|HOME||PUBLIC LIBRARY||ANOTHER PERSPECTIVE||INFOPERSPECTIVES||CONTACT|
The number of bitcoins in circulation is more than 12 million and growing. They have an aggregate value with an order of magnitude of $10 billion. Are bitcoins money? That depends in on how one defines money and where in the world one happens to be. In China it illegal for banks to trade bitcoins but in the USA bitcoins are kosher. Overstock.com sells IBM x servers for bitcoins. But there is no simple way to buy Power or mainframe servers from IBM with the bitcoins they resemble.
Briefly, a bitcoin is not a physical coin like a penny. It is a data record, defined by software, various conventions and the principles laid down by its pseudonymous inventor, Satoshi Nakamoto. In 2009 Nakamoto put in the public domain software, Bitcoin-Qt, that creates and manages bitcoins. Part of the management scheme is a public record, called the block chain. The block chain is a publicly available data file that includes the inception of every bitcoin and every transfer of ownership of that bitcoin subsequent to its creation. The software defines a process, bitcoin mining, which creates fewer and fewer bitcoins as time goes on. Mining is designed to end in 2140, after which no more new bitcoins will be created. When mining ceases, there will be 21 million bitcoins.
This constrained and ultimately limited supply stands in contrast to the population of computers, particularly if one defines computers broadly to include tablets, smartphones and gaming systems. And if one defines computers to encompass even more things, including everything that contains a system-on-a-chip (SOC) device, the population of computers becomes truly vast. Whether every smart gadget is included in one's count or not, the number of computers is growing at a stunning pace. What is true in general is not the case for every type of computer. Some are booming, some are disappearing, and some are going nowhere. Computers of any particular type have functional lives that end when they are no longer practical, often because better, smaller, less costly and otherwise more attractive computers take their place. Often, computers give way to improved but architecturally similar models, at least for a few generations. But few computer architectures survive a very long time. IBM, for example, has some classic product lines, notably the 64-bit mainframe and the Power/i family, that have persisted for quite a long time. But most IBM processor lines, like those of other vendors, lasted a while and then faded away. The extensive list of IBM's retired product lines includes the S/32, S38, 1800, 7094, 1400, plus the 24- and 31-bit ancestors of current 64-bit mainframes.
The aggregate market for computing devices resembles the market for fuel. One type of computer, like one type of fuel, may replace another, but demand for MIPS or kilowatt-hours (or BTUs) just keeps growing. The demand for computing power and the demand for thermal energy grow faster as costs come down, with the net result that improvements in efficiency lead to increases in consumption. Even if costs rise, which has not occurred in the case of commodity MIPS but which has been the case for KWH, technology that delivers more bang for the buck enhances the value provided by computers and reduces the impact of rising fuel costs. IBM has distinguished its proprietary computer business from the computing industry at large by positioning its servers as premium products and pricing them accordingly. By going its own way, or perhaps despite going its own way, the company has been able to succeeded on its own terms. But IBM's business depends on customers who stay with legacy software and retain, in both form and content, their long-established strategic databases. Shaped and sometimes limited by its unique practices, IBM has been unable to successfully participate in the high-volume markets that constitute the lion's share of the total computer industry. It seems to be able to get sales volume or profits but not both at once. Critics say IBM can only attain leadership in these markets by buying it. Admiring investors say IBM is at its best when it concentrates on premium products and accepts that it cannot succeed in high-volume segments. Comparisons of IBM and Apple, which has attained high unit sales volume while maintaining premium pricing for its trendsetting products, seem to get both critics and admirers all riled up.
Judgments aside, it is obvious that IBM servers don't incorporate the technologies that are driving sales of mobile devices, shaking up most of the server business and forcing data communications equipment companies to repeatedly replace their hardware platforms. IBM does not make or use state-of-the-art SOCs, the most popular of which currently use ARM architectures. IBM neither made the processor chips used in the PC business it long since sold nor the more powerful but architecturally similar CPUs that define the X86 server trade it is now leaving. It is possible that IBM, without skin in the X86 chip game and without its own X86 operating system, could not succeed the way it has in proprietary segments. None of the X86 client or server manufacturers seem to be able to offer truly distinguished products . . . except, again, for Apple, which uses X86 chips but adds its own microcode and systems software, among other things, to define what has become a very popular line of laptop and desktop systems. Apple does not appear to be interested in the server market.
IBM is still in the chip manufacturing business, although speculation about its eventual sale of that part of its empire is rampant. The company makes Power and mainframe chips, which are well-regarded by analysts and apparently appreciated by customers. Still, IBM's chip business is quite different from that of Intel, AMD and the myriad makers of ARM circuits. IBM's chip fab business is a boutique by the mass production standards of the industry, producing only the modest volume of circuits needed for p and z machines. IBM never had enough faith in its Power designs to put its name on a PC family, even though it did create chips (made by IBM and Motorola) that for several years powered Apple computers.
In IBM's proprietary server trade, unit volumes are measured in thousands, unit volumes in PCs and PC servers are measured in millions, unit volumes in mobile standalone devices are measured in tens of millions and unit volumes in the myriad embedded computers that manage automobiles, appliances, communications devices, medical apparatus and most everything one can think of that runs on electricity are measured in the hundreds of millions. ARM Holdings, which licenses its eponymous chip architecture to just about all the leading SOC producers, counts the fruits of its output in billions. IBM's business models don't seem to include operations that grow revenue and profit with soaring hardware production volumes. That is not the case at Apple, Samsung, Intel, HP and other companies that are in today's computing mainstream, nor is it the case in the product divisions of computer industry participants with larger interests outside the device business, such as Google, Microsoft and Amazon.
The computer manufacturing industry leaders also include companies that once were the nearly invisible assemblers and manufacturers of machines for others, typically Taiwanese of Chinese industrial enterprises that today seek to be the primary suppliers of servers to the true giants of the services world such as Google, EBay and Amazon.
IBM knows all this and apparently prefers to choose a path that takes it away from the computer business. But its success, and perhaps its very survival, depends on another business that will ultimately obey the same laws of economics that govern the markets in computer hardware and energy.
In fact, it was the situation in the key energy market of his time, coal, that inspired William Stanley Jevons, an academic born and raised in Liverpool, England, in the middle of the nineteenth century to describe what he saw was in some ways a paradoxical development. Jevons saw that dramatic improvements in steam engine technology and other advances in engineering were making it possible for much less coal to do much more work. When other commodities became less expensive, and in his England wool may have provided a concrete example, business interests shifted focus and sought new opportunities. But in the case of coal-based energy, improved value set the markets on fire. Demand rose even faster than prices fell, with the net result that demand for coal kept soaring as coal technology developed. One of Jevons's books, The Coal Question, published in 1865, helped the polymath shine brightly in the emerging constellation of scholars whose collective work would come to be known as macroeconomics. Jevons also participated in related work on the marginal effects of pricing and demand on production and consumption.
IBM knows all about macroeconomics. Even among the company's top executives, a crowd in which an MBA is cheerfully passed off as proof of educational attainment, lots of people must have been compelled to read Samuelson or one of its ilk. At the very least, the company's Dutch dodgers, who obviously have the ears of Big Blue's economics enchiladas, must know a bit about this stuff. If they don't, now might be a good time for them to learn. As they do, they might trip over some of the more amusing tales about Jevons, including his work building a machine that could solve problems in Boolean algebra that he called his logic piano. Also, a little inspiration about the value of invention couldn't hurt whatever passes for a brain trust inside IBM these days.
The computer industry has long since adjusted to goods becoming standardized and treated as commodities. Now this has begun to happen in services, despite the best efforts of every service provider to create differentiation that adds value. IBM has been able to distinguish its offerings from those of competitors until now, but its methods for monetizing and branding services that in fact closely resemble the offerings of others could soon lose effectiveness.
IBM's success in services was built on the bedrock of its prowess in manufacturing and its claim that no other service provider could understand how to use IBM systems better than their developer and manufacturer. Now that IBM is leaving manufacturing (or vice versa) in one product line after another, the bedrock is crumbling. IBM is still the sole source of mainframe and Power servers, but that asset could become a liability if customers start wondering whether IBM is in the server business for the long haul. If doubt becomes widespread, the services structure, built on the foundation of hardware manufacturing, is unlikely to stand.
IBM is suffering because it seems to have fallen far behind the trend toward processor hives using large numbers of small engines to perform tasks for large user communities on big data. Moreover, if IBM catches up with that trend it might discover it has caught only the trailing edge rather than the leading edge.
The envelope trend in computing - the evolution from large engines to smaller ones and on to tiny but impressively powerful SOCs and their ilk - could give way or at least cede ground to an alternative: server farms based on specialized processors X86, ARM, Power, Sparc and MIPS may be great, but they are simply not the best possible solution to every problem. Consequently, in addition to the companies that produce general purpose servers, there is a much smaller but very lively segment of the computer business that builds specialty chips and products. This specialty has cousins in the form of adept server makers that integrate these chips into machines that excel at certain types of mathematical functions.
Arguably, the most prominent members of this class are the manufacturers of graphics cards, such as Nvidia. Nvidia products are used in workstations, where they drive displays, but the same circuits are also very good at the vector calculations used in math and science. Nvidia has gone very far creating software and hardware that adapts its specialty engines for use in compute-intensive tasks. As a results, some very well regarded supercomputers depend on Nvidia chips (and the software that manages them) to produce results faster and more economically than could be obtained using general-purpose CPU circuits.
A related application that uses special chips is encryption, where application specific integrated circuits, ASICs, make it possible to rapidly and, of great importance, inexpensively encode and decode messages. Currently, there are chip developers creating ASICs that are used to perform the hash encryptions on which bitcoin mining is based.
Bitcoin mining does not involve physical mining as done in the hunt for gold. The creators of Bitcoin used the term mining to stand for a mathematical process that encodes each bitcoin transaction using a powerful one-way hash. This process, along with surrounding procedures and conventions, largely guarantees the integrity of the bitcoin system.
(This integrity has been breached in dramatic but relatively small ways, at least compared to the size of the total bitcoin market, recently. The bitcoin community is scrambling to remedy flaws in the bitcoin software, imperfections that have permitted malefactors to engage in double spending, something the bitcoin system is supposed to prevent.)
The ASICs are components that make bitcoin mining more economical, but they must be harnessed in special servers to produce useful results. These special servers include general purpose chips to do the processing that accompanies hashing in the mining process. This integration of ASICs with more-or-less standard X86 hardware is a growing part of a computing universe that seems to live in a parallel world not yet occupied by the established big name computer companies. It is a lively new branch of computing, and it seems to have an enthusiastic customer base.
Several emerging specialty server coprocessor makers, such as Butterfly, produce bitcoin mining apparatus that is in great demand. The only way to buy a mining machine, if a customer wants the latest model, is to pay in advance and sit on the waiting list of a perpetually backlogged manufacturer for weeks or months. Once built and delivered, this equipment is sometimes installed in users' data centers or ordinary server hotels, but the mining servers are sometimes found in exotic data centers located where power and cooling are extraordinarily cheap, such as up near the Arctic Circle in Iceland. Bitcoin is probably only one of the first of a bunch of synthetic moneys. The next one or the one after that - and we are inclined to think there will be quite a few of these launched during the next few years - may be defined to become a thousand times the size of bitcoin. If it could get to $10 trillion it would be large enough to make a real difference of some kind in the world, whether it succeeds or fails. And, if Big Blue got into the cryptocurrency game in a big way, creating its own currency, the Watson, and grossing, say, one percent off the top, it would be more or less doubling its intake. A prompt entry into the cryptocurrency universe might be quite timely for IBM. If Big Blue's services customers begin defecting to other vendors the way its product customers have, the mining business might be a way for IBM to continue profiting from all the servers it already owns and lights up. There might be an extra step along the way. In order to coin money at a satisfactory rate and at a satisfactory cost, IBM might have to add special coprocessors to its installed equipment base. It could, presumably, define its mining process and the special circuits to perform the encryption. This would give IBM considerable control over the money it invents.
IBM's next big money question might be, "How many bitcoins is this Watson worth?"
— Hesh Wiener February 2014