Chapter 4: The ecological chain of the computer industry

From: http://www.google.com.hk/ggblog/googlechinablog/2007/10/blog-post_9934.html

 

The entire information technologies (IT) industry involves many fields and links, and these links are interrelated. Like everything in the world, the IT industry is constantly changing and developing and has its own development rules. These rules are summarized by people in the IT field into some theorem called it laws ). We will introduce these theorems several times based on specific examples. In this chapter, we will introduce Moore's theorem, Andy-Bill's theorem, and antimoore's theorem. Together, these three theorems describe the law of the computer industry, the most important component of the IT industry.

1. Moore's Law)

There are many stories about Bill Gates circulating in the technology industry, one of which is a conversation between him and the boss of General Motors. Gates said that if the automobile industry can develop like the computer industry, today, it only takes us $25 to buy a car, and one litre of gasoline can run four hundred kilometers. The story tells us that the computer and IT industry are growing much faster than traditional industries.

The first thing we saw was Dr. Gordon Moore, the founder of intel. As early as 1965, he proposed that the integration of integrated circuits would double every two years in at least ten years. Later, we shortened the cycle to 18 months. Now, the performance of computer and other IT products will double every 18 months; or the price of computer and other IT products with the same performance will be halved every 18 months. Although this development speed is incredible, the IT industry has always followed the speed of Moore's theorem prediction for decades.

In 1945, the world's first electronic computer ENIAC was able to complete the addition and subtraction of 5000 fixed points in one second. This 30-meter-long, two-meter-tall giant, weighs 27 tons, consumes 0.15 million watts of power. Today, Intel Core's PC computing speed is 50 billion floating point operations per second, at least 10 million times that of ENIAC, making it much easier to compare the volume of power consumption. In today's (June 2007) world's fastest computer, IBM's Blue Gene (bluegene/L), the speed is as high as 367 trillion floating point operations per second, 73.4 billion times that of ENIAC, it is exactly doubling every 20 months, which is roughly the same as the prediction of Moore's theorem. Even with the speed of computers, storage capacity is growing faster, doubling around every 15 months. In, Apple's computer's floppy disk drive capacity was kb, saving around 80 pages of medium files. Today, the hard disk capacity of a desktop PC at the same price can reach 500 GB, 3 million times that of an Apple Computer at that time. It can store all the text in the library of Peking University. Moreover, over the past decade, the speed of network propagation has almost increased by Moore's theorem. Thirteen years ago, I had the honor to become the first online user in China. At that time, I went to the Stanford Linear acceleration lab through the Institute of high energy physics to connect a dedicated line with the Internet, at that time, the telephone modem speed was 2.4 kb. It would take eight hours to download the Google Pinyin input method. Now, the commercial ADSL can achieve a 10 m transmission rate through the same telephone line, four thousand times that of thirteen years ago, almost doubling every year. It takes about 10 seconds to download Google pinyin. Among the top five industries in the world economy, namely finance, information technology (IT), medical and pharmaceutical, energy and consumer goods, only one IT industry can continue to double.

People often doubt how many years Moore's theorem can apply. Even Moore himself initially thought that the IT field could develop at such a high speed for ten years. In fact, since the Second World War, technological advances in the IT field have been doubling every one to two years, and there are no signs of stopping. In the history of human civilization, no other industry has done this. Therefore, the IT industry must have its own particularity.

compared with any other product, the raw materials required for manufacturing IT products are very small and the cost is almost zero. Taking the semiconductor industry as an example, an Intel Core dual-core processor integrates 0.29 billion transistors, And the Intel 8086 processor was only 30 thousand transistors thirty years ago. Although the integration between the two is nearly 10 thousand times different, the raw materials consumed are not much different. The manufacturing cost of hardware in the IT industry is mainly the cost of manufacturing equipment. According to Applied Materials, a semiconductor equipment manufacturer, a production line capable of producing 65-nanometer core dual-core chips with a total investment of $20 to $4 billion. Last year, Intel's R & D cost was USD 6 billion. Of course, we cannot count it all on core, but intel may not be able to develop a product like core on average every year, so its R & D cost should be equivalent to Intel's one-year budget. If we split these two costs into the first 0.1 billion core processors, each of which costs nearly $ one hundred on average. In this way, the core processor can be greatly reduced after intel recovers the production line and R & D costs. Last year, Intel sold around 0.2 billion processors, so a new processor would not take more than a year and a half to recover costs. Generally, users can see that the price of a new processor will be lowered significantly after one and a half years of release. Of course, Intel's new products are now under development.

Moore's theorem leads the development of the IT industry. First, in order to enable Moore's Theorem to be established, IT companies must complete the development of next-generation products in a short time. This requires that IT companies must invest a lot of money in R & D, so that there will not be too many competitors in each product market. In the United States, most of the major IT markets have only one competitor and one competitor. For example, for computer processor chips, only intel and AMD; for high-end systems and services, only IBM and sun; for personal computers, HP and Dell (DELL is a bit strange, is the R & D investment large ?). Second, with strong hardware support, applications that were previously hard to think about will continue to emerge. For example, two decades ago, the computation of digitization of high definition movies (1920x1080 resolution) was not sufficient for IBM's mainframe; now, a laptop-sized Sony game machine could do that. This creates conditions for the birth of some emerging companies. For example, ten years ago, no one wanted to create a company like YouTube, because at that time, the speed of the network could not meet the Video watching requirements on the Internet. Now, YouTube has been integrated into the lives of ordinary people. Similarly, the current R & D must target the market after many years. Let's look at it for ten years. If I propose that the speed of surfing the internet will increase by one thousand times after ten years, some people may think that I am crazy. In fact, this is a fully achievable goal. If this is done, each of our families can simultaneously play three high-definition, surround-sound movies and watch them on three different televisions. You can also fast forward and jump to the next chapter at any time. After stopping at any time, you can continue watching it next time. When watching three movies, we can store our photos, videos, files, and other information on an online server, and access from home is as fast as on our own machine. This is not my own fantasy, but an ip TV plan implemented by companies such as Cisco and Microsoft. Once again, existing IT companies must have a way to eliminate the unfavorable factors brought about by Moore's theorem, that is, the price is halved every 18 months. We will discuss this in the next two sections.

2. Andy and Bill's Law)

Moore's theorem brings a hope to all computer consumers. If I think computers are too expensive to afford today, I will be able to buy them at half the price in 18 months. If it is so simple, the sales of computers will not be able to go up. People who need to buy computers will wait for several more months. People who already have computers do not have the motivation to update their computers. This is also true for other IT products.

In fact, the sales of personal computers in the world have continued to grow over the past two decades. In 2004, Intel estimated that sales of PCs (including personal machines and small servers) in the world would increase by 2009 in five years, which is far higher than the economic growth by 60%. So what drives people to constantly update their hardware? The IT Community summarized it into the Andy-Bill theorem, that is, Bill wants to take what andy gave
Andy gives, Bill takes away ).

Andy is Andy Grove, CEO of Intel, and Bill Gates, founder of Microsoft. In the past two decades, Intel's processor speed has doubled every 18 months, and the computer memory and hard disk capacity have increased at a faster rate. However, Microsoft's operating systems and other applications are getting slower and slower. Therefore, although the current computer is one hundred times faster than a decade ago, the running of software is still similar to the previous one. In addition, in the past, the entire Windows operating system was only a few megabytes in size, and it now takes several gigabytes, so is the application software. Although the new software features a little better than the previous version, the added features are definitely not proportional to its size. Therefore, how many applications can a computer install ten years ago?Program, But now it is not installed so much, although the disk capacity has increased by one thousand times. What's worse, users find that if they do not update their computers, many new software will not be available now, and connection to the Internet is also a problem. A car that can afford it a decade ago can still run.

at first glance, Microsoft is doing the right thing with everyone. In fact, gates and other vendors do not want to make the operating system and applications so big. It is understood that Gates himself said many times that he used to build basic with only a few dozen K jobs. You (Microsoft engineers) have to build a. net job with hundreds of megabytes, which can be optimized. Of course, we know that Microsoft's current. NET is much better than its basic functions two decades ago, but is it 10 thousand times better? I'm afraid no one thinks so. This shows that software developers are no longer as careful as they were two decades ago. We know that the basic interpreter was written in assembly language, so it could not be refined, or it could not run on the early IBM-PC. However, software engineers are required to use assembly language programming, which is extremely inefficient and the written program is poorly readable and does not meet the requirements of software engineering. Today, with sufficient hardware resources, software engineers focus more on their work efficiency and program standardization and readability. In addition, due to the increase in labor costs, in order to save the time for software engineers to write and tune the program, the programming language is getting better and better, while the efficiency is getting lower and lower. For example, today's Java is much less efficient than C ++, and C ++ is less efficient than C ++ two decades ago. Therefore, even for software with the same function, it is inevitable that today's hardware resources are more occupied than yesterday's.

Although users are very worried about the benefits of hardware upgrades brought about by new software, in the IT field, hardware vendors rely on software developers to use their hardware resources to survive. For example, as of the first half of last year, Microsoft's new operating system Vista was delayed, from Intel to devices such as Hewlett-Packard and Dell, to peripherals such as Marvell and Seagate, all sales are greatly affected because users do not need to update computers. The shares of these companies have fallen by 20% to 40% to varying degrees. At the end of last year, Microsoft finally went public on Vista, and of course Microsoft's own performance and shares were immediately improved. Intel, which had been depressed for more than a year, also reversed its decline at the beginning of this year, of course, both HP and Dell are also growing. This year, the shares of the three companies have all risen sharply. In the next few days, hard drive, memory, and other computer chip manufacturers are about to recover. Compared with the previous version of XP, Vista also provides many 20% features, but its memory usage is almost doubled, and the CPU usage doubles, unless it is a new machine, otherwise, Vista cannot be run. Of course, users can choose to use the original operating system XP, but soon, Microsoft and other software development vendors gradually reduced support for the XP system, thus forcing users to update machines.

We can see that the entire ecological chain of the personal computer industry is like this: software developers headed by Microsoft eat all the benefits of hardware upgrades, forcing users to update machines to make profits for companies like HP and Dell, these machine manufacturers then order new chips from half-guide manufacturers such as Intel, and purchase new peripherals from other factories such as seagat E. Among them, the profits of various companies have been correspondingly increased, and the stock has also increased. Each hardware semiconductor and peripheral company will invest in R & D of profits, speed up according to Moore's theorem, improve hardware performance, and prepare Microsoft for the next step to update software and eat hardware performance. Wall Street investors know that if Microsoft's development speed is slower than expected, and its software performance is poor, it will not be able to buy shares of companies such as Intel.

For users, buying a usable computer is similar to buying a usable computer ten years ago. If it is not affected by the "Made in China" effect, it will also increase slightly due to inflation. (Should this sentence be like this: if it is not the impact of the "Made in China" effect, for users, they will buy a usable computer and a usable computer ten years ago, the amount of money spent is almost the same, and even increases slightly due to inflation. The original meaning is that there is no difference with the original, because I feel cheap. The modified result is made in China and is now cheaper than the original one .) Of course, while Microsoft and other software developers eat most of the benefits of hardware improvement, they will more or less bring new things to users.

If the information revolution that began in the United States 20 years ago was based on personal computers and the Internet, then in Asia, mobile phones and mobile communications are the mainstream. Today, mobile phones generally have two processors: A Digital Signal Processor (DSP) and a general-purpose processor (CPU) similar to a microcomputer processor. Today, the computing performance of a mid-range mobile phone is as follows, personal computers more than five years ago, and the speed expected by the Moore's theorem is growing. Although no general operating system company similar to Microsoft exists in the mobile phone industry, the mobile phone manufacturers themselves, operators, and value-added service providers play Microsoft's role together. They are providing new services that consume more and more resources, and users have to update their mobile phones several years ago.

In this way, Andy-Bill's theorem turns computers, mobile phones, and other commodities that originally belonged to durable consumer goods into consumable goods, stimulating the development of the entire IT field.

3. Reverse Moore's Law)

Eric Schmidt, CEO of Google, pointed out in an interview that if you look at Moore's Theorem in turn, if an IT company sells the same products today and 18 months ago, its turnover will be halved. It calls it the antimoore theorem. The inverse Moore theorem is unfortunate for all IT companies, because an IT company spent the same effort, but only half of the previous revenue. The inverse Moore theorem forces all hardware equipment companies to keep up with the update speed specified by Moore's theorem. In fact, all hardware and equipment production plants work very hard. The following table lists the shares of the largest companies in various fields today and their maximum proportions.

IBM: 82%.
Cisco: 1, 40%
Intel: 1, 33%
AMD: 30%
Marvels: 60%
HP: 70%
DELL: 1, 35%
Sun Microsystems: 10%
Motorola: 33%

In this regard, apart from IBM's not just hardware vendors, there are strong service and software revenues that can keep stocks at a high level, and other companies are far from their best levels. Today, the U.S. stock market is almost the highest point in history. This shows that hardware-based companies are difficult to make a living because of the influence of the inverse Moore theorem. If you are interested in reading the financial reports of these companies, you will find that these companies are developing at a high volatility. Once they fail to achieve the development speed specified by Moore's theorem, their profitability will plummet. Some companies may even have a disaster tolerance, such as SGI, which was booming 10 years ago. Even if they develop well today, they cannot guarantee that they will continue to make double progress in ten years. Therefore, investment guru Buffett never invests in these IT companies.

In fact, the positive side of the inverse Moore theorem is even more important. It facilitates qualitative advances in the field of science and technology and provides emerging companies with the possibility of survival and development. Like the development of all things, technological advances in the IT field are both quantitative and qualitative. For example, the same processor does not change much in the system architecture, but only improves the clock speed. This progress is the progress of quantitative change. When the number of processors increases from 16 to and then to, there is a small qualitative change. If nano-or biological technology can be used one day, a qualitative leap will be achieved, and the integration of Semiconductors will be improved by hundreds of times. In order to catch up with the growth rate of Moore's theorem prediction, quantitative variation alone is not enough. Every technology, within a few years, will be tapped into the potential of quantitative change. At this time, there must be a revolutionary creative invention.

In the process of technological advances, new small companies cannot compete with old large companies, because the latter has unparalleled advantages in old technologies. For example, Nokia, born from a wooden factory, cannot compete with Motorola, the boss of traditional communication equipment, on older analog phones. However, in seizing the opportunity for qualitative changes, some small companies will do better than large companies, because they have no burdens and are more flexible than large companies. This is why many new technology companies have emerged in Silicon Valley.

Thirteen years ago, I first used a 2.4 kbps modem to access the Internet. Two years later, one of my colleagues, the founder of Dongfang Wangjing, the earliest Internet company in China, sent me a modem of the latest 14.4 kbps at that time, and I immediately felt that the speed was much faster. Because the digital telephone transmission rate is limited to 64 Kbps today, the modem transmission rate is up to 56 kbps, so by 1995, several of my colleagues predicted that the speed of accessing the Internet over a telephone line would not exceed this limit. If we stay at using the traditional method to speed up the modem, it does not take a few years for Moore's Theorem to apply. However, in 1990s, DSL technology was introduced to increase the data transmission speed on telephone lines by nearly two hundred times. Although DSL technology was first published by Bell's core laboratory, it was actually converted to practical technology by Professor John chafei of Stanford University. Professor Cha Fei became a senior member of IEEE (fellow) in his thirties and became an academician of the American Engineering Institute. In 1991, he started Amati, a small DSL Company, with several of his students. In 1997, he sold Amati to Texas Instruments (TI) at a high price of $0.4 billion ). This is a typical success story of Silicon Valley new technology companies. In the quantitative change phase of modem development, there will be no small companies like Amati, and even if they do, they will not be able to compete with Texas Instruments. However, once the modem speed is close to the original limit, emerging companies that can break this limit will have the opportunity to enter the stage of history.

The inverse Moore's theorem makes it impossible for the IT industry to pursue only quantitative changes like the oil industry or the aircraft manufacturing industry, and it must constantly seek revolutionary creative inventions. Any company that cannot meet Moore's theorem needs to develop technology will be eliminated in a few years. Large companies, in addition to maintaining high R & D investment, must always pay attention to the development of new technologies related to themselves and often buy small companies with revolutionary new technologies. They even make money to invest in small promising companies. Cisco is the most typical representative in this regard. It has bought many small companies that have invested in it over the past two decades.

At the same time, Moore's theorem makes it possible for new small companies to be at the same starting line as big companies in terms of developing new technologies. If a small company is successful, it can be acquired by a large company like Amati (this is a good thing for the founder, investors, and all employees ). They may even replace their status in their respective fields. For example, in terms of communication chip design, Broadcom and Marvell have largely replaced the original Lucent semiconductor department and even Intel's business in the relevant fields.

Of course, a company requires money, and no one can guarantee that investment in an emerging company will be profitable. Some investors who are willing to take risks and pursue high returns bring together the money and give it to professionals who know both financial management and technology, and give it to promising companies and individuals, this has gradually formed a risk investment mechanism in the United States. To build a high-tech company, you also need professionals who are both like-minded and willing to take risks. They are more interested in some companies than their relatively high salaries, so we have the option system for employees of high-tech companies.

Since its development, the IT industry has its own survival and development path. It is not shrinking because the price keeps falling, but is booming. We will introduce its development rules in the future.