IEE ELECTRONICS DIVISION: CHAIRMAN'S ADDRESS

The business of VLSIC.A.P. Foxell, B.Sc, C.Eng., F.I.E.E., F.lnst.P.

Indexing terms: Integrated circuits, Components, Electronic circuits

Abstract: Very large scale integration (VLSI) represents not only the threshold of a capability for combiningover 100000 electronic components on a die of silicon, but also yet another stage in the growth of a volatileindustry. The incredible feat of harnessing a wide range of technologies to achieve this degree of integration,and even hold out the prospect for evolution to 106 elements and beyond, often tends to obscure the fact thatthis is also being implemented in high manufacturing volumes and at ever decreasing costs. As a result, it isexpected that the characteristic rapid growth of integrated-circuit production will continue over the nextdecade. However, the scale of the industry has now reached a point where fundamental changes are likely andthe pressures resulting from the manufacturing commitments and the market reactions will come to dominatethe purely technical influences. VLSI, especially via the leverage of electronic systems, now represents a majorinfluence on world economies, and the commitment necessary to cost-effectively produce such devices in volumestretches the resources of even the largest companies. The objective of the address is to outline, against thebackground of the evolution of the integrated circuit industry, the wide consequences of VLSI technology,which, by the contribution it makes and the magnitude of the resources needed to even participate, raises issuesat national and strategic levels which are likely to be determining factors in the future direction of the industry.

1 Introduction

Very large scale integration (VSLI) provides the capabilityof combining over 100000 individual transistors and othercomponents on a die of silicon a few millimetres square asshown in Fig. 1.

Fig. 1 One of the first VLSI devices

A 32 bit CPU containing over 600000 elements on a die 0.250 in x 0.250 in[Hewlett-Packard].

Although I shall be describing the general technicalbackground to the evolution of this complexity of inte-grated circuit I shall be concentrating, as the main themeof my address, on the creation and implications of theestablishment of a major new industry in the worldeconomy.

Perhaps because most in the electronics industry are sofamiliar on a day to day basis with the expansion of the

Paper 2709A, delivered before the IEE Electronics Division, 12th October I983Mr. Foxell is Senior Director of Development & Procurement, British Telecom,2-12 Gresham Street, London EC2V 7AG, England

integrated-circuit industry, we tend to forget that with adirect turnover of $15 billion per year and by creatingequipment of at least an order of magnitude greater invalue, it is now comparable to the automobile, steel, oiland similar industries, and moreover, by contrast, hasmajor potential for further growth. Indeed, every year, asmany devices are produced as in the entire previoushistory of the integrated circuit industry.

Therefore I want to concentrate not so much on thetechnology involved or the applications, or as it is oftendescribed in terms of technology push or the market pull,but on the influence of the industry itself. For it is my viewthat it is the dynamics of the integrated circuit industrythat is the characteristic that makes it different from otherindustries and that it is the momentum of the business thatinfluences both the applications and the technology, ratherthan the other way about.

To illustrate this aspect I review the industrial changesthat have occurred in the transition from the point-contactera, to the transistor, and then the integrated circuit andthen to the main topic of the implications of VLSI.

2 From point-contact to planar

The transistor has a considerable prehistory, and by the1930s there was a rising groundswell of research on semi-conductors with the solid state equivalent of the therm-ionic valve as its main objective. The Second World Warprovided the necessary impetus for such research and inparticular it was the need for the sensitive detectors for thenew radar receivers that provided the stimulus for much ofthis work. Quickly, silicon and germanium were identifiedas being the most promising materials, and by the end ofthe war extremely advanced point-contact diodes wereavailable. The crucial factor was development of the rela-tively pure semiconductor material and in retrospect thelarge amount of effort expended on the technology of thewhisker contact was the less important. This developmentwas undertaken by the traditional valve suppliers andtherefore, inevitably, the devices were treated commerciallyas yet another type of valve.

This was the situation in 1947, and in December of thatyear Shockley, Bardeen and Brattain were performingexperiments at BTL with point contacts on a germanium

IEE PROCEEDINGS, Vol. 131, Pt. A, No. I, JANUARY 1984 17

wafer and were determining the field distribution aroundthe point contact by means of another whisker. Theyfound that the potential applied to one contact affected thecurrent flow through the other, and thus the transistor wasborn.

In spite of the comparison with the valve—which waslarge, hot, fragile, demanding a great deal of power, anddoomed to wear out—the significance of the transistor wasnot immediately apparent, as these first devices wereextremely limited in performance with no two having thesame characteristics. This was so because the point-contacttransistor action took place between the two preciselypointed wires which had to be placed, laboriously by hand,less than a few thousandths of an inch apart. Again thevalve companies entered the field, and established manu-facture largely for protective reasons, some even calling thedevices 'crystal valves', but the sales were small and disap-pointing, owing to lack of reliability and reproducibilityinherent in their construction. The vital step in turning thetransistor into a viable product was the introduction ofalloy junction technology. Again it was the question of thenecessary complementary materials and technologiesbecoming available at the right time. The search forhigher-purity germanium had led to zone refining and thensingle-crystal growth technologies, taking semiconductormaterials to a level of purity above that for any othercommercially available materials. Such relatively perfectmaterial fostered the development of alloying and diffusiontechnologies for fabricating contacts and junctions, repla-cing the unsatisfactory point contacts. The introduction ofthese technologies marked the threshold of the semicon-ductor industry as we know it today. Indeed with one veryimportant exception, that I will come to later, virtually allthe basic technologies were then in place that areemployed in the VLSI industry today.

This development had a suitably dramatic affect on thefledgling semiconductor industry, for the new technologygave relatively reproducible products with good life char-acteristics which were amenable to volume production.However, existing valve companies soon found that it wasa different market, for no longer was it mainly a replace-ment business, but every sale was a new one and a differentcommercial situation emerged with the beginning of theintensive battle for domination of a growth market thathas raged ever since. It is also interesting to note that atthat time the major preoccupation was the transistor radiomarket. So that with relatively moderate investment (bytoday's standards) Europe and Japan were motivated tocreate new semiconductor facilities which regarded them-selves as being on relatively equal terms with those in theUSA.

However, in the USA, spurred by the 'cold war' and thefirst Sputnik, a large new market requirement developedfor high-reliability devices for use in the aerospace anddefence areas. To meet these demands, particularly foroperation at elevated temperatures, attention turned to thehigher-energy-gap material—silicon, first using the alloyjunction technology as had been employed for germanium.By this time, Shockley had left Bell Laboratories and hadformed the Shockley Transistor Corporation, to follow upspecific ideas on PNPN 4-layer switches. This failed com-mercially, but the outstanding team of engineers he hadrecruited subsequently formed the nucleus of almost everymajor US semiconductor company. In particular, one suchgroup under Noyce discovered the basic planar process,which led to the integrated circuit and all that has fol-lowed.

Noyce's team found that a stable oxide could be grown

on silicon and used to define a diffused area. Thus regionscould be doped to produce contacts or junctions in siliconmaking it possible to fabricate diodes, transistors, capac-itors and resistors. With the concept of planar technologyfrom Noyce, who then founded Fairchild Semiconductors,and the interconnected solid-circuit concept of Kilby ofTexas Instruments, the integrated-circuit industry as weknow it was born in 1960.

Planar technology can be applied in a variety ofsequences to create a host of devices. No one device struc-ture or circuit form stands out as the best solution for allrequirements. In practice the skill with which a particularsolution is realised is often more important than the choiceof a particular technological approach. However, bipolarprocesses tend to be fast but more complex to fabricate,whereas MOS is slower but simpler, but in any case, asthese processes are extended there is an inevitable trend toconvergence in capability and manufacture.

3 The IC business

The market conflict between the transistor manufacturershad decimated the traditional valve companies, and nowthe fight for the IC market was contested by the remainderin contention with a number of new IC companies rep-resenting offshoots of the major electronics firms. The bur-geoning US defence needs, also reflected in computers,created the environment for another commercial war ofattrition, this time over the supply of the standard IC logicfunctions that had emerged, such as DTL, TTL and ECL.In this 'war of the logic families' various commercial stra-tegies such as penetration pricing and product innovationwere escalated to new levels, but so too were the problemsof fluctuating production yields and volatile user demand.These factors have remained characteristics of the ICindustry, which has become imbued with the achievementof growth of the order of 20-30% per year coupled with areduction in real unit costs of about 30% per year.

The use of aggressive pricing and marketing in order toachieve market domination (not always by the best techni-cal product) became just as important as the technologyinvolved. A typical forward pricing strategy is shown inFig. 2. This encourages market penetration, often in a vitalequipment development application at an early stage, bymeans of a low price, while hopefully recouping the lossover higher volumes later in production. Although in iso-lation this may seem a logical approach it tends to assumethat competitors adopt a passive stance, whereas in prac-tice many counter ploys are used, thus creating uncertaintyover the prospect of recouping early losses.

Integrated-circuit companies also sought to obtainmarket share by means of continual product innovation.However, with many competitors in the field as shown in

product A product B

^manufacturer 3

manufacturer 2"""""

timeFig. 2 Forward pricing strategies

18 1EE PROCEEDINGS, Vol. 131, Pt. A, No. 1, JANUARY 1984

Fig. 3, any product had to be outstanding in order tosucceed, but in achieving this it would be copied by others.Up to a point this is desirable, as users desire the assur-

15

o 10

1975 1980 1983

Fig. 3 Increase in number of variants of the 16 K DRAM

ance of a second source of supply. So the optimum strat-egy was to encourage a second or even a third source butto also ensure they were sufficiently lagging to enable theoriginator to dominate the market. Obviously in achievingthis ideal, technology often took second place to marketingand other pressures. Indeed such has been the complexityof this type of situation that to date no major IC manufac-turer has been able to move from one dominant product toanother of the same type. Often this is due to the concen-tration of corporate resources on the exploitation of theinitial product, which limits the ability to introduce asecond-generation development, while competitors canbenefit from becoming second in the field.

However, these strategies created by the suppliers ofintegrated circuits have to be set against the background ofthe rapidly fluctuating demand of the equipment userswhich in itself is, of course, an interaction with previoushistory of supply. Predicting the appropriate level of pro-duction commitment in these circumstances is extremelydifficult and is compounded by the complementaryproblem of assessing manufacturing output in a growthenvironment where there are conflicting trends from

(i) increased output due to the learning curve i.e. exist-ing operator experience, technical understanding, etc.

(ii) decreased output due to the introduction of newand different plant and technologies with new operatorsand a dilution of experienced supervision in order toexpand basic output.

Indeed the net effect of most attempts to dramaticallyincrease output has been to reduce it! However the long-term trend is as indicated in Fig. 4.

It also has to be remembered that such are the pressuresof the industry that new devices are often launched on thebasis of a yield of less than 1% satisfactory die/wafer.Although mature products may eventually achieve yieldsof 90%, where there is intense commercial pressure, assoon as a yield approaches 20-30% there will be a motiva-tion to redesign the device in order to eliminate any inher-ent design weaknesses or to scale-down layouts to increasethroughput. However, such design recycles in themselvesoften introduce new manufacturing problems which canagain reduce yields to a few percent as the tendency insuch critical situations is to correct symptoms of low yield,rather than causes (Fig. 5). Thus the average yield of a keyproduct is often around 10% owing to these continualchanges, for example the typical yield of 64 K DRAMdevices in 1982 has been estimated to be about 14% beforeimproving to a current figure of about 40%. Even so, suchyields are average figures and cover considerable wafer-to-wafer and batch-to-batch variations, and thus there are

usually significant difficulties in predicting manufacturingoutput.

100-

10

1 -

, \1977 \

-

•1978\

\ «1979\

\\

\ «1980\

\ .1981

\ .1982\

70*/» cost reduction \for doubling volume \

0.1 1 10 100 1000cumulative volume, millions

Fig. 4 Learning curve for the 16 K DRAM

100

50

timeFig. 5 Yield during product life

When these difficulties were experienced in the early1970s and compounded with the previously mentionedproblems of increasing output to a volatile market (Fig. 6),

S.-20

Fig. 6 Change in demand for ICs

it is not surprising that many of the protagonists retiredfrom the fray owing to their inability to penetrate themarket and inbalances of production and stocks. Thesetended to be the IC units of conglomerates where theorganisation could not respond to the rapidly changingmarket pressures created by the larger companies special-ising in semiconductors. However, in the USA they weresoon challenged by a new generation of entrepreneurialcompanies.

Often the initiators of these new companies came fromthe larger organisations, and they clearly saw the tremen-dous potential for specific LSI products and electronicequipment. Probably the most significant examples arosein the introduction of memory circuits and then the micro-

IEE PROCEEDINGS, Vol. 131, Pt. A, No. 1, JANUARY 1984 19

processor based on MOS technology, which, according toone's viewpoint, either initiated or sustained the massiveexpansion in computing demand during the 1970s. Manysuch companies failed, but a significant number not onlysurvived but generated outstanding growth to form agroup of IC companies rivalling the traditional leaderswho were largely supported by their discrete semicon-ductor product activities. During this period it is inter-esting to note that the rating of US semiconductorcompanies became much more related to their growthprospects rather than their current financial performance.

During this period, integrated-circuit technologyadvanced dramatically from SSI to LSI, and now to VLSI.

4 The implications of VLSIThis progress brings us to the present day. In practice,VLSI is basically no different in concept to the earlier ICtechnology. It was formally initiated in 1978 by theannouncement of the Japanese MITI (Ministry of Interna-tional Trade and Industry) organisation of a nationalprogramme to achieve an order of magnitude improve-ment in all basic IC capabilities (i.e. in speed, complexity,power and cost) when coupled to a complexity of morethan 100000 devices on a chip. I will return to the indus-trial motivation and significance of this programme laterin this paper, but I would first like to outline the technicalaspects of VLSI as these set the scene for many of thebusiness implications that follow. The overall strategy forachieving VLSI performance is, in essence, an extension ofthe previous approach, namely increasing the packingdensity on the chip which improves performance andreduces cost. This is mainly be means of

(a) ingenious circuit design(b) refined device structures and processing(c) larger die(d) better yield and throughput from improved repro-

ducibility and production control.

In spite of the immense effort devoted to IC design it isstill possible to create innovative designs which either gen-erate a more efficient layout or enable an extremelycomplex function to be realised within the limitations ofthe technology.

In parallel, advances in materials technology oftenbased on dielectric isolation or epitaxy have encouragedthe evolution of new device structures and intercon-nections, while the introduction of sophisticated litho-graphy, dry etching and ion implantation has opened upthe possibility of achieving one micron structures on aproduction scale.

The availability of larger wafers of silicon over 4 indiameter, makes the processing of larger die feasible andthe consequent yield problems can be overcome to someextent by redundancy techniques whereby defective ele-ments can be replaced by spare cells contained on thesame die. With wafers of this size the whole process cannow be linked together by automatic transfer machinery totransport wafers to create a truly continuous process in abenign environment. Such production lines are amenableto computer control and thus can be optimised to providebetter and more consistent yields and throughput.

From a business viewpoint the major increment ininvestment now necessary even to make one VLSI device,and the fact that such a plant will only be viable whenoperated at near capacity with an output of millions ofdevices each year is probably the major new dimension ofVLSI. In 1970 a $2 million wafer plant turned out $20million worth of products as the minimum commitmentnecessary to enter the field, now the threshold is of the

order of $50 million for the plant to produce over $200million worth of devices. In other words with VLSI thestandard product industry has changed from one favouringthe entrepreneur to one based on massive investment andgearing. Indeed, commitments on this scale are now onlylikely to be made against the background of a nationalpolicy.

The other significant factor that stems from investmentsof this magnitude is that they are only effective if the plantsare operated at full capacity. Not only does the VLSI tech-nology mean that extremely complex devices becomeavailable but, perhaps even more important, simplerdevices can be produced in volume at lower cost than everbefore (Fig. 7). This is precisely what is required for stan-

yield cost100*/. 0 l i

silicon

wafer processing

probe test

assembly

final test

total factory cost = $ 1.92Fig. 7 Cost and yield analysis of a typical 64 K DRAM

dard products, such as the logic families, memories andmicroprocessors and to sustain the dominant consumerproduct markets, but it is increasingly in conflict with theflexibility necessary as one moves to dedicated productsfor the telecommunications, automotive and defencesectors (Fig. 8). Some of these, such as a Codec, may be

100r

50

military

/

yS computers anddata communications

^

^ ^ ^ ^ ^ consumer

^ ^ telecommunicationsindustrial : : ^zr^rr^^r—-^ —

1962 1967 1972 1977 1982

Fig. 8 IC market shares

required in adequate volumes of the order of millions ayear which warrant loading a VLSI facility, but the widevariety of more specialised devices needed in significantbut smaller numbers tend to be incompatible with suchmainstream production facilities. Although unattractive tothese mainstream IC manufacturers the availability ofdedicated ICs and custom ICs are the key to the cost andperformance of a large segment of the electronics industryand as a result four approaches have emerged to try tomeet this demand:

(a) the creation of captive facilities (Fig. 9)(b) silicon foundries(c) gate array, multi-device wafer and cell technologies(d) programmable devices.

The relative importance of these is still difficult to judge,but between them it is expected that they could accountfor 50% of the market by 1990.

5 National strategy

VLSI manufacture is a major activity in its own right andis estimated to grow to a value of over $50 billion by 1990.

20 IEE PROCEEDINGS, Vol. 131, Pt. A, No. /, JANUARY 1984

The Japanese were the first to formally recognise thatthe availability of VLSI technology was of crucial strategicand economic importance. Against the background of the

open market

1974IC sales (USA)

1982

convergence between computing, telecommunications,office and consumer products (now termed 'informationtechnology') MITI concluded that it was essential for theirvertically integrated electronics equipment companies topossess VLSI capabilities. Significantly they recognised theadvantage of creating the support and the added economicleverage derived from a total infrastructure ranging fromthe relevant chemicals, silicon, process equipment, testersto software. Indeed in both hardware and software areasJapan has also encouraged the training of engineers tosustain this programme of expansion at a time when theWest has been retrenching. The Japanese co-operativeVLSI research programme as part of an overall nationalstrategy using their home market as a base to penetratethe US and European markets has had a marked successto date. All marketing aspects from high quality, readyavailability, aggressive pricing etc. have been pursued aspart of a concerted plan, and whereas their product rangehas not been outstanding it has been adequate andenhanced by high productivity in manufacture. Equallywith Japanese technical papers now dominating mostinternational conferences, an aura of competence has beencreated in their VLSI products. Initially targeted on estab-lished memory products, the concept has been that pro-ficiency in this area can be applied to either reduce costs orincrease technical sophistication in other devices. Not onlyare Japanese VLSI manufacturing units now being estab-lished in the USA and Europe but their complementaryinfrastructure now also has a dominant position in theworld in the supply to their competitors of essentialmaterials and production equipment.

As a consequence the USA and Europe have suffered adeclining share of the VLSI market (Fig. 10), although,with the overall growth in demand, their output in abso-lute terms has continued to rise. This loss of competitiveedge has been variously attributed to low profit margins,the high cost of capital, shortage of skilled staff and lack of

1960 1970 1980Fig. 10 World IC production shares

1990

a national research and development strategy. The USAreaction has been along the following defensive lines:

(a) improvement in quality levels(b) the launch of a DOD VLSI research programme

(VHSIC)(c) the creation of a cooperative civil VLSI R&D

programme(d) embargos on VLSI technology transfer to other

countries(e) sensitivity to commercial espionage and(/) concentration on products such as customer specific

devices not targetted by the Japanese.

However, in order to counter the direct Japanese com-petition there has also been a more pragmatic tendency inUS companies to harness the Japanese skills by investingin joint operations, establishing manufacturing plants inJapan and licensing their technology.

Somewhat similar responses have occurred in Europewith the EEC originated Esprit project identifying VLSI asa key technology for collaborative industrial research, andin the UK a similar initiative proposed by the Alvey Com-mittee is now being implemented. However, in the UK, theoverall picture of integrated-circuit activity is that shownin Fig. 11, indicating that the truly indigenous companiesthat have tended to concentrate on selective products arein a minority position in relation to the total market whichis dominated by multinationals.

6 The future

The escalation of VLSI technology shows no sign ofabating. It is likely to continue to be based on siliconowing to its low cost and availability (particularly in largerwafer diameters) in relation to the most likely competitor,gallium arsenide. The massive investment in silicon pro-cessing represents a considerable threshold for any newmaterial to overcome.

Process technology will also probably be extendedalong the present lines with the emphasis being on betterdefinition by moving from conventional projection towafer stepper and then electron-beam lithography. Pro-cessing will rely increasingly on vapour deposition and dryetching with the provision of adequate interconnectiontechnology becoming of major concern.

Although the scaling of existing structures still has con-siderable potential, a number of effects which were pre-viously of a second order and could be disregarded willnow tend to limit performance, particularly

(a) material uniformity and stability(b) electrical breakdown effects(c) sensitivity, i.e. soft errors due to alpha particles.

However, it is very likely that these problems can be over-come, indeed 256 K byte memories are now being sampledand research groups are already demonstrating, albeit in acrude manner (and assisted by built-in redundancy), thestructures needed for a 1 Mbit device. With reasonablescaling within the bounds of current theory and smallextensions of existing technology these devices should bereadily converted to production items. The same tech-nology base should enable exploratory research models tobe fabricated an order of magnitude beyond this complex-ity, but more radical changes in technology will berequired to convert these prototypes into viable manufac-ture.

Having predicted that there is no fundamental problemto technically achieving at least a further order of magni-tude increase in complexity, power, speed or performance,

1EE PROCEEDINGS, Vol. 131, Pt. A, No. 1, JANUARY 1984 21

nevertheless, from a business viewpoint it will again esca-late the consequences of both the creation and absorptionof such new products and also the consequent investmentnecessary to bring it about.

The creation of such devices, not only of VLSI com-plexity but in a wider variety and number than ever before,will increase the pressure to realise adequate CAD tools.Whereas numerous CAD packages are currently available

Although the Japanese electronics industry has alwaysbeen based on such a vertically integrated structure, it isonly recently that this trend has re-emerged in the USAand Europe. In the West this is now being brought aboutby two mechanisms. Either major electronic companies arecreating captive 'in-house' facilities oriented to meetingtheir own requirements for strategic reasons (e.g. IBM,Western Electric, NCR, General Motors, Kodak) or they

UK basedproduction€130 m

rest ofworld ICproduction£8300million

UK-ownedproduction£70m

Fig. 11 VK IC trade flow

their very variety and their inadequacy is prejudicing theirwider adoption. The use of gate-array and cell technologieshas, and will, ease this problem, and the embryonic siliconcompilers are beginning to emerge. However, the vitalingredient still needed to enable the electronics industry toexploit VLSI technology is a suitable silicon compiler withgeneration of the relevant test programmes for suchdevices.

The market of VLSI will represent the whole spectrum,from super megabit stores and complex microprocessors tostandard products, which will be of a complexity corre-sponding to today's state of the art and manufactured inimpressive volumes of ten to hundreds of millions of piecesper year. It follows from these volumes that the take-up ofsuch devices is most likely to come from not only what wecurrently regard as the consumer sector, but will increas-ingly reflect the pervasive employment of sophisticatedelectronics in automobile, telecommunications terminals,control, computing and similar applications.

Such volumes imply that the industrial commitment tocreate these products and manufacture them and generatethe necessary software will be raised to an even higher levelthan ever before. Again, although gate and cell technol-ogies will mitigate this pressure to some extent, it is diffi-cult to see how, in broad terms, such investmentconsiderations will not lead to further vertical integrationof the VLSI industry.

are acquiring existing mainstream VLSI companies andcontinuing to operate them on a vendor basis. This type ofacquisition has become possible as a result of intensepressures on the main VLSI companies arising from thepresent period of relatively low growth placing moreemphasis on profitability than before, while recognisingthat they need to match the massive Japanese investmentprogrammes in products and, facilities if they are to remaincompetitive in the long term.

The key to their future success will be the ability of suchWestern vertically integrated operations to optimise theiractivities to the extent apparently achieved in Japan wheresignificant benefit is derived from harnessing the in-housesystem requirements to the volume production for otherusers. In this way it is feasible to conjecture that thepresent Western captive manufacturers of VLSI will haveto seek wider markets outside their own organisations, andthe independent manufacturers will seek to ally themselveswith equipment suppliers in order to justify the increasinglevel of investment. In turn this will create a more difficultenvironment for the remaining independent manufacturersof volume and custom VLSI. Although the recent upsurgein availability of US venture capital has created some 20new such companies since 1980, many are attacking, albeitwith different technologies, the same market area of cus-tomer specific circuits. However, how many will survive inthe inevitable 'overkill' situation is a matter of conjecture.

22 IEE PROCEEDINGS, Vol. 131, Pi. A, No. 1, JANUARY 1984

7 Conclusion

VLSI technology can probably be extended for anotherdecade, the main potential limitation being the ability tocreate adequate CAD facilities capable of exploiting theprocesses. However, participation in this thrust to escalatesemiconductor technology to this level has consequencesin dramatically increasing the inherent capital intensityand viable device outputs. The response has either been ona national basis, typically as in Japan, or via conglomer-ates as in the USA and Europe, thus emphasising the shiftfrom competition based on technology to one orientatedtowards productivity.

Apart from compexity and low cost, VLSI also offersthe improved performance that is vital to the competi-tiveness of the professional equipment industry, but suchdevices will not be required in the large numbers thatresult from the conventional investment in manufacturingfacilities. New methods are being sought to meet this stra-tegic demand, either by utilising standard processes or bycreating dedicated facilities (i.e. custom, gate-arrays etc.),but the relative success of these routes is problematicalwhen, as so often in the history of semiconductors, attrac-tive concepts have to fight for survival in a volatile com-

unit growthinflationprice dropdollar growth

mercial environment. Indeed, as the pressures are nowmercial environment.

After two decades of tumultuous growth it is reasonableto ask whether the integrated-circuit industry, now trans-formed into a VLSI dimension, shows any sign of matur-ing. Fig. 12 would seem to indicate that, in percentageterms, there is a trend for the business parameters to sta-bilise. However, as Fig. 13 shows, the absolute values are

$15B

$10B

1960 1970

Fig. 13 Trends in the IC business

1980

fin" 1 9 6 5 1970 1975

Fig. 12 Growth in world IC sales

1980

still increasing dramatically so that in pragmatic terms thevolatility remains. Indeed, as the pressures are now escalat-ing to strategic and national levels, the only certainty is theextreme uncertainty of the business of VLSI.

8 Acknowledgments

Although the content of this address represents my person-al views, I would like to thank my colleagues with BritishTelecom for their valuable assistance in preparing thispaper, and also those colleagues in the semiconductorindustry who have provided background information.

IEE PROCEEDINGS, Vol. 131, Pt. A, No. 1, JANUARY 1984 23