A Strategy for The United States to Regain its Position in Semiconductor Manufacturing
The United States is awakening to the remarkable diminution of its capacity to manufacture semiconductors, an industry that was invented and established in the U.S. Why do we care? Today our economy, our military, and America’s political system depends upon information. As a nation we live by the written word, as well as audio-visual information, gathered, stored, analyzed, and displayed in real time. We live in a world of instant communication everywhere including outer space. We demand it, we need it, and above all we expect it. Not only is the demand for information in all its forms insatiable; without even knowing it, we have grown to expect that the availability of that information and its functional value and purpose will grow exponentially.
In the mid-1990s few had any idea about smart phones. Now you cannot do without them. They are the real personal computers of today, with over 100,000 times the computer processing power and millions of times more memory than that available in the Apollo 11 computer used to get us to the moon. Software programs provide financial analysis, mathematical calculations, games, and search engines. Today’s software programs drive cars, fly drones, guide missiles, and in time will be expected to recreate the world as we know it with advancements in artificial intelligence. None of this is possible without a semiconductor.
Without that industry, the United States as we know it, would become nonfunctional, as would most of the rest of the world. Semiconductors are devices so small that the human eye cannot observe their functionality, yet they are the engine that provides the functionality for almost any electronic product. If the United States loses its competitive edge in the semiconductor industry relative to its competition it will become subservient to that competitor: economically, politically, and certainly, militarily.
Rebuilding the U.S. Semiconductor Industry
In the beginning the U.S. represented 100 percent of the manufacturing capacity of semiconductors. Today that capacity is now 8 percent. What does it take to rebuild our position? Lots of money, for sure. Much more than most people realize. In 1983 a fabrication plant for the manufacture of semiconductors could be built for $50 million. Today, 50 years later a fabrication facility for advanced semiconductor chips can cost between $20-30 billion. In the 1980s a piece of photo lithography equipment designed to put a pattern for a semiconductor part on a silicon wafer could cost about $450,000. Today, for the most advanced chips, the latest lithography equipment may cost $400 million.
The CHIPS Act, recently enacted, committed $52 billion spread over five years to support investment in research and development as well as additional capital investment in the semiconductor industry. The industry was offered the incentive with strings attached including rules relating to control of intellectual property and what products can or cannot be shipped to certain customers. The semiconductor equipment and fabrication industry has taken issue with some of these rules. But that is not the most important problem. The $52 billion pales in the face of the amount of funding required to be competitive in the U.S. semiconductor industry and that investment is climbing exponentially every year. TSMC in Taiwan, by far the largest and most advanced semiconductor manufacturer in the world today, spends nearly $40 billion on capital equipment and research and development a year by themselves. TSMC now represents 80-90 percent of advanced semiconductor manufacturing.
Thus, there remains a fundamental question which nobody seems to address. Why did we lose such a significant part of an industry that was invented in the United States, developed In the United States, dominated by the United States and fundamental to the long-term future of the United States? The answer is both simple and highly complex.
The U.S. willingly, eagerly, in its search for short-term profits, exited the customer base of electronic products that supported the semiconductor industry. And now the industry is faced with a question: who wants to invest billions of dollars in an industry whose customer base is outside the country and can get their semiconductors from more entrenched and sophisticated competitors located near to the markets they serve? And what made up that customer base that we lost: electronic products in general, and, specifically, consumer electronics. All electronic products are functionally dependent on semiconductors. In the main, these electronic products are no longer fabricated in the U.S. Just how did this happen?
How Did This Happen?
The loss of the VCR was a pivotal issue in the loss of the American consumer electronics industry.
On the wall as you entered the consumer electronics show in Las Vegas this year, was a picture of the first Ampex Video Tape recorder demonstrated at the National Association of Broadcasters convention in 1956. No one had ever seen a demonstration before of a broadcast quality video, available literally as it was taken. In that same year, 20 minutes from Ampex Corporation’s front door, in what is now called Silicon Valley, William Shockley, whose team had developed the first transistor at Bell Labs in 1948, established Shockley Semiconductor. In 1952, along with several other companies, Akio Morita, founder of Sony Corp. purchased a license from Bell Labs to manufacture semiconductors, for $25,000. Ampex did not. In 1970, Ampex introduced the VCR (five years before the introduction of the VHS format in Japan). Ampex owned most of the patents for video recording technology. Yet the semiconductor (not video tape) was about to change the form, fit and function of the VCR. In time, it was Sony who would show Ampex how to incorporate semiconductors in their products in exchange for certain licenses to produce video tape recorders.
Ampex dropped their VCR project in 1972 because of severe financial problems. At that time, their Board of Directors had no idea what that decision would mean to Ampex, let alone the United States. They were trying to save the Company from going bankrupt. In 1970 Ampex had entered into a manufacturing agreement with Toshiba in Japan, to make the VCR. Ampex did not, by themselves, have the financial resources or expertise to produce the video recorder in volume. When Ampex cancelled the project, though Toshiba lost that relationship with Ampex, they did not lose what they had learned when preparing to manufacture the product.
In the 1970s the United States dominated 98 percent of its domestic market for semiconductors and about 70 percent of the international market. Prior to the introduction of the VCR the total market for video recorders was in the thousands of units. By the mid-1980s, VCR production in Japan exceeded 50 million units per year. At its peak, almost every home in America had a VCR. Virtually none were made in the United States. Most were made in Japan. The VCR was called an integrated circuit (IC) hog. It used more semiconductors than any other electronic product up to that time. The bulk of those semiconductors would be produced in Japan.
The VCR became the most successful electronics product ever introduced up to that time. More importantly, every VCR needed a television set, and in many instances a camera, all of which used semiconductor devices and related electronic components. The volume of VCR’s and related electronic products forced a major effort by the Japanese to automate their factories. It also infused their economy with a massive need for machine and sheet metal shops to support the consumer electronics demand for packaging and automation. This in turn required major investments in machine tools to support fabrication requirements.
What they learned from their use of semiconductors in consumer electronics allowed the Japanese to provide additional functional capabilities and features to other electronic products including automobiles and computers. All this activity caused the world to view the Japanese economy in those days as an economic miracle. By 1982, the U.S. share of the world’s semiconductor market fell to 51 percent. Japan’s share increased to 35 percent, up from 15 percent in 1972. By 1989, the U.S. share of the semiconductor market had dropped to 35 percent and Japan’s increased to 51 percent.
But There Was More
In 1965 two events occurred which upended the competitive position between the United States and Asia. First, the Boston Consulting Group produced a matrix designed to strategically improve the profitability of corporate business. Second, Gordon Moore postulated, based on experience at the corporation he helped found (Intel), that for the next decade we would be able to double the amount of transistors (computing power) on a given area of silicon at roughly the same cost every two years.
The BCG Matrix divided businesses (and industries) into four quadrants. Stars, Cash Cows, Dogs and Question Marks. If you were smart you put your money into the Stars-high growth, great margins, technical or other superiority, and limited competition. Cash Cows-limited growth, good competitive position, good margins. If you’re smart, you milk the Cash Cows and put the money in the Stars. Dogs-growth but low margins, excessive competition, relatively excessive investment. If you’re smart, you sell the Dogs and put the money in the Stars. And Question Marks—you need to decide where they fit.
In the 1960s America was still riding the wave of victory in the Second World War. It could add or drop (or so it thought) industries with little apparent damage to its competitive position. Making money was the name of the game. In 1965, U.S. GDP was eight times that of Japan. Competition from Japan was not to be feared. At that time, Japan, who was trying to rebuild a nation, saw massive market growth in consumer electronics if prices remained low.
They were correct. In the BCG matrix, consumer electronics became a dog industry. This concept became widely accepted in U.S. business schools and board rooms. The U.S. began to outsource to Japan, and ultimately the rest of Asia, components supporting consumer electronics and ultimately the entire end use product. The money went into the stars including software, establishing parameters and specifications for products to be outsourced, and services. In time, virtually all the major U.S. companies manufacturing consumer electronics like recorders, cameras, television systems, computers, phones and so on, exited those fabrication businesses, were sold, or closed their doors.
Based on Moore’s law, which has now gone on for nearly 60 years, everything changed. The static world of Stars, Cash Cows and Dogs became virtually obsolete the day it was announced. The power of semiconductors to process information grew exponentially. Thus, the electronic products that used those semiconductors grew in power and performance exponentially. Accordingly, their demand for more and more technically advanced semiconductors grew exponentially. The problem was that those products, either because of bad management, or bad strategy, were no longer made in the United States. These dog products had moved overseas. And with them went much of the semiconductor industry.
Contrary to the strategy implicit in the BCG concept—because of the exponential growth in technology embodied in the semiconductor—dogs often become stars. Just look at your land line and compare it to your smart phone. When that metamorphosis takes place and you want the star back, it is often too late. Why? Because the technological and capital investment required to manufacture these products goes up exponentially as well. At those times, the cost of reentry can appear so high that nobody wants to take the risk. In the case of Ampex, the technological advances in logic and memory semiconductors replaced the need for tape. By 2008, Ampex was bankrupt. And when Zenith Corporation sold its high-definition television technology to LG in South Korea in the 1990s for approximately $360 million, it was only a matter of time before a single production line using that technology to build HDTV systems would cost several billion dollars.
The result of this customer base moving offshore is now apparent. There are basically no more television sets and displays made in the United States. And while this was happening, as we lost those electronic products that depended on the machine tool industry for packaging, between 1982 and 1987 half the U.S. machine tool industry closed its doors. And as to semiconductors, 83 percent of total manufacturing capacity for semiconductors now resides in South Korea, Taiwan, China, and Japan as does most all of consumer electronics manufacturing and that of electronics in general. The U.S. share of the world’s semiconductor fabrication capacity is now 8 percent.
So, What Do We Do Now?
There are three legs of the stool that support the technological development and manufacture of semiconductors. 1. The semiconductor equipment industry 2. Micro architecture planning and system level design and 3. The end use market.
Contrary to the demise of the machine tool industry, the U.S. position in the semiconductor equipment industry is still intact. Three of the five principal equipment companies, Applied Materials, Lam Research, and KLA reside in Silicon Valley, and are flourishing. Why? Because of Moore’s law. The exponential growth in the technological complexity, and power of semiconductors is made possible by the equipment companies. Though the unit volume of semiconductor manufacturing equipment is relatively small, the equipment is very expensive. The cost of a single piece of equipment can run into millions of dollars. In the case of lithography, hundreds of millions. Moreover, the technological requirements placed on the design and production of semiconductor equipment have rapidly exceeded the technological demands of virtually every other industry. Recognizing this, as compared to the machine tool industry, outsourcing of the semiconductor equipment industry was and is very difficult. Moreover, because of the technological lead this industry has developed, their end use markets have stayed intact regardless of location. That will not always be the case—other nations are trying to build their own semiconductor equipment. Were the U.S. to lose its way in the semiconductor equipment market, it would probably be game over. But today, the United States is still in a strong position.
Regarding micro architecture design and system level planning: the U.S. is still in relatively good shape in determining and laying out designs to meet the specifications required to achieve appropriate chip performance. The major problem is in the end use markets. It is essential that we onshore some of the manufacturing of those electronic products that we have lost. This will help ensure a reasonable domestic market in support of semiconductor manufacturing. In addition, the process of onshoring the design and manufacturing of those electronic products will provide additional insight into what the future of chip design should be.
When a product is manufactured, the process provides great insight into what can be done to improve its manufacturability, its reliability and over time, its performance. This information is vital to the company manufacturing the equipment that is used to produce the product, as well as to those who are engineering its design and product specifications. Manufacturing a product advances the expertise required to make the product economically and at scale. As such, the manufacturing process is fundamental in the development of proprietary intellectual property. All of this helps determine the course of future semiconductor design. Onshoring will also help reestablish some of what we have lost in the machine tool industry as well as laying the groundwork for the reeducation of the workforce necessary to rebuild manufacturing. It can be done. Tesla did it.
In the case of Tesla, it was common parlance that no one had tried successfully to enter the automobile business in the U.S. in 100 years. So, to do that was at least considered foolish. But Tesla did. Tesla set out to change the manufacturing process to support automation never seen before in the industry. The United States was basically out of the battery business, but Tesla was producing electric vehicles. So, Tesla, in joint venture with Matsushita, set up a giga factory to build batteries in the U.S. It had to design the tools required to make the parts suited for its concept of automation, so it did. Tesla now has a market capitalization over three times that of all other U.S. car manufacturers combined. Moreover, the company, based on its automotive concepts, has developed highly proprietary AI capability, robotics capability and ancillary businesses such solar technology and related use of battery technology. Yet, because of TSMC’s foundry capability, Tesla still must get their chips from Asia.
In this regard, there have been individual operational problems as well. As electronic products using semiconductors grew in complexity and functionality, the markets for and sales of those products grew exponentially too. As such, the companies marketing those products expanded in lock step. As they grew, their management strived to control the functionality of their products—often to fit into an ecosystem that was forming around them. Apple is a good example. TSMC in Taiwan was formed to meet this need. They created a foundry that would accept architectural and system designs from client firms. They produced semiconductors that fit customer needs and supported the specific functionality and proprietary features of the end use product.
There are only three major semiconductor manufacturers of logic chips: TSMC, Samsung and Intel. TSMC and Samsung offer foundry services. Only now is Intel planning to provide this service. This lack of a major chip manufacturer in the U.S. providing the services of a foundry has turned out to limit the nation’s competitive position, especially in advanced logic chips. It goes without saying that without Intel the United States would be in real trouble. Lacking a foundry, as the years progressed, Intel’s customer base was weakened as was its ability to keep up with technology.
But, the point remains, bringing back America’s industrial base can be done. It must be done, or we will continue to slide downward as a competitive nation, ultimately losing control of our future. Starting with the President and the leaders of our economy, in and out of business, we need to think strategically as a nation. Because of the semiconductor and its ability to collect, process and distribute information, everything is becoming connected. Some products in this kind of environment become strategic—like displays, medical instrumentation, telecommunication products and computers. As a nation, we need to be sure that the U.S. has control of and can depend on its strategic products and markets. This requires money to be sure, but it also depends on tax structure, antitrust laws, and the determination, as in the case of Tesla, to be willing to take the long view. Otherwise, our history as a democracy may be short.
Fundamental to this paper are two key points: A nation’s lead in technology and therefore demand for its products comes principally from its ability to advance and commercialize that technology faster and more effectively than its competition. Secondly, the end use product provides the market for semiconductors, not the other way around. If you lose access to the end use market the cost of reestablishing that relationship will often require that the resources of the nation must be harnessed rather than relying on a specific government outlay. With this in mind: strategy is everything.
Richard Elkus, Jr., is an international councillor at the Center for Strategic and International Studies in Washington, D.C. Elkus wrote the product plan and introduced the Ampex VCR in September 1970. He was also a founder of KLA and director of Lam Research.