Moore's Law at 40: Part 5
- 2005-May-13
These captions and transcript were generated by a computer and may contain errors. If there are significant errors that should be corrected, please let us know by emailing digital@sciencehistory.org.
Transcript
00:00:01 which we'll be taking after a break, which we'll take after the next two speakers.
00:00:06 But we want to stay very much on schedule and it falls to me to have the great honor of introducing
00:00:12 the most famous person in the room and the person that we're all here to celebrate.
00:00:19 Pretty much everybody here knows that he invented a rule of thumb that became a law,
00:00:25 that became the theme and really a prime mover in what could be argued the world's most important industry in our times.
00:00:35 He also, quite remarkably and very uncommonly, I will say, as a venture capitalist,
00:00:41 we invest in founders and chief technology officers quite a bit.
00:00:45 I've been involved in five venture funds.
00:00:48 Very seldom do you see a career quite this successful on both the academic and the business side.
00:00:54 As you will note, he was co-founder of Intel, rose from executive vice president to CEO,
00:01:00 president and very successful chairman of the board in arguably one of the world's most important corporations at Intel.
00:01:09 Very intriguingly, the Chemical Heritage Foundation, I think,
00:01:14 thought up a delightful series of requests for the great Gordon Moore as to what to speak about.
00:01:20 As you'll see, the title is The Untold Story of Moore's Law, Economics, Strategy and Chemistry.
00:01:26 He's been asked to speak about the underappreciated dimensions of the economics and strategy surrounding Moore's law,
00:01:34 the economics that have been driven by his prediction and the way Moore's law has served as an explicit strategy for an entire industry
00:01:44 and the way it has become something of a self-fulfilling prediction.
00:01:48 And then, of course, to take a look at the way that the chemical and materials challenges have been involved in the semiconductor industry.
00:01:56 Please join me in welcoming a great and legendary leader, the wonderful Gordon Moore.
00:02:02 Well, thank you. But I think after today, there aren't any untold stories about Moore's law.
00:02:19 In fact, maybe I'm not turned on here. Is that better now?
00:02:24 OK. I think we may have gotten our slides all out of the same box or at least the data that went into them.
00:02:35 I'm going to show you several slides that you've seen before, either last night or today,
00:02:42 but ones that I think are important in getting us going here.
00:02:49 First of all, Moore's law is really about economics.
00:02:56 Boy, my glasses are getting too weak.
00:03:00 The industry is certainly best understood by looking at some of the underlying economics.
00:03:05 And let's take a look at the industry overall.
00:03:09 This is the semiconductor revenue versus time.
00:03:14 As you see, it's been a pretty good growth industry.
00:03:17 It's grown something like 100 fold during the time Intel has been in existence.
00:03:23 I think most of you would like to be in a market that grew 100 fold, especially with numbers with all these zeros on the end.
00:03:32 But this really underestimates the true rate of growth the way I look at it,
00:03:39 which is the products we turned out, counting the number of transistors.
00:03:43 This is something I started doing several years ago,
00:03:47 trying to look at the production of all the various devices,
00:03:51 estimating the number of transistors in them and plotting it versus time.
00:03:59 And the period in here, we were actually more than doubling every year.
00:04:04 Every year we made more electronics than existed in the world on January 1st of that year.
00:04:10 It slowed down somewhat recently, but it's still a pretty good growth curve.
00:04:14 Notice there are no bumps and wiggles in this curve like there were in the previous ones.
00:04:21 This one here has just continued to expand.
00:04:24 Now something over 10 to the 18th transistors a year.
00:04:30 That's a number that's kind of hard to contemplate.
00:04:32 I think Pat this morning said it was equal to the number of grains of rice every year.
00:04:37 Over the history, I've used a variety of different things.
00:04:40 At one time, Ed Wilton, the ecologist at Harvard,
00:04:48 estimated there were maybe 10 to the 16th ants on Earth.
00:04:52 So we'll pass 10 to the 16th transistor for every ant.
00:04:56 Now the poor little ant has to carry 100 of them around if he's going to have his share.
00:05:00 I estimate that the total number of printed characters in the world every year,
00:05:06 all the newspapers, magazines, Xerox copies, computer printouts, and the like,
00:05:12 is between 10 to the 17th and 10 to the 18th.
00:05:15 So we make more transistors than there are printed characters.
00:05:19 We sell them for less.
00:05:22 That's really the amazing factor.
00:05:24 You can take these two numbers, the revenue and the number of transistors,
00:05:30 divide one by the other, and get the average cost per transistor.
00:05:36 This has dropped down to about 100 nanobucks for the average transistor.
00:05:42 If you're talking DRAMs, it's somewhat less than that.
00:05:44 Microprocessors, probably a bit more.
00:05:47 But it's an amazing rate of change in the industry.
00:05:51 Of course, this has given us the large impact we've had
00:05:57 in making electronics generally much more available.
00:06:03 You've all seen a picture of the first planar integrated circuit two or three times today.
00:06:09 I've always been unhappy that this ugly beast is the only picture of it that got preserved.
00:06:19 But at the time I was doing the paper that became Moore's Law,
00:06:27 I wasn't the only one doing projections.
00:06:30 There was a conference in New York that same year, actually.
00:06:35 I think probably in March.
00:06:38 I forget if it was the IEEE then or if it was still the old IRE,
00:06:42 but they got the heads of the leading semiconductor companies together,
00:06:46 Texas Instruments, Motorola, Westinghouse, and Fairchild.
00:06:51 From Fairchild, the representative was Bob Noyce,
00:06:54 and he didn't say anything that I've extracted,
00:06:58 but the rest of them each made a prediction.
00:07:01 Pat Haggard, he said, he thought the industry,
00:07:04 looking out the next 10 years, would use 750 million gates a year.
00:07:10 A huge number.
00:07:12 We thought, boy, that's perceptive.
00:07:14 How did he get to something like that?
00:07:17 Harry Knowles, who was kind of the wild man among the group,
00:07:21 said we're going to get 250,000 gates from a single wafer.
00:07:26 We were struggling to get a few.
00:07:28 We thought this was ridiculous.
00:07:32 Les Hogan said the cost of a fully processed wafer will be $10.
00:07:38 Well, you can combine these things again,
00:07:40 and you can do the arithmetic,
00:07:43 and it says we'll only need 3,000 wafers a year.
00:07:47 At $10 a wafer, the total manufacturing cost
00:07:50 for all the semiconductors we need will be $30,000.
00:07:56 Somebody was wrong.
00:07:59 And it turned out the person that was most wrong was Haggerty,
00:08:05 the one we considered the most perceptive.
00:08:08 He so far underestimated the magnitude
00:08:11 of the number of gates that would be used.
00:08:13 It was ridiculous.
00:08:15 We actually achieved what Harry Knowles suggested
00:08:19 that we thought was ridiculous.
00:08:21 And even Les Hogan's $10-per-process wafer,
00:08:24 if you do it on a per-square-centimeter basis,
00:08:28 was pretty close, if you allow for inflation in particular.
00:08:31 We don't do $10 wafers, but we do wafers that are very much larger
00:08:36 from those 3-quarter-inch wafers,
00:08:38 1-inch wafers we're looking at at that time.
00:08:41 So it was really in the ballpark.
00:08:45 So this was the general view of the industry
00:08:48 at the time I did my 1965 projection.
00:08:52 I was given the task by Electronics Magazine
00:08:55 of trying to predict what would happen in the next 10 years
00:08:59 in the semiconductor components industry.
00:09:04 This was pretty early in the semiconductor integrated circuits era.
00:09:11 The situation in 1965 was
00:09:13 integrated circuits were used primarily by the military.
00:09:16 They were too expensive to be used in commercial systems.
00:09:20 The integrated circuits cost significantly more
00:09:22 than the individual components to put them together.
00:09:25 And the customers had all kinds of objections against them.
00:09:32 You couldn't be sure of their reliability
00:09:36 because you couldn't get in there and measure the parameters
00:09:38 of each of the components.
00:09:40 You couldn't measure the transistors and the resistors
00:09:42 and one thing and another separately.
00:09:45 There were cost arguments, and indeed the cost bore them out.
00:09:50 People argued that our yields would vanish.
00:09:53 They knew that transistors were made with yields
00:09:55 in the 10% to 20% range in that time.
00:09:59 A circuit with 8 transistors in it,
00:10:01 if you took 0.2 to the 8th power,
00:10:04 you got a darn small number.
00:10:07 So clearly we wouldn't be able to make very many.
00:10:11 The performance was below that
00:10:13 that you'd get using individual components
00:10:15 because of all of the parasitics and things that you had to do.
00:10:20 And of course, the yields wouldn't allow you
00:10:25 to get the supply you needed in any case.
00:10:29 So from my perspective,
00:10:33 looking as the laboratory director at Fairchild,
00:10:37 I could see some of the things that were coming down the pike,
00:10:40 and I wanted to get the message across
00:10:43 that integrated circuits were really going to be the route
00:10:46 to significantly cheaper electronics looking forward.
00:10:48 That was really the principal message I was after.
00:10:51 And I started out by looking at the cost per component
00:10:56 versus complexity and projecting it in a series of curves,
00:11:01 suggesting that there was a minimum cost at some complexity.
00:11:05 Less than that, you weren't taking full advantage
00:11:08 of the processing technology.
00:11:10 If you got beyond that minimum,
00:11:12 the die got so big that yields did drop considerably.
00:11:15 But this was something that was coming down pretty fast.
00:11:19 And from this, I took the original few points
00:11:24 and plotted this curve extrapolating
00:11:30 for the 10 years that I was asked to predict.
00:11:33 Now, the last point on this curve, the 1965 point,
00:11:36 was something we had in the laboratory
00:11:38 we were going to introduce.
00:11:40 So not to detract from people looking at this
00:11:43 and saying that's a reasonable extrapolation,
00:11:45 I used the rather obscure log to the base 2 scale
00:11:50 on the vertical element.
00:11:53 The fact that I was extrapolating from 60 to 60,000 roughly
00:11:58 wasn't quite so easy to get,
00:12:00 but it did seem to make sense with the existing data.
00:12:03 The first point down here was the original planar transistor
00:12:07 that you saw pictured earlier,
00:12:09 which was kind of the basic technology.
00:12:11 So it deserved to be on this same curve.
00:12:14 Now, I never really expected this to be very precise,
00:12:19 but as we...
00:12:24 newer slide...
00:12:26 threw on the points,
00:12:28 they actually scattered pretty well along that.
00:12:31 But at the end of 10 years,
00:12:34 I gave a paper at one of the device conferences
00:12:39 showing what had happened
00:12:42 and trying to analyze why we had made that degree of progress.
00:12:46 And I broke the curve into the principal contributors.
00:12:52 One of them was we were just making bigger dice
00:12:55 and we had more area to put them on there.
00:12:58 A contribution that was slightly larger
00:13:01 was due to the shrinking dimensions,
00:13:03 the increase in the density.
00:13:05 And if you multiplied those two together,
00:13:08 you get this curve here,
00:13:10 which is more than half of the progress we'd made,
00:13:14 but was still a very considerable element
00:13:17 that came about from some other factor,
00:13:20 which identified essentially squeezing waste space out of the chip.
00:13:24 We'd gotten rid of isolation structures,
00:13:26 a variety of other things.
00:13:28 And the last point on in 1975 was the CCD memory.
00:13:34 CCDs are active areas
00:13:38 as close to one another as you could get.
00:13:41 So there was no room left to squeeze.
00:13:43 So my argument was going forward from 19...
00:13:46 Whoops, didn't want to do that yet.
00:13:48 Going forward from 1975,
00:13:51 we were going to lose this factor.
00:13:53 And the slope, instead of being what it had been up to then,
00:13:56 was going to change to the slope of this line of these two products,
00:14:00 which is somewhat more than half,
00:14:02 but I just said half.
00:14:04 The slope is going to change from doubling every year
00:14:06 to doubling every two years.
00:14:08 But I knew too much,
00:14:10 because I saw these CCD memories,
00:14:13 and the one we had going into production was 32 kilobits.
00:14:18 We had a 64-kilobit about ready to go
00:14:20 and a 256-kilobit not too far behind it.
00:14:23 I could see those were going to keep us
00:14:25 on the doubling every year for another few years.
00:14:27 So I said, well, we're not going to change the slope right away.
00:14:30 I'd give it a five-year rollover time.
00:14:32 What I didn't realize is that CCD memories
00:14:35 were going to be a disaster.
00:14:38 The thing that makes them a good imaging device in your camera
00:14:41 makes them a terrible memory
00:14:43 because they're sensitive to any kind of radiation.
00:14:45 And in particular, an alpha particle out of the packaging material
00:14:49 will completely wipe out a bit or two or three.
00:14:53 And this was a major problem,
00:14:55 a soft error problem, non-repeatable errors
00:14:58 that we just started to find in DRAMs as well.
00:15:01 So CCDs prove very valuable for studying that phenomenon
00:15:04 and finding out what the problem was, again, with the subsolutions.
00:15:08 But we never introduced CCD memories beyond the first one.
00:15:13 The net result was I predicted this five-year hiatus here
00:15:18 before the slope changed,
00:15:20 but in fact the slope changed right away.
00:15:24 And as you see, had I started from the correct point,
00:15:30 I would have been much more accurate.
00:15:32 It turns out I've been historically.
00:15:38 I did a lot of other extrapolations at that time.
00:15:42 Some of them just to prove how ridiculous it was to extrapolate exponentials.
00:15:47 Again, you saw one of these plots earlier.
00:15:50 One of my colleagues dug the thing out,
00:15:53 and the 57-inch wafer that the extrapolation predicted for the year 2000
00:15:58 didn't quite come to pass.
00:16:00 But I'll have to admit I'm about equally surprised
00:16:03 by the 300-millimeter wafers that we use today.
00:16:08 Wafer size has certainly grown dramatically in this business.
00:16:13 This shows the size of the wafers we were using
00:16:17 when the first planar transistor came out
00:16:20 compared with the 300-millimeter wafer.
00:16:22 And I couldn't find one of the original wafers.
00:16:24 I would have sliced one off Harry's crystal
00:16:27 if I realized it was still around.
00:16:30 But we used 3.25-inch wafers in 1959.
00:16:34 In fact, one of my contributions to the industry at that time
00:16:39 was to show if you went above 3.25-inch,
00:16:43 your yields went to zero because of the quality of the material
00:16:46 deteriorated so rapidly.
00:16:49 The amount of technology that has gone into growing and slicing crystals,
00:16:56 single crystals now 300 millimeters in diameter,
00:17:01 is really fantastic.
00:17:04 I haven't seen a modern crystal grower.
00:17:06 I don't know how you do a 300-millimeter wafer.
00:17:09 Maybe I'll wait until we have the 450-millimeter ones to try.
00:17:14 It's been an amazing evolution of the technology.
00:17:18 Can you imagine our next wafer size is the size of the kind of pizzas
00:17:24 you get at Price Club, 18 inches.
00:17:26 Those are monster pizzas.
00:17:29 But in fact, this has become less and less a semiconductor technology
00:17:35 as a percentage of the total
00:17:37 and more and more an interconnection technology.
00:17:42 This is an electron micrograph of the copper interconnections,
00:17:49 actually a technology generation that's a couple of years old now.
00:17:55 These are copper interconnections where all the insulators
00:17:58 have been dissolved away, so you can see the complexity of the thing.
00:18:03 You can see the crystal grains in here.
00:18:06 As you go down further, we get smaller and smaller interconnections.
00:18:10 I like this slide because I think it shows it more dramatically,
00:18:13 but a more modern technology, the 90-nanometer one,
00:18:17 which is going into production now,
00:18:20 has these seven layers of metal interconnects,
00:18:24 low dielectric insulators here and there,
00:18:29 and buried down here at the bottom,
00:18:31 there's a thin layer of active silicon.
00:18:35 It really is an amazingly complex structure
00:18:39 that we've evolved.
00:18:41 And it's not just silicon.
00:18:43 If you look at the other things that are involved in here,
00:18:48 you see nickel silicide in some areas, silicon nitride.
00:18:54 This uses strained connectors,
00:18:57 excuse me, strained silicon in the transistors.
00:19:04 This one gets strained by putting some germanium into the silicon.
00:19:10 This one gets strained in the other direction
00:19:12 by using a silicon carbide nitride kind of a mixture.
00:19:17 For one of the devices, you want to compress the silicon.
00:19:20 For the other one, you want to expand the silicon,
00:19:22 depending on either the holes or the electrons,
00:19:26 the mobility you want to increase.
00:19:29 It really becomes really a pretty complex mess.
00:19:34 Things like tantalum nitride go in here as barriers
00:19:37 to prevent the copper from doing funny things.
00:19:40 The simple old silicon-silicon oxide aluminum
00:19:43 has really been pretty well replaced by much more complex systems.
00:19:52 Of course, one of the principal drivers here
00:19:56 of staying on the complexity curve
00:19:59 is the continual decrease in the dimensions.
00:20:04 The blue line here is actually what Intel did.
00:20:06 We got a little bit behind the industry there, I guess.
00:20:09 But this has been on a pretty constant slope,
00:20:13 the slope being a new generation about every three years,
00:20:17 and a new generation doubles the density.
00:20:20 That's been the algorithm the industry has had.
00:20:23 In the beginning, that just kind of happened,
00:20:27 and then increasingly we recognized this was what was happening
00:20:30 and tried to continue it.
00:20:32 When the first semiconductor industry,
00:20:35 this isn't the first, this is, I guess,
00:20:38 the second issue of the roadmap
00:20:42 where the Semiconductor Industry Association came out,
00:20:46 they plotted a continuation of this curve
00:20:49 with a generation every three years.
00:20:52 That was reasonable to do,
00:20:55 but the nature of this business is
00:20:58 you have to be at the leading edge to be competitive.
00:21:03 As Carver pointed out, if you are behind that,
00:21:07 you suffer a performance disadvantage,
00:21:10 and you also suffer a cost disadvantage
00:21:13 because what we really are doing is selling real estate.
00:21:17 The price of the real estate and the chip has been about constant
00:21:20 as long as I've been in the business.
00:21:23 It's been on the order of a billion dollars an acre.
00:21:26 It used to be a few billion.
00:21:29 I think microprocessors are now about $2 billion an acre.
00:21:32 Memory is maybe $0.8 billion, but the order of a billion dollars.
00:21:35 In fact, I used to claim that was why the Japanese
00:21:38 were such formidable competitors
00:21:41 because it was about the price of land in Tokyo in those days.
00:21:47 Anyhow, you stay near the leading edge
00:21:50 if you want to make leading edge products,
00:21:53 or you suffer both a performance and a cost disadvantage,
00:21:57 which is catastrophic competitively.
00:22:00 So what happens?
00:22:04 What happens, of course, is that we change the slope.
00:22:09 When everybody started to recognize
00:22:12 we were on a generation every three years,
00:22:15 we started changing to a generation every two years,
00:22:18 trying to get a bit ahead of the competition.
00:22:22 So rather than slowing up,
00:22:24 this trend to smaller and smaller dimensions
00:22:27 has actually accelerated
00:22:29 as a result of people understanding these trends.
00:22:32 I think this is an example of a case where
00:22:35 the so-called Moore's Law kind of plots
00:22:38 have really driven the industry.
00:22:40 Everybody recognizes they have to keep up with this curve
00:22:44 or they fall behind.
00:22:48 And, you know, this kind of thing is...
00:22:52 you can understand it intellectually,
00:22:55 but to understand it from a practical point of view,
00:22:58 I think this slide really shows it.
00:23:01 This is what the qualitative changes do over a period of time.
00:23:04 This was one contact in a generation of technology from 1978.
00:23:11 A few years later, more than a few,
00:23:14 a whole six-transistor memory cell
00:23:18 by the same scale sits in it like that.
00:23:21 These qualitative changes over a period of time
00:23:24 really make dramatically different products.
00:23:31 You know, a modern exposure system...
00:23:34 This is one of the 193-nanometer things
00:23:38 using an argon fluoride, exomer laser, and so forth.
00:23:42 It's really become a very sophisticated
00:23:46 and complicated piece of equipment.
00:23:49 But we'll have to make a step beyond that before too long.
00:23:52 And what we're working on is this 13-nanometer
00:23:56 extreme ultraviolet, essentially about a...
00:23:59 more than an order of magnitude decrease
00:24:02 in the wavelength that we're using.
00:24:05 And these machines have to have completely reflective optics.
00:24:10 Nothing is transparent, really, in this range.
00:24:15 Mirrors aren't very good reflectors either.
00:24:18 You go through 30-plus reflections in one of these systems,
00:24:21 each with only something like 0.6 or 0.7 reflectivity,
00:24:25 it becomes a real challenge.
00:24:28 The optical surfaces have to be better than the Hubble telescope
00:24:32 in order to get the kind of things we want.
00:24:35 It's really an interesting challenge.
00:24:38 And it keeps us on another exponential,
00:24:41 the cost of the lithography equipment.
00:24:44 In fact, you can plot an exponential like this
00:24:47 for most of the pieces of equipment.
00:24:50 It's an interesting economic challenge.
00:24:53 All the equipment keeps going up in cost exponentially.
00:24:56 The industry isn't growing especially rapidly anymore.
00:24:59 But you've got to stay on this leading edge of technology
00:25:02 or you are at a cost disadvantage.
00:25:09 So...
00:25:12 Oh, and also the other dimensions we keep pushing.
00:25:16 You know, Carver talked about the problems you get into
00:25:20 when you get to very thin layers.
00:25:24 This is one nanometer here.
00:25:26 So we're talking dielectrics
00:25:29 whose electrical properties start to deviate more and more
00:25:32 from the actual physical width.
00:25:35 I guess it takes some room for the electrons to locate themselves.
00:25:38 And to show just how dramatic this is,
00:25:42 these are transmission electron micrographs
00:25:47 of the gate area of some of these devices.
00:25:51 And that is a 90-nanometer process.
00:25:54 You can see the individual atoms in the silicon here.
00:25:57 The silicon oxide layer, as you see,
00:26:00 is a couple of molecular layers.
00:26:03 No more than that if we do that.
00:26:06 With the gate, which is, again, silicon on top.
00:26:09 And this has a problem.
00:26:12 It has leakage current that we'll call 1.
00:26:15 The other hand, if we can change materials
00:26:18 and go to something with a high dielectric constant,
00:26:21 which is something that allows this big a change we're working on now,
00:26:24 you can still get the capacitance.
00:26:27 In fact, you get a higher capacitance here,
00:26:30 which means higher electric field transferred to the substrate,
00:26:33 giving us more performance.
00:26:36 But with the leakage going down 100-fold.
00:26:39 And those are the kind of changes that new materials allow us to make.
00:26:42 They're not easy.
00:26:45 A tremendous amount of work went into finding a material
00:26:48 that had the correct dielectric capabilities
00:26:51 but also was stable enough
00:26:54 and could tolerate these high electric fields
00:26:57 to make this happen.
00:27:03 Well, one might ask,
00:27:06 when is it going to end?
00:27:09 In fact, I think I've been asked that at least 100 times this year.
00:27:12 And the answer is, not very soon.
00:27:16 I can see at least as far now as I ever could.
00:27:19 In fact, I always thought I could see maybe 2 or 3 generations
00:27:22 of what we were going to do.
00:27:25 I talked to the Intel R&D people
00:27:28 and they're looking out further than that now.
00:27:31 I can probably see what we're going to do for 4 generations,
00:27:34 which is further than I've ever been able to look before.
00:27:37 A lot of things have to happen to make that occur,
00:27:40 but they're confident that the things are going to be in place
00:27:43 to make it happen.
00:27:49 It's amazing, really, what a group of dedicated scientists
00:27:55 and engineers can do.
00:27:58 I don't see an end in sight.
00:28:01 With the caveat, I can only see a decade or so ahead.
00:28:06 Thank you.
00:28:14 Thank you so much.
00:28:20 I think we are going to try to hold questions for the panel.
00:28:23 Is that right?
00:28:26 We will hold questions, but please write them down.
00:28:29 This is a once-in-a-lifetime opportunity to get to ask Gordon more questions.
00:28:32 Let me just make sure.
00:28:41 We do have an unusual circumstance
00:28:44 in that Annalise Saxenian, I understand, is catching a plane
00:28:47 right after her remarks.
00:28:50 We're trying to be considerate of the wheels-up session here.
00:28:53 One way of thinking about this is, having thought about the chemistry
00:28:56 and really an industrial revolution in and of itself,
00:28:59 led by Gordon Moore,
00:29:02 there is the question about the broader context of community.
00:29:05 Next to speak about that, we get to hear, as you will read,
00:29:08 from the Dean of Information Management and Systems
00:29:11 out at Berkeley
00:29:14 and a professor who is also a professor
00:29:17 in the Department of City and Regional Planning,
00:29:20 which is germane for today's discussion.
00:29:23 Professor Saxenian is an internationally recognized expert
00:29:26 in economic development and information technology.
00:29:29 She has published extensively on the social and economic organization
00:29:32 of production in technology regions.
00:29:35 Interestingly, writing lately about how immigrant engineers
00:29:38 and scientists are transferring
00:29:41 technology entrepreneurship to East Asia.
00:29:44 Many in conversations around here have been talking about
00:29:47 how they expect so much to flow now from East Asia.
00:29:50 Annalise Saxenian has been asked,
00:29:53 and again I congratulate the Chemical Heritage Foundation
00:29:56 for its thoughtful requests as to topics,
00:29:59 to discuss the transformation of San Francisco's Bay Area
00:30:02 in particular, on which she has written extensively,
00:30:05 in connection with the development of the semiconductor
00:30:08 and associated industries.
00:30:11 She will be discussing the developmental dynamic
00:30:14 between the industry and associated industries
00:30:17 and the region and community.
00:30:20 Please join me in welcoming Dean Annalise Saxenian.
00:30:23 Applause
00:30:26 Applause
00:30:29 Applause
00:30:32 Applause
00:30:35 Can you hear me?
00:30:38 Great. Perfect.
00:30:41 Well, Patrick said this morning that he fell out of place
00:30:44 because he wasn't a chemist.
00:30:47 I feel a little bit out of place here,
00:30:50 but I have a different agenda,
00:30:53 so hopefully that will save me.
00:30:56 I guess I need to say that the first time
00:30:59 that I met Gordon Moore was about a decade ago.
00:31:02 I had just published a book comparing Silicon Valley
00:31:05 and Boston's Route 128 area.
00:31:08 It was really based on my dissertation,
00:31:11 and I had been asked to give a talk at Stanford
00:31:14 by a group at the Economic Policy Research Group there.
00:31:17 It was a group of industry leaders,
00:31:20 including Gordon.
00:31:23 I had only given this talk to people at MIT
00:31:26 or on my dissertation committee,
00:31:29 and I was never more terrified giving that talk
00:31:32 about Silicon Valley to Silicon Valley.
00:31:35 I'm going to tell a story that I told then,
00:31:38 and I suspect that Gordon is the only one
00:31:41 that's heard this story before, I hope.
00:31:44 It's just so apropos of this story about people
00:31:47 continuing to see the limits of miniaturization,
00:31:50 and yet they keep receding in front of us.
00:31:53 Well, I was one of the first people to study Silicon Valley.
00:31:56 I was a master's student at Berkeley in the 1970s,
00:31:59 and I decided to write a master's thesis
00:32:02 about Santa Clara Valley, really, it was called then.
00:32:05 It was the fastest growing city in the country at the time.
00:32:09 I wanted to understand it better,
00:32:12 and I went around and interviewed a lot of industry people.
00:32:15 In fact, I interviewed Bob Noyce a couple of times.
00:32:18 I never did meet Gordon at that point.
00:32:21 But I concluded this thing by writing a master's thesis,
00:32:24 which confidently predicted that Silicon Valley would stop growing.
00:32:27 I argued that it was too congested.
00:32:30 This was in 1970, I guess I published it in 1980.
00:32:33 It was too congested, the housing costs were too high,
00:32:36 the traffic was too congested, the environment.
00:32:39 And essentially what I argued, which is really what the people
00:32:42 that I interviewed had been telling me was,
00:32:45 that the semiconductor industry, while headquarters might remain in the valley,
00:32:48 all the growth would occur in Austin, Texas, or in Seattle,
00:32:51 other parts of the country where costs were lower.
00:32:54 And this is the point when they were moving fabs out
00:32:57 to different parts of the country.
00:33:00 Of course I was completely wrong, and then I had to go and write a dissertation
00:33:03 about why I had been so wrong,
00:33:06 and why it is that Silicon Valley continued to flourish
00:33:09 really for several more decades.
00:33:12 And I'm going to come back to that story.
00:33:15 What I want to do first is back up and give a bit of an introduction
00:33:18 to what I want to say.
00:33:21 Essentially the big picture is that
00:33:24 not only are integrated circuits
00:33:27 technologically revolutionary,
00:33:30 but they are transforming our entire industrial apparatus,
00:33:33 our organizational models,
00:33:36 and our geography in ways that we don't
00:33:39 really even fully understand now.
00:33:42 There was a Russian economist, Nikolai Kondratiev,
00:33:45 who wrote about these 50 year long wave cycles,
00:33:48 that essentially when you have
00:33:51 general purpose technologies like that,
00:33:54 they not only affect their immediate vicinity,
00:33:57 but transform the entire
00:34:00 global economy.
00:34:03 I think the IC industry, for the reasons that Gordon just mentioned,
00:34:06 because of the dramatically falling cost, is one of those industries
00:34:09 that is now transforming how we organize work,
00:34:12 and how we organize companies, and the geography.
00:34:15 I'm not going to touch on all of those,
00:34:18 I'm just going to really talk about the Silicon Valley region,
00:34:21 and then the geographic elements,
00:34:24 and how that's being transported around the world.
00:34:27 So this is an overview
00:34:30 of what I'm going to be talking about.
00:34:33 First I'll talk a bit about the Silicon Valley model,
00:34:36 and regional growth.
00:34:39 Then I want to talk about the ICs, not the integrated circuits,
00:34:42 but the Indians and Chinese that have really
00:34:45 played a huge role in Silicon Valley over the past couple of decades.
00:34:48 Then I want to talk about the shift from brain drain
00:34:51 to brain circulation, the creation of a partner
00:34:54 for Silicon Valley, really a manufacturing extension
00:34:57 of Silicon Valley in Taiwan.
00:35:00 And finally, talk a bit more speculatively
00:35:03 about China and India, and the role that they'll play
00:35:06 in this rapidly expanding dynamic industry.
00:35:13 Let me just say a few words about Silicon Valley's
00:35:16 technological evolution.
00:35:19 The integrated circuit itself got more and more powerful,
00:35:22 but it's fueled different waves of
00:35:25 new technology, new products.
00:35:31 There's Moore's Law in 1965,
00:35:34 the integrated circuit wave.
00:35:37 Each time you have a wave of innovation,
00:35:40 then you have a bit of recession in the valley,
00:35:43 and then a new technology.
00:35:47 The important point is that the value added
00:35:50 keeps rising as we move historically,
00:35:53 and this corresponds to the rising cost of living
00:35:56 in the Bay Area as well.
00:35:59 Then finally, the internet boom of the last decade.
00:36:02 We're sitting now on the tail end of one boom,
00:36:05 and awaiting the next wave,
00:36:08 where it will be, we don't know.
00:36:11 I want to explain what I concluded
00:36:14 about Silicon Valley, and why it continued to grow
00:36:17 in spite of all my predictions that it would stop growing.
00:36:20 Essentially, what I learned is that
00:36:23 it's a system,
00:36:26 this is many elements on the board,
00:36:29 but let me just say the three critical elements
00:36:32 to the Silicon Valley system that distinguish it
00:36:35 from older industrial regions.
00:36:38 What first was just a tremendous entrepreneurship,
00:36:41 and the entrepreneurship, we've heard about it already.
00:36:44 We had Shockley, the traitorous eight left,
00:36:47 Fairchild, Intel, just the flourishing,
00:36:50 the spawning of Fairchildren and Grandchildren,
00:36:53 the sort of genealogical story of the valley
00:36:56 where entrepreneurs, engineers really felt free to leave
00:36:59 when they found some technology opportunity
00:37:02 and start a new business.
00:37:05 The first piece is entrepreneurship.
00:37:08 Entrepreneurship is the implementation of the industry,
00:37:11 fueled by venture capital, by the way,
00:37:14 which has been mentioned a bit here.
00:37:17 One of the things that Intel did
00:37:20 was they forbid the manufacturing
00:37:23 or internal production of semiconductor equipment.
00:37:26 Why was that so important?
00:37:29 It turned out that Intel decided they would specialize
00:37:32 on the devices, manufacturing the ICs,
00:37:35 and started a train of specialization within the valley
00:37:38 where firms, as they spun off,
00:37:41 focused on a very narrow area and then relied on
00:37:44 the infrastructure in the region
00:37:47 rather than trying to build it all themselves.
00:37:50 That specialization has turned out to be, I think,
00:37:53 extremely important to the way the industry
00:37:56 has spread geographically.
00:37:59 Then finally, a pattern of open information exchange,
00:38:02 which I think was unprecedented in U.S. industry at the time.
00:38:05 I like to call it a community,
00:38:08 and this is a technical community that we've seen here.
00:38:11 It's Gordon and Carver and Harry
00:38:14 and a bunch of other people that we've been talking about
00:38:17 that formed the nucleus for a community
00:38:20 that built this industry, that pushed this technology forward.
00:38:23 In my book, Regional Advantage,
00:38:26 I compare the system that emerged in the valley
00:38:29 from entrepreneurship, specialization,
00:38:32 and information exchange to the Boston...
00:38:35 Oh, here's the entrepreneurship, I'm sorry.
00:38:38 This is just the entrepreneurial churn in Silicon Valley.
00:38:41 You can see thousands of new companies
00:38:44 created yearly in Silicon Valley.
00:38:47 The orange on the top is firm openings.
00:38:50 On the bottom is firm closings.
00:38:53 A lot of firms die, but you learn from failure.
00:38:56 We've heard here already,
00:38:59 a few companies move in and out all the time,
00:39:02 and then the line traces the net firm churn.
00:39:05 Even after the recession,
00:39:08 even after the stock market in 2001,
00:39:11 you still have 10,000, 15,000 companies
00:39:14 being started each year in the valley,
00:39:17 net new companies.
00:39:20 Entrepreneurial churn, unprecedented, I believe,
00:39:23 actually in the world still.
00:39:26 Specialization, this is the second thing I mentioned.
00:39:29 When I compared Silicon Valley to Boston's Route 128,
00:39:32 I came down pretty hard on DEC
00:39:35 because Digital Equipment Corporation,
00:39:38 which had been a leader in the mini-computer industry,
00:39:41 got completely bypassed by Silicon Valley's PC industry.
00:39:44 The argument that I made was that
00:39:47 the DEC model, vertically integrated,
00:39:50 they tried to do their equipment internally,
00:39:53 their architecture, their microprocessors,
00:39:56 their operating systems, all of the components,
00:39:59 they tried to do within one big company,
00:40:02 whereas in Silicon Valley, each of these layers
00:40:05 was either one single firm or often
00:40:08 hundreds of firms competing intensely,
00:40:11 each of them pushing technology forward.
00:40:14 So the specialization brings not only
00:40:17 technology forward more rapidly.
00:40:20 The problem in the DEC model was that
00:40:23 it was very difficult to support as many
00:40:26 different competing experiments as you found
00:40:29 within Silicon Valley in each of these different technologies.
00:40:32 So it became constrained.
00:40:35 It was hard to push technology forward.
00:40:38 They often got locked into sort of old technologies
00:40:41 that they had developed internally
00:40:45 Okay, so that's really the story
00:40:48 of the Silicon Valley model.
00:40:51 This is just to reassure you that Silicon Valley
00:40:54 is still alive and well.
00:40:57 We had a big blip in the late 90s.
00:41:00 Venture capital spending went off the charts.
00:41:03 I think it was healthy that we had the bust.
00:41:06 People were starting firms who never
00:41:09 should have started firms.
00:41:13 It's continuing to rise and there's a sense
00:41:16 in the Valley now that there are many new opportunities
00:41:19 emerging in many different places.
00:41:22 There's more optimism this year than there was
00:41:25 last year and more than the year before.
00:41:28 Of course, the big change is that that innovation
00:41:31 is not just happening in Silicon Valley anymore.
00:41:34 I think for the first 20 or 30 years,
00:41:37 you could easily say, and in fact I wrote a book
00:41:40 about Silicon Valley, really within itself.
00:41:43 Really, you could study all of the development.
00:41:46 Of course, in the 60s, semiconductor assembly
00:41:49 was moved off to Malaysia and Singapore.
00:41:52 There were some fabs set up in other parts of the US.
00:41:55 As we heard this morning, there were some alliances
00:41:58 in Europe.
00:42:01 Their capabilities were building elsewhere,
00:42:04 but really the action was still in the Bay Area largely,
00:42:07 and we certainly can't say that today.
00:42:10 The story that I want, here's a little more
00:42:13 about productivity growth.
00:42:16 This is just data on the health of Silicon Valley.
00:42:19 Each new wave of technology has brought with it
00:42:22 higher and higher value added.
00:42:25 Even though jobs have not grown,
00:42:28 if you normalize to jobs,
00:42:31 essentially you see that productivity has continued
00:42:34 to grow dramatically.
00:42:37 That's another measure of the impact of this industry,
00:42:40 this dramatic growth in value per worker hour.
00:42:43 If you want to look at it, it's compared to the US as a whole.
00:42:46 Look at that.
00:42:49 Even through the recession,
00:42:52 we're now almost double the US value added per employee.
00:42:55 This is, by the way, in the high-tech industries.
00:42:58 This isn't in the economy as a whole.
00:43:01 It's very entrepreneurial,
00:43:04 growing continually, but also now,
00:43:07 I would argue, growing in many other regions at the same time.
00:43:10 For this, I need to go back and talk
00:43:13 just quickly about the brain drain.
00:43:16 I think all of us in the room probably know the brain drain,
00:43:19 but just a small bit of data.
00:43:22 This shows the percentage of
00:43:25 foreign-born scientists and engineers working in the US.
00:43:29 You see that as you go towards,
00:43:32 first of all, it's grown historically between 1990 and 2000,
00:43:35 but also as you go up the degree levels,
00:43:38 the higher and higher educated,
00:43:41 the presence of immigrants is higher and higher as well,
00:43:44 so that in 2000,
00:43:47 7-38% of all scientists and engineers
00:43:50 working in the US were foreign-born.
00:43:53 Those of you that teach
00:43:57 know that the university engineering chemistry departments
00:44:00 reflect that trend directly.
00:44:03 I think this is a bit of what Patrick was worrying about this morning,
00:44:06 that we are not training a new generation
00:44:09 of scientists and technologists.
00:44:12 This is another measure,
00:44:15 just to show you the geographic distribution.
00:44:18 It turns out that the overwhelming majority of the students
00:44:21 that are studying in the US, getting higher education in the US,
00:44:25 are from East and South Asia.
00:44:28 Then you have a smattering from Europe,
00:44:31 but that includes all the countries of Europe,
00:44:34 and then far smaller numbers from the rest of the world.
00:44:37 That large share from East and South Asia
00:44:40 is really what's driving the transformation of the global economy now,
00:44:43 I would say.
00:44:46 Here's my IC story.
00:44:49 This tracks immigrants to Silicon Valley.
00:44:53 One thing that has become apparent by looking closely at the data
00:44:56 is that Silicon Valley absorbed the lion's share
00:44:59 of the foreign-born engineers.
00:45:02 I'll show you a slide that demonstrates that.
00:45:05 When I've looked at other regions,
00:45:08 I've looked at Boston, Seattle, Austin,
00:45:11 none of them have the share.
00:45:14 They have maybe half or fewer scientist engineers.
00:45:17 Somehow Silicon Valley became a magnet
00:45:20 for a lot of people.
00:45:23 They didn't all come from Stanford and Berkeley.
00:45:26 They came from Illinois, Philadelphia, Florida, all over the country,
00:45:29 but in the 70s and 90s,
00:45:32 they were sucked into the Silicon Valley labor market disproportionately.
00:45:35 This simply shows the immigration
00:45:38 of professional and technical workers to the Bay Area.
00:45:41 You can see almost 40% from China,
00:45:44 20% from India,
00:45:47 20% from Iran, Russia, and so forth.
00:45:50 This shows the trend.
00:45:53 This is only interesting
00:45:56 because it shows between 1970 and 2000
00:45:59 the trend in immigration to the Valley.
00:46:02 In 1970, about 14% of Silicon Valley's workforce
00:46:05 was foreign-born.
00:46:08 If you look at the breakdown,
00:46:11 it was mainly European immigrants.
00:46:14 The balance shifted dramatically.
00:46:17 By 2000, foreign-born workers represent 53%
00:46:20 of the workforce,
00:46:23 of the science and engineering workforce in the Valley.
00:46:26 You see the European share
00:46:29 has fallen.
00:46:38 These are all students who came and studied,
00:46:41 got graduate degrees in the US,
00:46:44 got hired,
00:46:47 worked for companies.
00:46:50 They quickly learned the Silicon Valley model.
00:46:53 Some of Gordon's work
00:46:56 has talked about the shift
00:46:59 from being an engineer to being a manager.
00:47:02 They learned how to be managers, not just engineers.
00:47:05 They learned how to work outside of research labs.
00:47:09 By 2000, they were starting
00:47:12 about a third of all the technology companies in the Valley.
00:47:15 They were very successfully
00:47:18 learning this Silicon Valley model.
00:47:21 They were marinating in it.
00:47:24 There is some data
00:47:27 on the number of companies they started.
00:47:30 Now we are aware of some of the high-profile successes
00:47:33 like Hotmail or Yahoo.
00:47:37 We couldn't list them all,
00:47:40 but a tremendous number of very successful companies.
00:47:43 The important fact about this
00:47:46 is not simply that they learned the model of the Valley,
00:47:49 but that
00:47:52 their successes were recognized
00:47:55 in places like India and China.
00:47:58 It started to change the reputation
00:48:01 of those countries as places to do business.
00:48:05 In 1990,
00:48:08 very few people would have thought of sending business to India.
00:48:11 It would have been very hard to imagine.
00:48:14 It was seen as backward, unreliable, corrupt.
00:48:17 In fact, IBM left India
00:48:20 only in 1978
00:48:23 because it wasn't willing to take
00:48:26 the shareholders that were required by government
00:48:29 onto its board.
00:48:33 There was a period in India starting in 1978
00:48:36 where there were no foreign technology companies at all.
00:48:44 I'm just going to show you a bit of the data.
00:48:47 The story that I want to tell
00:48:50 is the story of how these immigrants
00:48:53 have started to go back home.
00:48:56 They first created communities, I would argue.
00:48:59 The Chinese tend to start companies with one another.
00:49:02 The Indians tend to start companies with one another.
00:49:05 Probably, understandably, if you're a first-generation immigrant,
00:49:08 you trust people that speak your language
00:49:11 that celebrate your holidays and whatnot.
00:49:14 Here's the evidence
00:49:17 that they learned how to be entrepreneurs.
00:49:20 More of the foreign-born engineers in the Valley,
00:49:23 62% said they planned to start a business
00:49:27 and this was in 2001,
00:49:30 which is not a good time to start a company.
00:49:33 Have you advised or arranged contracts
00:49:36 for companies in your home country?
00:49:39 Again, this is old data.
00:49:42 People may be lying,
00:49:45 but even if some of them are lying,
00:49:48 there's a huge number of people.
00:49:51 34% of the mainland Chinese
00:49:54 and 46% of the Indians.
00:49:57 Likewise, they've advised companies in their home country.
00:50:00 Often, the advising is through your social network.
00:50:03 You advise your old friends
00:50:06 who went to the Indian Institute of Technology
00:50:09 with you or National Taiwan University.
00:50:12 All of a sudden, by 2001,
00:50:15 we were seeing tremendous circulation of ideas
00:50:18 and also people, what I'm calling brain circulation,
00:50:21 have you invested your own money in startups?
00:50:24 I was surprised that the numbers came out
00:50:27 even as high as they did here.
00:50:30 More than once, only once,
00:50:33 10% of the mainland Chinese in 2001
00:50:36 were investing money in China.
00:50:39 Sure, some of these people were just angels.
00:50:42 They were funding their friends, their family members.
00:50:45 Some of them were also formal venture capitalists.
00:50:48 Some of them were capital funds,
00:50:51 specializing in doing business between Silicon Valley and India,
00:50:54 between Silicon Valley and Taiwan, Shanghai.
00:50:57 There's a lot of expertise that's built up very quickly
00:51:00 transferring the professional service knowledge
00:51:03 from the valley to these other countries.
00:51:06 Would you consider returning to live in your country
00:51:09 at birth in the future?
00:51:12 Quite likely, 14% to 25%.
00:51:16 But somewhat likely, close to 50%
00:51:19 for both the mainlanders and the Indian immigrants.
00:51:22 Now, this may seem commonplace.
00:51:25 It's gotten a lot of media attention lately,
00:51:28 people going back home.
00:51:31 But I assure you that 10 years ago, 15 years ago,
00:51:34 very few engineers wanted to go back home.
00:51:37 In fact, the data from the National Science Foundation
00:51:40 from 2002 still shows that 97% or 98%
00:51:43 of both Indians and Chinese plan to stay in the U.S.
00:51:46 They don't plan to go home.
00:51:49 Whether they ultimately do or not, we don't know yet.
00:51:52 But we're in a period of rapid transition in that regard.
00:52:05 One of the things that...
00:52:08 Remember, I talked about the specialization
00:52:11 within the industry.
00:52:14 So the semiconductor industry became specialized
00:52:17 into people who made the integrated circuits,
00:52:20 who did the equipment, and whatnot.
00:52:23 What does that do?
00:52:26 Well, it creates opportunities for people to split off
00:52:29 and become entrepreneurs in other regions
00:52:32 but link into the value chain here.
00:52:35 And what you saw in the 70s, 80s, and 90s
00:52:38 was returning to Taiwan.
00:52:41 In fact, this is rather...
00:52:44 This curve I find pretty impressive.
00:52:47 Very few people returned from the U.S. to Taiwan
00:52:50 until the mid-80s.
00:52:53 And then in the late 80s, what happened?
00:52:56 The stock market in Taiwan took off.
00:52:59 Acer went public.
00:53:02 Taiwan Semiconductor Manufacturing Corporation was established.
00:53:05 Pioneered the foundry model
00:53:08 where they only manufactured chips.
00:53:11 They didn't design them.
00:53:14 They manufactured them in other independent design firms
00:53:17 who could be anywhere in the world.
00:53:20 So the specialization allowed Taiwan to carve out
00:53:23 a particular niche for itself.
00:53:26 Separate from but complementary to what's going on
00:53:29 in Silicon Valley at the time.
00:53:32 As people started getting rich in Taiwan,
00:53:35 you saw a flood back home.
00:53:38 And that's the flood that you see in the late 80s and early 90s.
00:53:41 And the Hsinchu Science Park took off in this time.
00:53:44 And the IC industry in Taiwan
00:53:47 as well as the related personal computer industries
00:53:50 really took off through people going back and forth.
00:53:53 And I would say that that industry in Taiwan
00:53:56 is like an extension of Silicon Valley.
00:53:59 The manufacturing that was once done in Silicon Valley
00:54:02 is now being done in Taiwan.
00:54:05 There we go.
00:54:08 The IC foundry model.
00:54:11 The two of them now control 70% of the foundry business in the world.
00:54:14 They're not leading technological innovation.
00:54:17 That's still in the U.S.
00:54:20 But they're cranking out a whole lot of chips for a lot of products
00:54:23 that are serving different markets in the world
00:54:26 The specialization of the manufacturing infrastructure
00:54:29 is very advanced in Taiwan.
00:54:32 You have really sophisticated IC packaging,
00:54:35 lead frame makers, mask makers,
00:54:38 design, wafer manufacturing.
00:54:41 In each of these areas, you've seen developed companies
00:54:44 that are at the leading edge.
00:54:47 And a lot of this is through incremental process innovation
00:54:50 that was mentioned earlier.
00:54:53 Just really innovating in the process on the shop floor.
00:54:56 Taiwan now looks very much like Silicon Valley, interestingly.
00:54:59 It's not as advanced,
00:55:02 but this infrastructure of small firms, highly specialized,
00:55:05 lots of venture capital,
00:55:08 very innovative, very dynamic.
00:55:11 The other case that I just passed over
00:55:14 and I want to quickly mention
00:55:17 that's very parallel to Taiwan
00:55:20 where they took chip manufacturing
00:55:23 and they really captured PC manufacturing
00:55:26 for the world in the 80s and 90s.
00:55:29 Another example, an earlier example,
00:55:32 is Israelis.
00:55:35 I know that Intel set up an R&D lab in Israel
00:55:38 for a guy who wanted to go back home.
00:55:41 And I think that you saw in this period,
00:55:44 in the 70s, 80s,
00:55:47 a large community of Israelis who had worked in the U.S.
00:55:50 who wanted to go home.
00:55:53 Again, transferring the know-how and the technology.
00:55:56 In this case, contributing at a higher value level.
00:55:59 Things like Internet security, software, and whatnot.
00:56:02 You have these two small nodes
00:56:05 that are connected to Silicon Valley
00:56:08 but specialized differently and complementary.
00:56:11 And now we sit at the crossroads
00:56:14 and the ones that are about to come online are the big ones.
00:56:17 Taiwan, Israel, they had to serve our markets.
00:56:20 They really had to partner with us.
00:56:23 There was no question that they didn't have the domestic market
00:56:26 to support an industry of their own.
00:56:29 PC, this is just data on how the Taiwanese dominate markets.
00:56:32 Over 80% of wireless,
00:56:35 local area networks, 65% of notebook PCs.
00:56:38 That's the global market share.
00:56:41 It's not impressive.
00:56:44 They're moving into cameras and cell phones and whatnot, PDAs.
00:56:47 But of course, what's happened to Taiwan?
00:56:50 Essentially, the competitive pressure has been so intense
00:56:53 that most of Taiwanese manufacturing
00:56:56 has now been pushed across the Taiwan Straits into China.
00:56:59 And China is now the leading IT producer in the world.
00:57:02 Here we go.
00:57:05 Silicon, Shanghai.
00:57:08 You have two new fabs established,
00:57:11 SMIC and Grace Semiconductor,
00:57:14 both established in 2001 in Shanghai,
00:57:17 largely through the know-how and technology
00:57:20 transferred by Taiwanese managers, engineers,
00:57:23 as well as a community from Silicon Valley.
00:57:26 So now what was never thought possible,
00:57:29 nobody saw China on the horizon in fabs.
00:57:32 Nobody thought that they would.
00:57:35 We're several generations behind,
00:57:38 but you've seen this tremendous push across the Straits.
00:57:41 So we have companies that are manufacturing in Silicon.
00:57:44 The entire PC infrastructure from Taiwan
00:57:47 has now moved over
00:57:50 to the Shanghai and Dongguan,
00:57:53 the southern Pearl River Delta in China.
00:57:56 So we're in a new world.
00:57:59 Again, they're not competing directly with Silicon Valley,
00:58:03 but they're certainly lowering the cost of all these commodities
00:58:06 that are built on ICs very rapidly.
00:58:09 And this is just a bit of the geography.
00:58:12 In fact, these bubbles are a little outdated.
00:58:15 The bubble on Shanghai should be the biggest one now.
00:58:18 And these are really the areas,
00:58:21 the new, you know, if you want to call them Silicon Valleys,
00:58:24 I hate that, but the new districts
00:58:27 that are producing IC technology.
00:58:30 China's a huge market.
00:58:33 China is starting to use the technology
00:58:36 not to create the next generation of technology,
00:58:39 but to create products that serve the Chinese market,
00:58:42 a lower income market,
00:58:45 and ultimately probably to serve other developing economy markets.
00:58:48 Patrick this morning said we're going to get
00:58:51 to all 6 billion people in the world.
00:58:54 Well, the way to do that is to design with existing technology
00:58:57 and develop products that are more appropriate in those markets.
00:59:00 And I think we're going to see in this decade
00:59:03 many, many more products of that sort.
00:59:06 Bangalore.
00:59:09 Bangalore's not manufacturing chips,
00:59:12 but they are designing them.
00:59:15 And there's some very capable chip designers.
00:59:18 TI has analog devices.
00:59:21 A lot of companies, Intel's apparently developing
00:59:24 processors there.
00:59:27 Bangalore, which was a backwater,
00:59:30 this is staggering.
00:59:33 The last time I was in Bangalore was like three years ago,
00:59:36 and it didn't look like this.
00:59:39 There were still cows in the roads.
00:59:42 I mean, this is a huge transformation for this country in particular.
00:59:45 It's hard to believe how fast,
00:59:48 you think that Moore's Law is changing technology fast,
00:59:51 but it's changing the geography of the world very, very fast.
00:59:54 And we have several regions, mainly in the south of India,
00:59:57 that are booming on the basis of this technology.
01:00:00 I guess the real big question
01:00:03 for this next century
01:00:06 and for the years ahead of us
01:00:09 is how fast will it spread to other parts of the world
01:00:12 and how will this technology
01:00:15 begin to help the lives of peasants in Africa
01:00:18 or Latin America be used and applied in new ways
01:00:21 that we probably can't imagine because we don't even understand
01:00:24 those domestic customers in those markets.
01:00:27 We have just begun to see
01:00:30 the implications of this technology,
01:00:33 and I want to applaud Gordon Moore for being there at the beginning.
01:00:36 Thank you very much.
01:00:40 Now, I understand that you will not be able to join the panel.
01:00:43 I'm so sorry.
01:00:46 Do you have time for one question from the audience?
01:00:49 Yes.
01:00:52 Is there anybody with a question they're pressing to ask?
01:00:55 If I don't see a hand, I am going to ask one question
01:00:58 that's pleasantly provincial.
01:01:01 On sort of behalf of greater Philadelphia,
01:01:04 as you take a look at positioning regions to be among the players,
01:01:08 you talked about the sort of diaspora out of Silicon Valley.
01:01:11 What's some of the generic advice that you give to knowledge regions
01:01:14 that maybe don't have as many Indian and Chinese immigrants
01:01:17 as Silicon Valley does, for example?
01:01:20 It's a hard one.
01:01:23 The thing that I failed to mention that I'll just add now
01:01:26 is the push factor for going to India and China in the late 90s
01:01:29 was the labor shortage.
01:01:32 Everybody was desperate for skill.
01:01:36 They saw, oh, I know some guys back home.
01:01:39 The people I went to school with, they can do this.
01:01:42 You could move into India or China at a very low wage level
01:01:45 and then move up the value chain, which they've done very quickly.
01:01:48 Pennsylvania doesn't have that luxury.
01:01:51 Your wages are higher, your cost of living is higher.
01:01:54 The one piece of advice that I would give,
01:01:57 not knowing as much about Philadelphia as I should know
01:02:00 before I give any advice whatsoever,
01:02:03 is get people together, break down the barriers
01:02:06 between the financial industry, manufacturing, and the university,
01:02:09 the technologists, get people having conversations
01:02:12 about what problems can be solved with ICs.
01:02:15 What is it that you can solve?
01:02:18 Then start figuring out what stands in the way of doing it,
01:02:21 helping people get started building new things,
01:02:24 starting new companies around the technology.
01:02:27 Is there anything that's sort of uniquely attractive
01:02:30 going out of these universities around the country,
01:02:33 these 50% who might otherwise return home?
01:02:36 I'm here in May. This is a beautiful city.
01:02:39 I'm not here in February.
01:02:42 The quality of life matters a lot.
01:02:45 Engineers love to be in places where they can afford to raise a family here.
01:02:48 I don't see how anybody just graduating from Berkeley
01:02:51 can afford to live in the Bay Area anymore,
01:02:54 whereas they could probably buy a lovely house here
01:02:58 Wonderful. Any other questions from the audience?
01:03:01 One last thing as an investor, how do you invest in China?
01:03:04 Everybody has it on the tip of their tongue.
01:03:07 If you're saying, okay, that's where there's going to be a lot of growth,
01:03:10 how do you invest in China?
01:03:13 I think that you have to be very careful. I think there's a lot of hype about China right now.
01:03:16 Everybody's sort of scrambling in.
01:03:19 There's a lot of overvalued Chinese companies right now.
01:03:22 I think if you want to invest wisely in China, get a Chinese partner.
01:03:26 Get somebody who's around there who understands the local markets,
01:03:29 who can build the Guangxi with the government,
01:03:32 who can really act in that environment. It's a very different environment.
01:03:35 I think the people that have had the hardest time in places like China and India
01:03:38 are the people that have flown in and thought it would be the same as it is here.
01:03:41 You have to really understand the local institutions, the local culture,
01:03:44 in order to invest wisely.
01:03:47 I would be careful.
01:03:50 There's a lot of money to be made, but be careful.
01:03:53 Thank you.
01:03:59 We'll be in touch. Thank you.
01:04:02 We are exactly on schedule, quite remarkably,
01:04:05 but I want to point out to people that, well, first,
01:04:08 one word on the Chemical Heritage Foundation.
01:04:11 Please, please have a boomerang effect here.
01:04:14 I mean, you've probably all enjoyed this wonderful physical plant
01:04:17 and gotten a sense of the wonderful mission
01:04:20 that Arnold Thackeray and his colleagues here have engaged in.
01:04:23 This is the central place for the central science.
01:04:26 Nobody's a better spokesperson for how central that science is than Gordon Moore.
01:04:29 The innovative work they do with fellowships and scholarships
01:04:32 and their Innovation Day and so forth,
01:04:35 they're deeply meaningful.
01:04:38 This is the kind of organization that we at the Eastern Technology Council
01:04:41 love to work with.
01:04:44 We hope you'll join us in thanking Arnold
01:04:47 and the entire Chemical Heritage Foundation for this tremendously
01:04:50 thought-provoking set of sessions.
01:04:56 I will end by reminding you that there's going to be
01:04:59 a terrific opportunity with this panel discussion.
01:05:02 You will never again have an opportunity to quiz people
01:05:05 so deeply shaped in industry.
01:05:08 We will be returning promptly at 4.20,
01:05:11 so please come back at 4.20. Thank you so much.
01:05:17 Thank you.
01:05:47 Thank you.