National Fiber Optic Engineers Conference

Sean Maloney
Dallas, Texas, USA
September 16, 2002

SEAN MALONEY: Good morning, everybody. Many of you are probably wondering why are we sitting here in this hall listening to somebody from Intel, which is a company most famous for the microprocessor and consumer advertising and ads with people dancing to 1970s and '80s music?

Well, you are still in the right place. As some of you may know, over the last four to five years, Intel has been steadily increasing its research and development in the core area of optical technologies. It's been clear to many in the industry that optical technologies would move into the mass production mainstream, inevitably, in just the same way as the microprocessor did. As computer performance gets more and more and more, copper runs out of its abilities to carry signals, and so optical technologies move out of the long haul area into the metro area, inevitably into the enterprise, and ultimately, of course, into the home as well.

So Intel, along with many others, has been considerably increasing its R&D, and we're in the position now where we have a whole series of technologies going all the way from photonics all the way into optoelectronics and into advanced transponders. I'm going to give you a snapshot of some of those technologies this morning and what they mean to the industry.

Now, of course, it has been a really, really, really tough year. The question I'm going to try and answer is, "Is there light at the end of the tunnel?"

Coming from a chip background and the semiconductor business, we have been used to recessions. For the communications industry, this, of course, has been a particularly brutal recession. We had an explosion of growth '97, '98, '99, probably the biggest boom the industry has ever seen. And then, right now, we're in the biggest recession, almost like a hangover after the world's biggest party.

The semiconductor business, which increasingly is linked and tied with the optical technologies industry, the semiconductor business has been through these recessions multiple times before. In fact, pretty much since the beginning of this industry, it's been through those recessions six times.

A couple of these have been macroeconomic recessions produced by things like oil shocks. But there's also been a series of oversupply, undersupply. Every single time the industry's gone into these recessions, some of them very, very severe, but every single time, it's come out stronger than it went in, with the industry being not only bigger in terms of units, but bigger in terms of revenues, technologies, research and development, and profit and so forth. Every single time it's gone in, every single time it's come out, and the industry's been stronger at the end of it.

And we believe that's very much also the case for this industry as well.

But when an industry goes into a recession, often there is a catalyst or there is some technology that pulls us out. There was an economic recession, albeit on a smaller scale, at the beginning of the 1990s.

At the beginning of the 1990s, after the Gulf War, there was a bit of a slowdown. There was already an infrastructure out there beginning to emerge of computers and communications. But then along came the browser, which at the time seemed a very simple, obvious thing, but turned out to be one of the great breakthroughs in information technology.

The browser enabled ordinary people to use computers without any training and kicked off an incredible growth cycle of PCs and information technology. In many ways, the browser was the entry path into the whole of the Internet boom of the middle and late 1990s.

Now, once again we are in a recession, albeit more severe. There are similarities to the one at the beginning of the 1990s. Once again we have a president called Bush. Once again we have issues in desert areas.

Right now, again you'd look around and you'd think, what is there out there that will kick start us to emerge from this recession? High tech and IT, of course, is a critical part of the world economy. What is there that's going to reignite optimism and growth?

Well, one area that continues to do well is broadband. Many people for a while have been saying broadband is the next big thing.

Broadband has not been happening as quickly as many of us would have liked. We'd like to see it doubling on an annual basis. But many service providers would come back and say, no, that's not fair. It's growing at 25, 30 percent a year, on par with almost any previous technology.

Of course, wonderful things happen to our industry with broadband. As people start to use broadband instead of conventional links to the Internet, not only do they send more data backwards and forwards, generating more demand for our networks, but also they spend more time on the terminal. In fact, the average broadband user now, just recently the amount of time they spend on their broadband PC is now greater than the time they spend watching television.

So broadband is a generator of traffic and an area of hope for us. We are in the middle of a conversion from narrowband to broadband, a conversion that's going to take the best part of the next five to 10 years.

Now, inside of that broadband area, one of the key technologies that's going to be driving that conversion from narrowband over to broadband is 802.11.

802.11 has the ability to transform the way that people use computers. In many ways, 802.11 is the analog to the browser. For the first time, the consumer electronics industries, the computer industry, and worldwide service providers -

(Phone rings.)

(DEMO begins and ends.)

SEAN MALONEY: The demo is all live. None of that is demo technology. It's all live stuff. More and more houses have 802.11 networks in them. Over the next year or so, 802.11 is going to become standard in notebook computers. Within two to three years, it'll be standard in PDAs. It's going to become standard in desktop computers, and it's going to change the way that people do computing.

Up until now, with the exception of a cell phone that generates a small amount of traffic, you will only generate a traffic load on the network when you're in one place tied to a piece of thick copper. That's all about to end. And it's going to end very, very, very fast. People are going to be installing 802.11 networks throughout places like this. They're going to be installing 802.11 networks in public spaces, in libraries. They're going into schools already. My daughter's school has 802.11 throughout it. It has an 802.11 umbrella spread across the school. You're going to see 802.11 going into not only places like Starbucks you've already heard about, but also into hotels all around the world.

It is one common communication standard. It is not like ISDN, which had disagreements for years. It's not like the differences between SDH and SONET and so forth. This is one standard for the entire world, service providers, computer industry, communications electronics companies have all jumped to.

The consumer electronics industry is going towards 802.11 as a common way of tying up digital TVs, PCs, set top computers and handhelds, moving to 802.11a because of the quality of video delivery, up to 50 megabit per second. Hot spots are going to be using 802.11b, 802.11g, and 802.11a as well to go and reach out into broader areas.

If I look at the area near where I live -- San Mateo, California -- 802.11 is being put in by the public service community so that the police, the ambulance, and the fire brigade can download video clips, information, pictures wherever they are traveling around the metropolitan area.

All of this means that people are going to change the way that they use computers. It's going to happen in the next three to five years. And it's going to end up generating a lot more traffic.

802.11 is absolutely a transforming technology and it's potentially the light at the end of the tunnel in terms of new traffic and new ways of using the network.

Now, let me again just drill on this for a second. Think about the impact of Napster. With Napster, you had a technology that spread very, very quickly where people went from sending, essentially, small amounts of text to music. Napster generated enormous amounts of traffic, unpredictable traffic spikes, quickly.

802.11 has the ability to take this thing one stage further and then also move into video. You're talking here in terms of three to four orders of magnitude step up from data over to voice. And then another two orders of magnitude increase as you move from voice over to video.

These are all huge amounts of traffic, and they're going to be aided and abetted by another phenomenon that will take place over the next year, which is PVR down in the PC. That means personal video recorder is going to be built as standard into PCs. So PCs are going to be able to have terrestrial TV signals digitized and stored on their hard drives. This, again, has the potential to generate another traffic spike.

So all of these things are likely to keep this traffic growth moving ahead very, very rapidly.

On the slide here, I'm showing you what the traffic growth looks like, supposing we just keep at our existing rates. So this doesn't factor in a big exponential increase if video comes onto the network. This is just the same doubling that's been happening over the last five years, that we've seen no impact of recession on the Internet traffic. It's carried on growing. A little bit slower in the United States, but Asia and the rest of the world has come on to keep this steady traffic growth.

Of course, by the rules of arithmetic, if we double every year, that means in 10 years' time, the traffic volume around the Internet will be 1,000 times higher than it is now.

That sounds absolutely incredible and absolutely impossible. I find it difficult to see demographic reasons, when you look around the world, you look at the number of people who aren't on the Internet, the number of people who aren't on broadband, the number of people who aren't doing anything like audio or video, it's very difficult to see how this traffic growth is going to slow down.

So it's almost like an avalanche of traffic. But it's not necessarily an avalanche of traffic where we get more revenue for it.

So at Intel, we're figuring out what can you do in order to meet a traffic growth that's basically doubling every year with a service provider community for which at the moment the revenue growth seems to be flat. The only way that you can do it is if you drive down your CapEx and your OpEx exactly the same rate that your traffic growth is doing.

Interestingly enough, what's almost a parallel to the growth of traffic is Moore's Law. If we can apply Moore's Law in its entirety to the problems of the network, then we could probably make that begin to add up. We would also get some revenue growth as a consequence.

Cost is obviously the biggest issue facing the industry right now. People are not installing new networks unless they can see return on investment within very short periods of time, six months or nine months. Everybody at the moment is CapEx and OpEx focused and that's absolutely inevitable in a recession. It's inevitable that's the way people will behave.

The way we look at it, again, we're down at one far end of the food chain, which is at the kind of components, manufacturing end. We're at the photonics, electronics, CMOS, digital end of the food chain.

The way we look at it is, let's try and use advanced manufacturing technology to try and drive that cost down.

A month ago, we announced the world's breakthrough semiconductor technology just going into production. That's 90 nanometer CMOS. It's a continuation of the existing technology that we've been on now for the last five years. But it takes it through to another level. Now, in addition to that, we're also looking at ways of using high integration, integration of components, to drive down the optical interconnection cost.

Using the concept of modularity, going towards modules in network assembly, what can we do to reduce the cost, increase the reliability?

What I'd like to do now is to show you an interview with Dr. Marc Verdiell. Many of you know Marc. He has been working for a period of time on using advanced manufacturing technology to reduce the cost of building transponders.

(Video plays and ends.)

SEAN MALONEY: One of the things that we've been working on as an industry for several years is in the tunable laser area. Marc described a standard way of assembling and significantly cost reducing by reducing the length of time it takes to assemble the modules.

What I have over here is an up to date demonstration of something that we announced last week. What you have here is a 300 pin MSA transponder package. But inside of it, we have a tunable laser. On the right hand side here, we have the control screen that I'm going to use to change the frequency of that laser. On the left hand side, we have a display from the optical spectrum analyzer so you can see the spectrum there.

I'm going to change from channel 1 down through to channel 8. This is an early version of the product. This is going to take about one minute to retune the laser on the fly, a CW tunable laser with a Mach-Zender modulator inside.

The advantage of this technology is that you get flexibility for the service provider in terms of inventory levels out in the field. If we can tune the laser on the fly when it's out in the field, that means we don't have to have so many line items in inventory, it means that the configuration becomes simpler, it means you don't have to have so many spares, the complexity of managing the whole thing comes down.

Tunable laser is something we've been working on inside of Intel for a considerable period of time. We recently did the acquisition of some technologies from a company called New Focus as well. You're going to see this, the tunable laser inside the transponder, coming in as a standard feature very soon.

Now, I guess some of you will be saying, "Well, is all of this technology necessary? Are there applications that are going to drive it in addition to things like 802.11?"

We believe passionately, yes, there are applications out there. There are needs the end customers ultimately have that can't be met now because they don't have the bandwidth.

Now, to illustrate that in a personal way, I'm going to call up Dr. Bruce Silverstein, who works at the medical school in Washington. Bruce and I were having dinner recently, and he ran through an example of a bandwidth problem. Hi, Dr. Bruce. Good to see you.

(DEMO begins and ends.)

SEAN MALONEY: But they were basically right. We have to go fix it, right? We have to get that bandwidth cost way down and reliability up. And we have to get very, very high speed links so that medical professionals like Bruce can use the latest technology to help people out.

It's not just in hospitals. Also we need to get the cost of optical technologies way down inside of the enterprise. Something that we have to do, and even if we don't have to do it, we have to do it anyways, and the reason why we have to do it is because one gig is happening to the desktop. One gig is a copper technology, obviously. One gig Ethernet is about to happen big time, we're going to have a crossover.

Now, the last time there was a step function increase in the client ability to generate traffic was in 1997. The original Ethernet specification, 1982, you remember Digital, Intel, and Xerox. In 1997, the industry crossed over from Ethernet to Fast Ethernet. Happened in a period of about a year.

The same thing's about to happen over the next 12 months. We're going to cross over into Gigabit Ethernet.

So the client, the PC, over the next 12 months, is going to have the ability to chuck traffic out at 10 times the rate that it's been doing previously.

Now, when you have all those one gig links together, the way you aggregate them is through 10 gig. And that's the industry problem, because 10 gig isn't going to run over copper. Now, maybe we can get 10 gig to run over copper the length of this stage. I don't think we're there yet, but maybe soon.

Ten gig sure as heck isn't going to run over copper anything over 300 feet or probably anywhere near 300 feet, which is the Ethernet floor specification.

So that means we have to carry on driving down this modularity, driving down the cost, using technologies like the XPAK module that Intel announced a few weeks ago.

This is a 10 gig module, so it has similar bandwidth to that 300 pin MSA, that tunable MSA that I showed you up there. But it's in a much, much, smaller form factor. It's small enough to fit onto a PCI card and obviously takes low power and can support fiber channel and so forth.

So we're on a path to drive down the cost of these 10 gig connections down to, first of all, one order of magnitude cheaper given that the bill of materials for a 10 gig link. We need to get it down to the point where it can go down on the server, the motherboard, or connecting up the storage subsystem. And that has to be less than US$100.

The way to get there is through integration. And this brings me to the next stage of the integration message, which is, today, I'm delighted to say that we're announcing another breakthrough on the manufacturing front. Earlier on, I said we had announced the 90 nanometer digital process. Today we're announcing the industry's first true mixed signal manufacturing process.

If I could take you back to 1968 and the formation of Intel, the work we did in the first couple of years was the development of the DRAM and then the microprocessor in 1971. All of the manufacturing technologies from that point to here have all been pretty much purely in the digital domain.

And when we have done analog, it's done in a totally different factory, totally different process.

So today we're announcing that our new manufacturing technology for the first time is going to combine both the analog world and the digital world in a single technology.

What's been the way of doing this up until now is that if you designed analog circuitry, you've tended to use a material like silicon germanium or maybe indium phosphide, and you've done it in a particular way with a different set of design rules, different design teams over on one side, and then on another side of your building, you've had digital folks who have been doing traditional digital design.

This has severely limited the ability of the industry to integrate together, and it has essentially meant that the communications industry hasn't really been able to ride down Moore's Law because you couldn't get all the passives in systems like RF systems or the various discrete components in opto, you haven't been able to get them into single packages so you can carry on shrinking.

So this technology is designed to integrate those two worlds together so that complex real communication systems like opto systems or RF systems can truly go on to Moore's Law and they can truly get the benefits of mass manufacturing.

The process itself has extremely high digital performance, as you'd expect. We have a fully functional 90 nanometer SRAMs right now. They're on these big, fat, 300 millimeter wafers.

We're just going through the conversion to these extra large wafers. These 12 inch wafers. They're built on factories that cost a little bit over US$2 billion to build. And this wafer here is one of these first wafers that combine digital and analog signals circuitry inside the same. Actually, if I move this thing backwards and forward, you can see the die there. You get a phenomenal yield increase in terms of productivity when you go from eight inch wafers over to these 12 inch wafers. We're planning on using this for some very good communications and optical components.

It uses something called strained silicon, which you may have read about in the newspapers over the last week or so.

The second thing that you can do with this is you can reduce power consumption pretty substantially. If you go along and look at a big voice switch or if you go and look at a server farm that's used for Internet data storage, or if you look at a bunch of gateways for doing Voice over Packet, one of the biggest problems is power consumption.

As you rack these things up more and more, the operating cost is significantly impacted by the amount of power these things absorb.

So, when we go back out to our customers' customers, one of the things we want to be able to do is say we're going to significantly reduce the power consumption of these products and it'll help get down your OpEx.

So we have some clever stuff in there for power consumption reduction. Then also for mixed signal, there is some special technology in there so that we can put passives like capacitors, inductors, and resistors into a single integrated chip that previously had been done in discrete components.

So if you pop the lid off a cell phone or if you pop the lid off a transponder like this, you see a whole bunch of little discrete components that haven't been able to be integrated because they've been on different processes. Now we're going to be able to put them into a single chip.

The circuitry itself is using silicon germanium, which is particularly useful for very high speed circuitry in both the optical space and in areas like RF and 802.11.

We're down one end of the food chain to many of you, supplying rural components. We believe this new manufacturing technology is going to make a substantial improvement. It's going to enable us to do higher integration, reduce the cost, shrink down the package size, it's going to enable the circuits to go much faster so we're going to be able to do single chip OC 192 network processors, things that can pull those packets off the network without dropping any packets and inspect them.

Serialization is also something that kind of comes for free with this technology. What that means is, instead of having to use a bunch of lines to send signals, because you can do super high speed I/O, you can go down to serial lines. Serial lines means you can reduce your pin count. Those of you who are in manufacturing know that apart from anything else, reliability is a function of the number of pins you have in a product. So you can go for lower pin count, higher density, higher reliability.

Digitalization is something you get when you go towards very, very high speed switching digital circuits, along the lines you get at 90 nanometers, you can actually start to do a bunch of things in digital that you used to have to use analog circuitry for.

So right now there's a lot of excitement over this technology, excitement in areas like cost reduced 10 gig OC 192, super small packaging, very high performance FECs and so forth. Over the next couple of years, a whole series of products are going to be coming out in that area.

One of the ways of looking at it is, from an optical space, you can look at something like an optical module or what some folks in the industry now are calling optical modems. And you can pop the lid off it. Right now, today, it takes about 10 chips and it has four different process technologies in it. Over a period of a couple of years as we roll this stuff out, you're going to go towards just three chips and two process technologies. Much simpler manufacturing process, so you can get into high volumes, reduce the cost, increase the reliability, and so forth.

So, I want to wrap with that point. The summary is, we do feel very, very optimistic. The traffic continues to grow. Our customers' customers have still a burning desire to generate traffic, get on the Internet and experience it. We believe that there are some very interesting technologies coming along that may accelerate that, probably the most interesting of which is 802.11.

Cost reduction is essential: cost reduction flowing from manufacturing technologies, modularity, all of us agreeing to go towards industry standards, whether it's MSA packaging or industry standards on software interfaces.

And recessions end. We always make a mistake during a boom. We always think the boom's going to last forever. And then during a recession, we always make the mistake of thinking the recession is never going to end.

You know, this one will end, too. We have to have the confidence and the will to see this one through, to see it through by continuing to invent, by continuing to push ahead, and not giving up in our confidence.

As Mark Twain said, "The answer is all around us. We just have to see it."

There is tremendous desire for our technology collectively as an industry. And we believe strongly that the future is bright.

And with that, I wish all of you here at NFOEC a fantastic conference and look forward to working with all of you over the next couple of years on some of these new technologies. Thank you very much.

About Intel
Intel (NASDAQ: INTC), the world leader in silicon innovation, develops technologies, products and initiatives to continually advance how people work and live. Additional information about Intel is available at www.intel.com/pressroom and blogs.intel.com.

* Other names and brands may be claimed as the property of others.