller (MAC/PHY) completely of its own design. The company plans to eventually use its 90-nanometer mixed-signal technology in this market, taking control of its own destiny.
Vitesse Semiconductor has been a leader in Gigabit Ethernet switch chips but did not have its own PHY technology. The company filled this hole by acquiring Cicada and will now offer a complete solution.
Agere Systems' sizable ASIC business includes many Gigabit Ethernet designs, but this revenue stream is threatened as OEMs shift to standard products. By acquiring two small startups, Massana (Gigabit Ethernet PHY) and TeraBlaze (Gigabit Ethernet switch), Agere will be able to sell its own standard products, but not until later this year.
Market entry will be tougher for other vendors. The acquisition spree has left no remaining startups with proven Gigabit Ethernet technology, forcing other prospective entrants to develop technology internally.
That's not easy. While Fast Ethernet MACs and PHYs are as common as jelly beans, a Gigabit Ethernet PHY is a highly complex design full of the analog and digital signal processing required to force 1 Gbit/second through uncopper wiring. Few have been able to demonstrate a working Gigabit Ethernet PHY.
The 10-Gbit Ethernet PHY will be 10x more complex. The standard for 10-Gbit Ethernet over copper is not yet defined. We have a few years before we need 10-Gbit Ethernet, but as the Gbit Ethernet transition shows, a head start could be crucial.
Linley Gwennap is founder and principal analyst of The Linley Group and co-author of "A Guide to Storage Networking Silicon" (www.linleygroup.com/npu).