The global demand for smartphones will foreseeably fall during this second quarter, according to the Taiwanese consultancy Digitimes, mainly due to the lack of components by large manufacturers and delays in the delivery of materials due to the controls caused by the pandemic. The first quarter was exceptionally good, because it was compared to the Covid-19 boom in China a year before, but now the stocks accumulated last year have run out and it is more difficult to replace them, due to the breakdown of the logistics chain and the shortage global chip. In any case, the forecast of selling between 1.35 and 1.4 million smartphones this year is still maintained, similar to the global quantity dispatched during 2019, which would have to materialize in the second half of 2021.
One after another, products that consume a lot of semiconductors have been experiencing many supply problems for months, especially computers and tablets during the pandemic due to the increase in demand and since the autumn cars, when a totally lethargic demand began to wake up. . During the first quarter, the boom in smartphone sales, including the 5G models, which include many more chips than 4G and at a very reasonable price despite the novelty they represented, could be supplied without major problems.
On the one hand, the lower demand for smartphones throughout 2020 meant that there was an excess of finished products at the end of the fourth quarter, which could be dispatched during the past quarter. On the other hand, the smartphone manufacturing industry collects the necessary components well in advance and does not discuss prices so much with chip producers, because it requires products of the highest quality, which means that it has more resources to buy what it needs. .
When Morris Chang founded TSMC in 1987 to make custom chips, at a time when chip design and manufacturing were integrated, they took it for nothing short of crazy; now everyone is queuing and begging TSMC to make the chips they have designed
However, strong pressure from companies that have had to stop production for several weeks due to the lack of semiconductors, especially car manufacturers, has caused some deliveries to be accelerated compared to others and that those destined for smartphones stopped having the highest priority, even if they were the ones that provide the highest profit margin for suppliers.
A traditionally cyclical industry
Semiconductor manufacturing, especially memory, has always been very cyclical and supply has rarely been accompanied by demand: sometimes there has been a shortage of supply, with the consequent rise in the price of available products, and in other cases a shortage of demand, with prices down. To this is added that orders are made several months ahead, in anticipation of a demand for final products that is not always met, and causes strong imbalances in stocks.
The problem with memories, both the DRAM type used in CPUs and the NANDs or flash in most electronic products, is that their manufacture requires investment in production lines that cost several billion dollars. The start-up of the production lines, from the moment they are commissioned until the first chips with a reduced percentage of defects come out, takes place after several months or even years, because they are generally state-of-the-art equipment, which is very difficult to process. adjust.
To further complicate the situation, the semiconductor manufacturer does not order a new line until it anticipates that there will be sufficient demand, which has to be much higher after a few years because production is typically doubled or quadrupled with the new line. of production (for example, it is possible to go on to manufacture memories with double the capacity per unit of production; if on top, as now, 3D memories are made, of several floors, the production capacity increases exponentially). Demand must necessarily increase at the same rate.
It is reversed when the demand is higher
The availability of capital to invest has always been a capital issue. Typically, you invest in a new production line when demand is very high, because prices (and profit margins) are high and there is money to invest. But it can happen (and it is not uncommon for it to happen) that when the new line is prepared to manufacture a lot of product without defects there is not enough demand and it is necessary to sell at a low price.
Because the most common is that when a manufacturer has decided to invest in a new plant, its competitor or competitors have done the same, because all have seen the same opportunity. Production capacity, then, suddenly increases considerably and, even when demand grows, it is not uncommon for there to be excess supply. And it is more difficult to amortize the investment by the semiconductor manufacturer
The design of the chips, without an associated factory, has been revealed with the pandemic as a great lack of sovereignty and an excessive dependence that the United States, Europe, Japan and China want to correct; while South Korea and Taiwan seek to maintain their current leadership
Nor is it strange that the roadmaps are not fulfilled as planned at the beginning of the process, as is happening to Intel in recent years, when in its first four decades the production machinery of the veteran company worked like a Swiss watch. In addition, if suddenly someone decides to seize a large manufacturer and consumer of semiconductors, as Trump did with Huawei, and, on top of it, a pandemic like the current one has come, the strange thing would have been that the logistics chain had continued to function as if nothing had happened, especially considering that she was already extraordinarily stressed.
Marketing divorce with production
To finish complicating an already very uncertain panorama, it must be taken into account that the large marketing departments begin to sell the product long before it is available and even when it is in the design phase, with hundreds of evangelists who explain why everyone the excellences of the future product. Just remember when people started talking about 5G and its extraordinarily low latency of one millisecond, which has yet to be achieved.
Over the years, everyone has learned to distinguish marketing from things to come, and production chains have adjusted to the fluctuating nature of supply and demand and to touch the ground. But for some time now, the introduction cycles of new products have accelerated and, in addition, China, in addition to producing, has gone from buying to selling highly competitive products, leaving all the commercial, productive and now logistical gear without room for maneuver (and with the price of raw materials through the roof).
In this extraordinarily complex environment, a new factor has appeared, which has always been known to exist but has not been taken into account: sovereignty or technological independence. Half a century ago, when Japan began to make consumer electronics products very competitive against Europeans or Americans, Ronald Reagan imposed heavy tariffs to try to remedy the situation in the United States, which he failed. Half a century later, the same recipe has been tried with China, but raising trade tension to the maximum, without any evidence of success either.
Manufacturing the chips as well as designing them
Both the functional design of a chip and the physical manufacture of a wafer, made up of tens or hundreds of chips, are extremely complex tasks, requiring in the first case considerable human resources and, in the second, a great deal of human and material resources. higher, because it takes an extraordinarily powerful production infrastructure, several billion dollars and hundreds or thousands of people to make tiny chips, like the ones in the Samsung image below.
Morris Chang, now 90, was already working at Sylvania Semiconductor in the mid-1950s, in what was later Silicon Valley. What the forerunner Fairchild and later Intel, AMD, Motorola, Texas Instruments and a few others did more than half a century ago was design semiconductors and figure out how to make them. In the late 1950s, Chang went to work in Texas, which paid for his studies in electrical engineering in 1961 (he was already a mechanical engineer from MIT) and received his doctorate three years later from Stanford University, the hotbed of electronics pioneers. .
Chang spent 25 years at Texas Instruments, until 1983, but realized that, since he was Chinese, he could not climb very high, even though he had designed very advanced semiconductors and had American citizenship despite being born in Taiwan. He thought of doing what was crazy to everyone: making custom chips. Sun Yun-suan bought the idea from him in 1985 and appointed him president and CEO of the government-run Taiwan Industrial Technology Research Institute after leaving General Instrument and earlier Texas Instruments. In 1987 he founded TSMC, which is now the most valued semiconductor company in the world and where everyone queues to get some of their chip designs manufactured.
In the last three decades, Chang’s opposite idea has spread in the United States: designing chips and having others make them. Actually, it was the perfect complement: NVidia, founded in 1993 by Jensen Huang and still CEO, was one of the first companies dedicated to designing chips and one of TSMC’s big customers, such as Qualcomm, AMD and later Apple. Huawei has also been one of TSMC’s big customers for the more sophisticated chips it designed, until Trump got in the way of both companies.
This division of tasks in semiconductor design and manufacturing has paid off well for the entire electronics industry ecosystem, but the pandemic, like so much else, has fully exposed its weaknesses. Now, suddenly, the United States sees the need to maintain its powerful chip design industry and especially to have all-American semiconductor manufacturing plants. At least, what refers to the manufacture of the wafers, because their cutting, encapsulation and verification of the chips will already be seen: now it is concentrated in China and Southeast Asia, with encapsulations that are a mechanical prodigy, apart chip, like the one in the picture above. Trump’s idea of getting TSMC and Samsung to install chip factories on US soil no longer works, because dependence on other countries remains.
A race against the clock
What Biden wants to do, and to do this he has asked Congress to approve a $ 51 billion program over ten years, is to stimulate the creation of an integrated chip-making industry. It is the same as the European Union, Japan and obviously China, which was the forerunner in 2015 with its Made in China 2025 program, while South Korea and Taiwan want to strengthen it, both with multi-million dollar investments backed by their respective governments.
Like an auction, the stakes for being a chipmaker have skyrocketed. The European Union wants to make 20% of the chips it consumes, double that of now, while Japanese Prime Minister Yoshihide Suga plans to announce next month a major program to boost the country’s chip manufacturing industry.
Several Japanese companies are building plants in South Korea and Taiwan for the production of basic materials for the manufacture of semiconductors and photolithography equipment, as well as in mainland China. Japan is a leader in the production of material for the manufacture of chips: Shin-Etsu Chemical and Sumco, for example, control 60% of the world market for silicon wafers and between several Japanese companies they account for about 90% of the world market of products photoresists, including for extreme ultraviolet lithography (EUV), which etches chips with production lines from the Dutch company ASML.
Precisely Lee Jae-yong, vice president and one of the top managers of Samsung, traveled last fall to the ASML headquarters, in the midst of a pandemic, to negotiate the delivery of several complete EUV lines to make chips, whose only manufacturer is precisely the Dutch company. ASML has so far delivered 100 EUV lines worldwide and 70% are owned by TSMC. The remaining 30% are shared, for now, by Samsung and Intel. Controlling the entire chip production chain is a practically impossible task, but for all manufacturing of the advanced US-designed chips to be done on an island a stone’s throw from China is equally absurd, from a strategic and financial standpoint. National sovereignty. If it has been done so far, it is because it is only thinking about the immediate benefit.
Samsung plans to invest a total of 150,000 million dollars in the manufacture of semiconductor memories and logic chips until 2030, while the Government of South Korea has just announced a plan to support the semiconductor industry of 450,000 million dollars, including grants and tax deductions until the end of the decade. Apart from Samsung, the other big Korean manufacturer is SK Hynix, which makes memories.
In the city of Pyeongtaek, south of Seoul, Samsung has just started the construction of a new chip production line with a floor area of 200,000 square meters, equivalent to 25 football fields and an investment of 25,000 million dollars. (the largest building seen in the top photo of this article). Taiwan is also doing the same, with strong state support that has already led to the creation of TSMC, a company that plans to invest $ 100 billion to increase its chip production.
China does not want to be left behind either and, despite the obstacles imposed by the United States, last year alone it spent 33,000 million dollars in subsidies to the industry, especially semiconductors and defense, which is 14% more than the amount invested. in 2019, according to the Japanese newspaper Nikkei. Despite this massive injection of capital, it seems that China will not achieve its goal of manufacturing a quarter of its total demand for chips in 2025. According to consultancy IC Insights, China will reach 19.4% of the total in 2025.
The problem with all these countries that want to be sovereign in the manufacture of chips is that it is not just about investing money. It takes three to five years to set up a large semiconductor production line, in addition to several billion dollars and a large pool of highly qualified personnel. A short staff, which both South Korea and Taiwan closely monitor to prevent brain drain, especially to China. All together, the current shortage of semiconductors is not the result of chance.