In the first of two articles, our sister publication, Computer Business Review examines how microchips are sold and investigates the emergence and increasing influence of the independent distribution and broking market for semiconductors – especially memory. The second article tomorrow, ‘The megabit bandits’, will examine how the economics of the open markets have helped to create the ideal conditions for the chip crime which is plaguing industry, and the efforts of the authorities and the computer industry to tackle the problem.

In an old building in a small town nestled in the rolling hills of the English countryside, employees of a small company, Deep Blue Technologies, are testing, tagging and packaging memory chips ready for dispatch. Each box is labeled with a bright yellow sticker declaring that these single-inline memory modules (SIMMS), are guaranteed original parts, built by IBM Microelectronics, a prestigious supplier. Then, after being boxed, the address labels are stamped, revealing the name of a customer that is no less prestigious: The IBM Corporation. It needs the chips urgently to meet an order for disk controllers. Deep Blue, says James Willard, the American expatriate who founded the company, was able find the IBM chips more quickly, and more cheaply, than IBM itself. Across the Atlantic, in the New England town of Peabody, Massachussetts, telephone traders at NECX are making the same boast to their hundreds of customers. While the latest prices roll past on the electronic trading board, the 125 brokers work ‘the pits’ – matching buyers and sellers, hunting down specialist components, consulting a database of some 20,000 parts, and feeding in the latest data from trading partners and customers across the world. It’s just like a stock exchange, says Peter Cannone, worldwide head of sales for NECX, It’s an international components exchange. Meanwhile, at Moscow’s international airport, an ‘entrepreneurial’ Russian is negotiating his way through customs. In his hand-luggage, wrapped in anti-static bags, customs men find tens of thousands of dollars worth of SIMMS strips of uncertain origin. He has, he says, bought them from dealers in the West, and they are destined for one of the many small PC manufacturers which have sprung up in Russia in recent years. After showing some influential paperwork from the Russian authorities, he is waved through. These are just three faces of what has become one of the computing and electronics industry’s fastest growing, most pro fitable and most controversial markets. The multi-billion dollar international trade in semiconductors is a business that spans the internationally-regulated, multinational agreements between manufacturers, governments and equipment builders at the top end, and moves down through many shades of the free-market grey areas into the black market at the bottom. Although little understood and often publicity shy, component traders are playing a key role in the changing dynamics and economics of the computer industry. Semiconductors have been sold in increasing quantities since the early 1970s, but it is only in the past three years that the independent channels (the grey markets), have become so large and important, and – although few suppliers will admit it – so central to their strategies and logistics.

It is a business which has never been accurately sized, but which – if the independent SIMMS card manufacturers (which build memory strips from chips) are included, is certainly worth in excess of $30 billion a year. It is a business which is only just beginning to earn a mantle of respectability, but which is driving dramatic changes in the economics of the semiconductor industry; a business where the dynamic and complex problems of planning huge investments in naturally volatile markets not only encourages spot market dealing, but which is creating new possibilities in the options and future trading markets. It is also a market which, because of its size, fluidity and lack of regulation, provides the ideal backdrop for the black market eers, traders in stolen, faulty and counterfeited chips.

THE BROKERS

Behind the razor wire, video cameras and the other hidden security measures at the historical listed building which houses Deep Blue, Willard cautiously eyes the crumpled business card which is proffered to him. He is suspicious of visitors. The com pany specializes in SIMMS, the most saleable and most valuable form of memory, and like many memory brokers, the company is a not just a natural target for thieves, but has been visited by police investigating the wide channel into which hundreds of millions of dollars worth of SIMMS have disappeared without trace. Willard’s willingness to receive visitors is unusual. In the past, many of the memory brokers, especially the smaller ones, have been pilloried by the major manufacturers as parasites, as disreputable as used car dealers, making them shy away from publicity. A casual enquiry to one such supplier, Drakus Marketing, meets with the response we do not do any publicity. Another simply states: We’re not prepared to answer any questions. But Deep Blue is one of a small group that are attempting to emerge as bona-fide, quality-conscious traders, offering warranties, branding, and conforming to quality requirements. Overcoming his reservations, Willard outlines the way his memory broking business works. It is not dissimilar to trading commodities in the City, he says. And, like City traders, he says, We are not afraid to speculate. Chips are located, often from abroad, at one price, and then sold on at a higher price. Inventory, which is all tested and marked, is increased or decreased, if the company anticipates price movements. The key is intelligence gathering and contacts. On a whiteboard, he and his colleagues have scrawled down recent prices for DRAMS, v arious configurations of SIMMS and other similar components. Well thumbed trade magazines, databases of contacts, and the Internet all help to provide contacts and data. But mostly, the intelligence is gathered on the phone, while the expertise to make the buying decisions comes from years of experience. This year, says Willard, Deep Blue will see revenues of 7 million British pounds ($12 million); next year, he says, it will be lot higher.

While the scale is a lot larger, the methods of the US’ largest trader in memory, NECX, are much the same. The traders bark out telephone orders like commodity dealers. Acronyms and jargon – DRAMS, DIPS, flashes, boot blocks, SOICs and PLCCs – litter the negotiations, along with brand names such as Intel, Samsung, Texas, Kingston and AMD. Who are they talking to? Resellers, original equipment manufacturers (OEMs), manufacturers, Fortune 2,000 companies, says Peter Cannone. We make connections between people who want to sell and people who want to buy. Mostly, he says, these people are not speculators. They need to buy or sell parts for pressing commercial reasons. Early in the year they were receiving calls from big PC manufacturers and memory makers with warehouses crammed with inventory they no longer needed. Cannone describes NECX as a ‘solutions provider’. We make these people look good. Cannone knows that, mostly, his customers do not like to be named. Publicity material and press cuttings reveals the names of AT & T, Bay Networks and LSI Logic. Among the rest almost every major computer and telecommunications equipment supplier in the US and many more outside – as well as the major semiconductor manufacturers. In fact, most brokers can list some surprising names among their customers. Even Samsung [the world leader in building memory chips] has ‘gone to the street’ to meet extra demand, says Willard. This, he says, explains how his company came to be supplying IBM memory chips to IBM itself. It’s pretty normal. When a big supplier needs parts, they will buy them from where they can, he says.

SECRECY

It is one of the most striking features of the component distribution industry. Like shares, small companies sell to large, and vice versa. Chip lots are frequently split and sold off, then joined with other lots for larger sales. In the process, at tempts at quality control and part-number tracking become difficult. For this reason, reputable companies prefer to do their buying quietly. Over lunch, one executive confides in CBR that his company, which adheres to international quality standards , will occasionally source components from the spot markets at traders’ prices as opposed to contract prices. But you can’t print that, he adds. The secrecy is at odds with the profile and the size of NECX. Co-founder, Henry Bertolon, is feted at conferences, and was recently described by Joel Birnbaum – the inventor of the RISC processor – at an IDC conference, as one of the brightest men working in the computer industry. NECX is expanding into Internet sales and has broadened into official distribution of components and subsystems. In 1995, NECX had sales of $640 million, and revenues may reach $1 billion this year. The main difference between us and a stock exchange, says Peter Cannone, is that we take title to what we buy and sell . That means that NECX, for a few days at least, becomes the legal owner of the property. Goods are brought into the company’s 50,000 square-foot-high security warehouse, checked and shipped. Satellite warehouses will eventually be put in place in Ireland, through its alliance with Ctech International, and the Far East. Like all the others in the semiconductor trading business, NECX does not like to be caught holding stock. Bertolon compares it to fish, which after a few days, begins to stink. The usual objective is to move stock fast. But that does not mean that NECX is averse to a little speculation. If the market is going to move up, we will take a position, says Cannone. Bertolon boasts that NECX’s high-tech headquarters in Peabody, Massachussetts, was paid for by one gamble: Its dealers heard that demand was starting to soar for small outline integrated circuits (SOICs), and that the manufacturers were unaware of the pace of demand. NECX took a huge position in all the stocks it could find, later selling them at a vast profit. But NECX is unique only in the way in which it has formalized its trading systems, established a reputation for quality, and grown to a certain size. Across the world, there are probably 20,00 0 significant, established, bona fide organizations which generate a substantial part of their revenue from buying and selling chips – a figure that excludes consumer retailers, and the opportunistic interventions of individuals and companies which are not primarily involved in chip sales. Many of these are dealing in ports which they may never see and whose origins and whose ultimate destination they may never know. These organizations span the big memory manufacturers – of which there are ab out 30; the official distributors; those companies such as Kingston Technology, which manufactures SIMMS strips and boards from chips (and is the world’s biggest single purchaser of memory); and the many semi-official or unofficial distributors, dealers, and system manufacturers, many of whom also buy and sell components. Virtually every major electronics distributor has a unit dealing in memory, says Joe D’Elia, semiconductor analyst for research organization Dataquest. So too, he says, do most PC makers, which have ‘broker desks,’ buying and selling SIMMS during times of shortage and glut. Two UK companies, Memory Corp and Syntaq, have even developed patented technologies for re-building faulty memory circuits into working SIMMS modu les. They add circuitry to a SIMMS module to route around faulty cells. The chips are quietly sold to them by the manufacturers, who always remove their names, and re-emerge in modules which sell below the market price. It is this variety, and the s ize and spread of the fast-moving distribution chain, that makes it possible to ‘launder’ even large amounts of stolen chips – especially DRAMS and SIMMS, which account for 90% of traded chips. According to Frank Barry, Deputy District Attorney of San Jose, one batch of stolen SIMMS modules tracked in California changed hands 18 times in 24 hours. The capacity of these dealers is staggering – especially in the far east – where large volumes change hands on a regular, but ad h oc basis. You could get rid of a few million dollars worth of memory in the far east, no problem, says Fergus Campbell of Datrontech, a publicly listed supplier of memory and a distributor for Kingston Technology. There are probably a hundred deal ers in the US, he adds, capable of taking those kinds of volumes in a single deal. Before the Chinese New Year, advises Willard, it is good to buy memory, since the lack of trading in the far east helps to firm up prices. Another memory company executive warns that prices fall at the turn of the year, because PC makers sell off excess stocks to lower the inventory on their books and improve their cash-flow. Campbell, of Datrontech offers some simple advice: Beware of buying before Easter becau se of two four-day weeks. Preoccupation with price is natural. Prices are extremely volatile says Cannone of NECX. If we publish a price at 10 am, it could have changed by 2 pm. NECX, which partly attributes its success to its real-time parts database and trading system is able to track prices well enough for some others to use it as a kind of market-maker system for the industry. But, emphasises Cannone, The market sets the prices, not us. Even the memory manufacturers, who were once ab le to price to en- sure good profits, are now saying the same. For a broker, of course, there is always some control of prices. The seller will set one price, the buyer another, and the dealer will try to push the two as far apart as possible. And u nlike in the stockmarket or in currency trading, where market makers will usually publish ‘the spread’ (and in most countries are obliged to), the memory trader’s spread is never published. The inefficiencies of the market mean that the profits are often huge – even in a straight deal which involves no speculation and no overnight price changes. The key to it all, says one dealer, is to buy in one market and sell in another. He buys on the spot market in the far east, and sells direct to P C retailers in the West. The other tactic is to ensure that the buyers and sellers never exchange business cards. The middleman’s role is always fragile. The profit structures in the memory business are, in spite of recent price falls, far more gene rous than in the PC industry which they largely serve. Prices, which always depend on quantity, on the configuration, the type and the manufacturer, have clearly allowed for good profit margins in the past. In fact, the continued run of high prices, in spite of the recent upturn, along with the promise of the continued high demand for DRAM, has encouraged the enormous new investments in memory plants and accounts for the high profitability of companies such as Samsung. An example of the profit s can be seen in the market for 4 Mbyte SIMMS strips, made up from eight 4mbit DRAM chips. The manufacturing cost of the 4mbit chip, say manufacturers, is about $3 to $3.50. One source representing a major manufacturer, says that $1 is more realisti c. In the fall of 1995, 4mbit chips were trading on the spot market for upwards of $12 each, allowing for a huge profit to the memory makers after selling, general and adminsitrative (SGA) expenses. On the spot market, these prices were reflected in SIMMS prices of $130 to $150 per 4Mbyte SIMMS – another 50% mark-up. Retailers were charging above $200 for 4MB. In Spring, spot market prices fell dramatically – at one point to just above $5 per 4mbit chip, or lower in larger quantities. This bro ught the price of 4MB SIMMS tumbling down to less than $50 on the spot markets, and in some cases down to a break-even of $40. But while many dealers sold at a loss or ran into trouble, many more did not, shedding their inventory, selling in advance , or holding on for an upturn. Many PC retailers, enjoying the flush of demand for extra memory caused by Windows 95, thrived. Even with spot prices as low as $50, one retailer was selling 4 MB SIMMS memory for $200 in April. While the pricing struc ture reveals how much money various dealers are able to make by dealing in memory, the volatility reveals the role of the big PC suppliers in determining prices. The price fall early in the new year, says Dataquest, can be attributed to PC vendors pushing DRAM chips which they had purchased on the contract market back into the spot market. This is not an ad hoc panic measure, but is a standard tactic for PC suppliers. In fact, the role of the independent dealer is now so clearly established, and the volume of business so large, that the free and open market is having an increasingly influential role in semiconductor economy. It is becoming increasingly difficult for the big companies on both the supply side and the equipment manufacturing side to avoid playing a more active part – and to be seen to be involved.

THE CHIP MAKERS

According to CBR’s estimates, between 60% and 70% of the world’s $150 billion semiconductor market is supplied under contract to the final manufacturer – leaving a third to go through distribution channels. But even if these chips are supplied under contract, an unknown proportion may be later resold into the market or become the subject of traded delivery options. The independent market initially sprang up to solve simple logistical problems and to match supply with demand, but its role has now become vital – partly because of the changing economics of the microelectronics industry and because of three clear technical trends. First, the miniaturization of the chip making process has pushed up the costs of the capital equipment required for state-of-the-art manufacturing. According to the VLSI Research of California, each tenth of a micron of improvement in manufacturing tolerance at the design level, can lead to exponential increases in the cost of capital equipment. New plants being planned for 0.4 micron level manufacturing by Intel, for example, will cost above $2 billion, and some new plants may soon cost $3 billion. The implications of this are simple: major manufacturers – especially those in Taiwan and Korea, where labor is cheap – will produce as much as possible, irrespective of what price they can eventually get, or what components they are contracted to deliver. Those products that are not ordered in advance under contract are sold to both official and unof ficial distributors. NECX is now used so routinely by major chip manufacturers that in some cases it has set up formal electronic data interchange links for the exchange of inventory information and trading documents. Second, there is a clear trend towards the standardization of parts, especially memory, which is helping to create a large market in tradable commodities. While there are still thousands of specialist devices, the major volumes are for Intel and Intel-compatible microprocessors, and PC-compatible SIMMS memory strips. Third, the expertise in chip making is moving inexorably away from manufacturing towards design. Any major organization with the capital and the willpower can buy the designs and the semiconductor manufacturing equipment and enter the business. Several Taiwanese and South Korean companies, such as Taiwan’s Nien Hsing Textile, which began as a textile company, have done just this, entering the high-margin memory business in spite of their lack of expertise . The result is more manufacturers, with fewer formal distribution chains feeding the spot market. At present, Taiwanese companies are spending about $17 billion on new memory plants. Some of these manufacturers have little choice but to feed the sp ot market. Their economic model requires that they do not lead the industry in terms of innovation, only in price. This can mean that they are ruled out of some of the leading-edge contracts with the big PC manufactures.

MONOPOLY ECONOMICS

All of these factors combined would have been enough to break down the old structure of the industry. But there is a further factor which is being fueled by a new orthodoxy in the business schools of North America. The theory is that in technology businesses in particular, competitors can be warded off by using huge volumes to drive down costs. At the same time, the barriers to entry are increased because of a faster pace of innovation, the use of state-of-the-art expensive equipment (such as in chip manufacturing) the need for high intellectual capital (such as in chip design). The application of this approach was clearly seen when Intel introduced its Pentium processors. Vinod Dham, a former Pentium chief engineer now with rival AMD/Nexgen, testifies that this ‘monopoly economics’ was a deliberate policy that came out of Intel’s strategic long range planning (‘SLURP’) sessions that the company undertakes twice a year. In 1993, he said We made a big change. We sped everything up, ramped it all up. We had never done anything like this before. The goal was to turn as much of the PC industry over to Pentium as fast as possible, sidelining competitors and moving to smaller design rules – 0.6 micron – while competitors were left behind. A more crude, but equally effective approach has been adopted by the Taiwanese and Korean DRAM and flash memory chip manufacturers, which, by investing in huge new factories, have built their share of the world semiconductor market up from 8% to 12% in the past three years. In each case, the traditional economics of balancing supply with demand has been replaced with a drive to increase manufacturing volumes and cut unit costs – the goal is to create demand and take market share. But the approach requires that these manufacturers secure immediate markets for their products. Building for inventory, except for very short periods for speculation, is not an option. This means feeding the spot markets and selling off options to take delivery of chips.

THE PC MAKERS

When, in early 1995, DRAM memory chip prices on the spot market plummeted, the brokers knew exactly what had happened. The top tier of PC suppliers – those that buy under large contracts directly from the big manufacturers – dumped their stock onto the spot markets after Christmas. They wanted to switch to new types of extended data out (EDO) memory, and flooded the market with the older fast page memory. PC producers offloaded their DRAM inventory onto the spot market and prices tumbled, said Joe D’Elia of Dataquest. Brokers unanimously agreed, naming most of the major PC suppliers. What drives the price down is people like Compaq putting chips back into the market, says Campbell of Datrontech. Strangely, however, not one first-tier PC manufacturer (usually meaning a top ten supplier) contacted by CBR at press time would admit to buying or selling on the spot markets – they always buy directly or from official distributors, they said. Only one admitted – off the record – to using the open markets. This reticence is surprising in the face of the evidence that most PC manufacturers maintain a broker desk, sometimes with the power to speculate on the chip price movements as well as negotiate prices. But there are reasons: in order to avoid shortages, most of the first-tier suppliers negotiate to take a regular supply of processors and memory chips from the big suppliers. These contracts normally specify that the goods are not to be resold. So the broker’s depart ment should not exist, said one former procurement specialist with a major PC company. Another reason why they prefer to keep quiet about their dealings and for keeping a low profile on the spot markets is that part of the sub-text of the marketing message from the top-tier PC manufacturers is that they control their quality, and monitor their suppliers, more tightly than the smaller PC companies. In fact, suppliers building to ISO 9000 quality standards are supposed to vet their suppliers fo r quality – something which is not easily done when sourcing components over the phone from a broker.

DENIAL

However much they may deny it, dealing on the spot market has become essential for most PC makers, which consume around 60% of the world’s processor and DRAM chips. The independent market’s ability to source components at low cost, and absorb large volumes, means it has become just as important to the computer industry as it has to the semiconductor industry. Again, there are some clear economic reasons. The first is the emergence of the low-inventory manufacturing model. During the past four years, PC margins have fallen so sharply, and the pace of technology has moved so fast, that the risk of being caught holding aging stock has become too great financially. For this reason, the big suppliers have followed the example of smaller suppl iers and are moving towards just-in-time, or make-to-order, manufacturing. In theory, this means that no inventory is held at all – everything is built to order, and parts are bought when they are needed, and not before. Everything is ordered on a rolling-contract basis with a small group of component suppliers. Dell, for example, claims to be able to build and supply any PC within 17 days – and says it never resorts to buying or selling on the spot market. But while Dell’s claim is difficult to assess and depends on the definition of approved, (Kingston Technology claims to occasionally supply Dell – but its quality control might classify it as an official supplier), in practice most of the make-to-order manufacturers not only buy on the spot market, but they have come to rely on its size and efficiency as a form of insurance. This insurance is vital. As the struggles of IBM have shown, the make to order model is extremely difficult to implement. Parts must always arrive at the right time and must be the right quantity, even during times of shortage or when there are unpredictable surges in demand. At Dell’s manufacturing plants in the US in Houston, and in Ireland, it gets chip deliveries several times a day; so too does IBM at its plants in North Carolina and Scotland.This is demanding for the supplier; and if the big PC makers are to be guaranteed this kind of service, they may have to pay heavily. Contract prices, which were once below the spot price because the me mory manufacturers wanted to reward regular customers, are now usually higher. This differential means that those suppliers which are prepared to buy more cheaply on the spot market for immediate delivery may have a crucial price advantage in the low margin PC business; memory and processor together typically account for 40% of the component costs of a PC. So a PC with components costing $1,500 might cost $80 to $100 less.This equation is making it increasingly attractive for PC manufacturers to organize final assembly points near the customer, and, moreover, for retailers to begin to build their locally produced own-brand PCs. The later in the cycle that the memory chips and processor are installed, the more likely it is that prices wil l be lower. A plentiful supply of chips, and the continuing news of new DRAM plants being built, has reduced fears of another global memory shortage.The same sums have encouraged the emergence of the grey market runners across the borders of Eastern Europe. These traders are exploiting the poor logistics in the East which means that pre-ordered components can take weeks to arrive, by which time they are often over-priced. They scour Western Europe for cheap chips – often working on commission from manufacturers and rarely asking questions about the origin of the components – and courier them back to the East. To further ensure competitiveness, the chip-couriers are often loosely affiliated to the wealthy individuals who hold tradable val ue-added-tax exemption certificates. The power of this arrangement can be seen by IBM’s decision to close down its Moscow assembly plant early in 1996, handing the rights to a local manufacturer which is prepared to enter into the grey market and ex ploit the tax anomalies. How are all of these changes affecting the world’s computer and semiconductor industries? The tail has started to wag the dog says one analyst. The spot markets, once a disreputable extension of big business, have changed the logistics of the industry and the nature of the trading relationships between companies. Several trends are already becoming apparent. First, the most simple changes will be seen in the way the industry measures itself. The book-to-bill-ratio, for example, will be complemented by other measurements that are more akin to stock market measurements: commodity part prices, volume of trade, even financial derivative indicators, which will themselves be sensitive to other market news. Second, el ectronic trading of component-commodities, perhaps led by bigger brokers such as NECX, will become the norm, making it more difficult for smaller traders but increasing the overall role of the spot markets. Equipment manufacturers, in spite of their contracts with big chip suppliers, will increasingly deal in this market, opening up and proudly displaying their ‘broker desks’ as if they were corporate treasury departments. This will further flatten out the traditional cycle of glut and shortage, with parts becoming available on the spot market in ever greater volumes.Third, the more ‘liquid’ market in certain key components will accelerate the trend towards the manufacture of more standard components (whose prices will then fall further) and systems. There are already signs of this: almost all of the big investments in manufacturing plants have concentrated on the most popular specifications for just three families of components in high demand: DRAM for PCs, Flash memory, mostly for portables and mobile phones; and S-RAM, for computer cache memory.

FRUSTRATION

The trend towards standardisation will carry through into computer system design, where the Intel microprocessor is, for a variety of reasons, already starting to dominate in servers as well as PCs. But even in Unix workstations, for example, manufacturing are starting to abandon specialist memory components in favor of industry standard SIMMS. A final trend will could frustrate the PC manufacturers seeking global dominance. The spot market for components will broaden out to cover more parts, such as motherboards, encouraging more and more final assembly to be done locally and moving the value-added function in the supply chain towards the end user. The spot market, a red-blooded capitalist phenomenon, will help to spread the wealth of the computer industry around.

SEMICONDUCTOR FUTURES?

In late December 1995 many South Koreans were worrying about North Korean troop excursions. In Taiwan, the continued aggression of the Chinese was a worry. But for the big memory maker executives, it was an exciting time for other reasons. In Seoul, Kwang-Ho Kim, chief executive of Samsung Electronics, the Korean giant, completed months of secret negotiations, finally clinching a hatful of deals which had even the restrained Korean executives ready to punch the air. Innovative financial engine ering was one of the keys to the deal, which helped seal the future for Samsung’s memory plants in South Korea, and planned new plants in the UK and Austin, Texas. Under the terms of the deal, revealed through Seoul news agencies, Samsung would supply memory chips to IBM, Texas Instruments, HP, Apple, Compaq and Sun Microsystems until 2000. The deal was potentially worth $65 billion. For Samsung, which already had semiconductor sales of $8 billion a year, the deal was a major coup, even if muc h of the business would eventually have come its way especially because memory prices plummeted a few months later. But there was a difference between these, and past deals. These new agreements last five years, instead of one. And, according to sources, the prices will be more flexible, linked to underlying spot market prices at the time of delivery. In effect, the six manufacturers have bought an option to take a proportion of Samsung’s capacity at a future price. It is not clear if the options on Samsung’s capacity are tradable – if one supplier does not want to buy the chips, it will have the right to sell to someone else. But this is the way forward. Earlier in 1995, Taiwan Semiconductor Manufacturing (TSM), the world’s biggest cont ract chip maker, began to offer its established customers tradable options on its manufacturing capacity. TSM took advice from US financial firms before drawing up contracts – details of which have not been revealed. It is believed the paperwork acts like an options contract – it can be traded at any price, with anyone, and TSM will honor it. How long before such contracts are traded, along with chips on the spot market? Or before large investments are hedged against semiconductor derivatives, based on a pricing index? Analysts are not certain it will happen, (an attempt to trade derivatives in the 1980s failed), but all agree that the spot market has become larger, more dynamic and bolder than anyone expected. It is only a matter of time before the speculation becomes more ambitious.

BOOK-TO-BILL BURNED OUT?

In late May 1996, the price of dynamic random access memory (DRAM) chips took a sudden upturn, sending a ripple of excitement through the spot-market, and ending a run of almost six months of unbroken falls. The stock price of the big memory manufacturers, however, barely registered. Few analysts watch the prices of semiconductors, preferring instead to react to the influential monthly Book-to-Bill figure published monthly by World Semiconductor Trade Statistics (WSTS) on behalf of the Semicon ductor Industry Association.The influence of the Book-to-Bill ratio can be tracked by mapping the effect of a good or bad monthly figure on to the stocks of the major semiconductor and PC companies, which consume about two thirds of the world’s processor and memory chips. Just like any set of government figures – inflation or public sector borrowing, for example – the Book-to-Bill is taken as a headline, market leading indicator, and as such it has a clear impact on stock prices and market expectations. In March this year, for example, the Book-to-Bill fell to a low 0.9 — suggesting orders for the next three months are 10% down on the last three months. Shares in Samsung Electronics Co, as the leading memory manufacturer, fell 5.9%. In contrast, the memory price changes have had little impact on stocks – even though these directly affect revenues. Texas Instruments chief financial officer Bill Aylesworth, for example, told analysts recently that Lower prices will mean lower revenues and therefore lower profits in the near term. The Book-to-Bill measures the amount of chips ordered (in dollars) against the amount being billed. It is a three month rolling figure, and in the crude logic of the financial analyst, is the most accurate way of measuring whether demand is rising or falling. And if demand is rising, they believe that it means PC sales are going up. But there is a problem – or rather a host of problems – with the Book-to-Bill figure which, ultimately, must reduce its importance to the high-tech industries and to stock market analysts. Just like government figures, the ratio is usually restated a month later, mainly because major manufacturers – especially far eastern ones which are increasingly importa t – fail to provide the necessary data on time. But that is only operational. The real problem is that the semiconductor industry has changed radically from a contract-only model, where demand from manufacturers is fulfiled by ordering goods from a first line supplier, to a spot-market commodities model, where fluctuations in demand are met by buying or selling on the open market. Although there is little hard data available, some analysts believe that as much as 40% of the world’s semiconductors are sold through independent channels and the spot markets. In addition, a small but significant proportion of those sold to first line manufacturers are later sold, unused, back into the market. There are several reasons for the rising importance of independent channels, but one stands out: the cost of the sub-micron level IC manufacturing equipment is so high that chip makers cannot afford to run below capacity. They must produce as much as they can and then sell excess stock onto the sp ot markets. Because these components may be bought more cheaply than those ordered in advance, computer manufacturers with growing demands may nevertheless actually order fewer chips, not more.This renders to the Book-to-bill ratio almost meaningless as a short term, leading indicator, which is how it is treated in world stockmarkets. Nor does the ratio serve its original purpose particularly well: It was initially conceived as a feedback figure, enabling manufacturers to make investment decis ions based on demand. To assess short-term demand (and hence quarterly revenues), spot market prices may be a better indicator. And as a guide to making major investment decisions, the long-term projections using an array of more detailed, analytica l data, available from the WSTS and from organizations such as Dataquest, are likely to prove more useful.