The race to get in on GPON

As early as this month, the country's three biggest telecom carriers could very well pick finalists in a contest that has been raging for nearly a year among equipment vendors: the fight to win a deal supplying the major U.S. carriers with Gigabit passive optical networking access gear. Many expect AT&T, BellSouth and Verizon Communications to narrow the group of hopefuls into a short list of vendors whose gear will move to the next round of lab tests. For the winning vendors, the prize is potentially large — ultimately of unknowable size, in fact — but beset by serious hazards. And by most accounts, the competition will be severe.

One analyst estimated as many as 17 or 18 equipment vendors may be chasing the Bell request for proposal (RFP) released last November. (As this story went to press, UBS Investment Research said it believed the list of potential vendors had been trimmed to four — Alcatel, Hitachi Telecom USA, Motorola and Tellabs.) But others suggest that number may include both sides of several partnerships, such as the Nortel Networks/Huawei Technologies duo. Counting such pairs as one entrant, the number of suitors is probably closer to 10, analysts said.

In aggregate, the Bells could spend $3 billion to $5 billion on GPON gear through 2010, according to Infonetics Research. However, many expect the price competition for those contracts to be fierce. Though home-side optical network terminals (ONTs) have long been a target of pricing pressure, at least one analyst has predicted the larger optical line terminals (OLTs) that sit in carrier central offices to get the same squeeze in the GPON bid. “The GPON RFP is shaping up as a potential blood bath,” Joe Chiasson, Susquehanna Financial Group analyst, wrote in a report last month. “There is every incentive to push equipment prices down very hard.”

Chiasson had heard that the RFP specified a high number of configurations involving low subscriber take rates, suggesting that carriers were focused on OLT prices. (When take rates are low, most of a PON system's cost lies in the OLT.) However, other sources familiar with the RFP say the low take rates it specifies refer to the first year of deployment, with take rates increasing over time. As the Bells see GPON as a long-term deployment, they're likely to examine costs over three to five years in evaluating a given system's price, a vendor source familiar with the RFP said. Bottom line: The specifics of OLT and ONT variations don't change the general anxiety the GPON vendors are facing.

“Regardless of what the take rate is, [the Bells] are going to have a lot of pressure on vendors to have a low price for the OLT,” the vendor source said. “They're still going to beat us up on price.”

Part of that expectation is guided by recent history. Advanced Fibre Communications (now Tellabs) famously “forward-priced” its bid to be the main equipment supplier for Verizon's current broadband PON rollout. After losing money on its ONTs for several months, the company was forced to acquire a start-up, Vinci Systems, to finally get its hands on some ONTs with positive margins. Alcatel also reportedly discounted its gear heavily to win the $1.7 billion deal as SBC Communications' (now AT&T's) primary fiber-to-the-node (FTTN) supplier. Many vendors expect history to repeat itself with GPON.

“Whoever wins this [GPON RFP], their number-one task for the first couple of years is to try to reduce their costs so they can get back to break-even,” said one vendor employee.

And the GPON market beyond the Bells is already overcrowded. Though eight to 10 players may be chasing the Bell GPON business, the global GPON market will probably support only about five or six suppliers, said Michael Howard, Infonetics' principal analyst.

So why are vendors flocking to GPON so fervently? For one thing, they've been watching carrier capex dissolve for so long that any big new rollout sounds practically mouth-watering to them. Second, whichever vendors supply the Bells will achieve economics of scale that will help them when selling to other customers, such as independent operating carriers, competitive local carriers and municipalities — none of which will apply nearly as much pricing pressure. (In fact, according to vendors, the Bell RFP has already spurred an increase in additional RFPs from other customers aligning their requirements with that of the Bells in order to enjoy the economic benefits that Bell-sized deployments could bring.)

It's also worth noting that Bell GPON contracts are likely to involve much more than products, just as Alcatel's FTTN contract made it the system's chief integrator. But perhaps the biggest appeal of GPON to vendors is that it represents the ground floor of a fundamental redesign of access networks that could span decades. In the near term, Infonetics predicts the market for GPON gear (OLTs and ONTs) will grow from $740 million this year to $1.4 billion in 2009. But no one really knows how big the long-term opportunity is. As one vendor employee put it, “There are some 200 million phone lines in the U.S. today. Those could all be redone over 10, 20, 30 years. That's a huge long-term market.”

One of the areas vendors typically rely on to lower the cost of new technology is the mass production of components inside them. But last spring, when the Bells jointly issued a GPON request for information (RFI), chip vendors were reportedly unenthusiastic about GPON. This was partly because GPON was (and still is) becoming a predominantly North American technology (Asian carriers are increasingly favoring Ethernet-based technologies), and even here, only one carrier, Verizon, is aggressively rolling out FTTP so far.

“It's not as if this is a broad market yet where every country's doing it, like DSL,” said Mark Abrams, marketing director for Passavé, which intends to bring GPON chips to market this year. “Because of that, you haven't had the huge commitment that would generate the fastest turnaround of silicon.”

Also, even as the Bells' RFI was circulating last year, the International Telecommunications Union's GPON standards (G.984.3 and G.984.4) still contained some unresolved aspects. A year ago, the standard offered a choice of three different speeds, but last fall, carrier interest quickly crystallized around one specific speed set: 2.5 Gb/s downstream and 1.2 Gb/s upstream.

“That caught a lot of vendors by surprise,” Abrams said. “If you were to have designed a chip a year ago, you would have had to start over.”

Broadlight announced the industry's first merchant GPON chips last May, shortly after the GPON RFI came out, but those chips would only be available “for sample” by December 2005, the company said then.

Some vendors, such as Alcatel and Hitachi Telecom USA opted to build GPON gear with their own chips rather than wait for merchant chip-makers. In contrast, Fujitsu Network Communications — whose public announcement of its GPON gear last month lagged nine months behind Alcatel's — opted to use commercial chips because the GPON standard was still in motion when the company made its decision (Fujitsu won't specify when that was).

“We didn't feel comfortable spending $8 million to $12 million to do the chip development with the standard still in a state of flux,” said Randy Eisenach, market development manager for Fujitsu's access products. “That wouldn't provide an interoperable solution.”

Fujitsu also considered using FPGAs (field-programmable gate arrays, which can be reprogrammed if standards or requirements change), which Eisenach believes might have allowed the company to bring a product to market sooner but left it vulnerable to cheaper, denser, lower-power gear from competitors four to six months later. Still, he downplayed the impact of silicon economics on GPON's overall cost equation.

“If you're paying $5 or $2 more per chip, I'm not sure that's what makes or breaks a cost structure for a service provider in a widespread deployment,” Eisenach said.

Perhaps because Fujitsu wasn't one of the quickest vendors to the GPON market, it is taking a longer-term approach. To differentiate itself from the pack of competing vendors, the company is offering a coarse wavelength-division multiplexing (CWDM) upgrade to its system that allows carriers to multiply bandwidth speeds up to eight times. (So instead of 2.5 Gb/s heading into a splitter to be divided by up to 32 households, those households would split more than 10 Gb/s, for an average of more than 300 Mb/s each.) Fujitsu isn't necessarily arguing that residential users will need more bandwidth in five or 10 years than standards-based GPON gear currently delivers; it's simply offering carriers insurance in the face of that uncertainty.

No one expects GPON to be the last access network. In fact, people are already talking about WDM PON as the putative next step. The ITU's FSAN group, which produced the GPON standard, hoped to begin work on WDM PON this year. But as one vendor put it, “It's being talked about as something that should be talked about.”

So, in the meantime, for access equipment vendors, the GPON market seems to stretch out pretty far toward the horizon. And at their feet, it begins with a short list of vendors to be selected as early as this month.

GPON TIMELINE

2005

APRIL

Bells issue GPON RFI

MAY

Alcatel announces GPON gear

AUGUST

Alcatel names first GPON customer

[FALL]

Carrier interest solidifies around 2.5 Gb/s downstream, 1.2 Gb/s upstream

SEPTEMBER

Wave7 Optics adds GPON support

OCTOBER

Hitachi names first U.S. customer of 2.5 Gb/s GPON gear

Lucent debuts multimedia access platform for GPON

NOVEMBER

Bells issue GPON RFP

Calix announces Optical Solutions acquisition

2006

FEBRUARY

Fujitsu announces entry into GPON space

Nortel, Huawei announce GPON joint venture

Calix closes OSI acquisition, announces 2.5 Gb/s gear

MARCH

Potential selection of Bell GPON supplier short list