The European Union has formed an industrial alliance to try to unite key players in semiconductor design and manufacturing and regain some technological sovereignty in the strategic chip market. The objective is to produce at least 20% of the chips consumed on European soil. The plan is still to be detailed, but they would like to have TSMC to manufacture the most sophisticated chips, which would be integrated into the multiple semiconductors already produced in Europe by the Infineon, ST, NXP, Bosch and GF plants, among others, together with the chip design tools and semiconductor manufacturing equipment of the Dutch company ASML, which are the most important and sophisticated in the world.
The Industrial Alliance for Processors and Semiconductor Technologies, launched on July 19 in Brussels, has the overall objective of “identifying current problems in microchip production and the technological developments that companies and organizations need to be successful, its size does not matter ”. This would help the competitiveness of companies, increase Europe’s digital sovereignty and drive demand for the next generation of secure, energy efficient and powerful chips and processors, the document added.
The Alliance wants to “strengthen and foster collaboration with existing and future European Union initiatives” and provide the necessary capabilities in semiconductor technologies that Europe needs to run its critical infrastructure and digital communications networks. In addition, “that supports a wide range of sectors and technologies, including automotive, industrial automation, health and systems based on artificial intelligence.”
The novelty of the European Alliance is that it has the commitment of the European manufacturers of specialized chips to pass orders to a manufacturer of sophisticated chips located in Europe
With this, it is specified in the document, the European electronic design ecosystem would be strengthened and the necessary manufacturing capacity would be established, which it considers to be the two main deficits that Europe has in this field. Any organization with existing or planned relevant activities is thereby encouraged to join the Alliance. A European Alliance for Industrial Data, Edge and Cloud has also been created, as specified in the statement released, as other technologies are equally key to the technological future of the European Union.
The purpose of these two Alliances is not new, because they were already part of the communication presented by the European Commission in mid-March, with the objectives for European digital sovereignty in 2030. The objective of “being manufactured in Europe at least 20% of world production in value of last generation semiconductors and in a sustainable way ”, compared to 10% calculated at the moment. Recovery funds should help achieve this in part.
The main novelty that has occurred since the initiative for Europe to be more sovereign in the manufacture of chips was formulated half a year ago is that it wants to resort to an effective public and private collaboration and to have the full commitment of the current European manufacturers of chips. chips. At first, it was felt that the main objective was to manufacture very sophisticated chips, of five nanometers or less, in Europe, when the European demand for these miniaturized chips is very low, as the main European manufacturers have made clear in recent months. of semiconductors, and who prefer to make specialized chips, with a greater market and profitability.
Various European chipmakers
Europe has several important chip manufacturers, such as the Franco-Italian ST Microelectronics, the Dutch NXP or the German Infineon or Bosch, which recently inaugurated an advanced plant in Dresden and which will start up this September. These companies manufacture chips essentially for automotive or industrial automation equipment, which must be very robust and withstand adverse conditions, which means that they must work at higher supply voltages and currents than, for example, a smartphone. The nodes of the chips, consequently, must be larger, 14 or even more nanometers, and thus more immune to interference.
The problem is that these “industrial” chips also need a very sophisticated “brain”, like a computer processor, to function properly. Companies, European or not, such as the Japanese Renesas, must go essentially to the Taiwanese TSMC (or, failing that, to the Korean Samsung) to manufacture these sophisticated chips to order, because they are the only ones that can manufacture them (with equipment from Dutch company ASML, by the way, because it is the only one capable of working with extreme ultraviolet technology, at 150 million dollars each production line).
With the pandemic, it is well known, the demand for cars fell and also that of industrial equipment, with factories stopped, which led to lower orders for automotive and industrial components. These industrial chips use sophisticated chips to a minimum and also their unit value is very low, a few euros, compared to the tens and hundreds of euros that processors for computers, workstations, graphics cards or smartphones are worth.
The key to success is, in the end, that all the parties involved agree on a good business plan and that everyone wins in the long term, because it is a business at least ten years from now.
As if that were not enough, the supply chain is very long and involves dozens of intermediaries. It was common for, in the normal times of yore, four months after an order was placed until the component was shipped to the car production plant, for example, without car brands, for example, being aware of it. it. While the makers of graphics cards and smartphones saw it coming and quickly grabbed all the chips they could at whatever price, automakers and others found themselves (and still are) without supplies of the prized and, for them, ignored components. .
A very extensive chip ecosystem
The semiconductor manufacturing industry, it has now become aware, is made up of hundreds of companies, involving thousands of subcontractors around the world, who are responsible for cutting, encapsulating and verifying the wafers with the billions of transistors and electronic circuits etched on its surface. A few tens of thousands of chips are tremendously sophisticated; the other billions not so much.
Furthermore, the most sophisticated ones are always made to order, for a client with very precise specifications. In the case of Intel, always (for now) for its own consumption and with its own factories; In the case of the other two members of this tremendously exclusive league (TSMC and Samsung), it would also be possible to distinguish between the Taiwanese TSMC, which has always manufactured for third parties, and Samsung, which a large part of its semiconductor manufacturing business is dedicated to manufacture memories and the other to order (including your own needs).
Companies that want a very sophisticated chip, such as Qualcomm, Apple, NVidia, AMD or in its day Huawei, before the American veto, to name VIP customers, should go to TSMC. Or, failing that, Samsung, which has recently greatly enhanced the custom chip manufacturing division, radically different from memory. The rest of the companies that require chips of less than ten nanometers must also go to TSMC or Samsung and wait months for them to attend them, because both have more than full order books.
There is a manufacturer of sophisticated chip technology, between 10 and 14 nanometers, which is GlobalFoundries (GF), owned by the sovereign wealth fund Mubadala of the United Arab Emirates. This company has factories in several countries, such as in the German city of Dresden (when it owned AMD), and a few days ago it announced the expansion of 1,000 million dollars in its plant near New York and 4,000 million dollars in construction. of a new plant in Singapore. Pat Gelsinger, Intel’s CEO since January, announced his intention to buy it for $ 30 billion, which GlobalFoundries CEO Tom Caufield does not seem to have paid any attention to and has responded with the announcement of the new investments of a company that works very well.
Negotiations at the highest level
On July 21, coinciding with the visit of the European Commissioner Thierry Breton to the chip factory that ST Microelectronics has near Grenoble, its main owner, Jean-Marc Chéry, took the opportunity to explain to Breton that the key factor in time to manufacture semiconductors is not miniaturization but specialization, as he commented a few days later when giving an account of the quarterly results.
Until a few months ago, Chéry was very reluctant to Brussels’s plans, precisely because of his commitment to nanometers. But now, especially after Breton’s visit, ST seems to be more willing to collaborate with the Alliance and, it seems, has committed to subcontracting part of the production of the sophisticated chips to the plant that the European Union wants to promote. , instead of passing the orders as before to Asian plants.
The other European semiconductor manufacturers also seem more inclined to join the Alliance, after months of explaining in multiple interviews the key role that Europe plays in specialized chips and where the vast majority are 14 or more nanometers. Commissioner Thierry Breton, from Brussels, praised the change in attitude of Jean-Marc Chéry. “We have an ambitious vision on a European scale and our industrialists who want to remain competitive will have our full support,” he told the French press.
It remains, however, to put the cherry on the cake. The most suitable candidate is TSMC, in case you want to build a highly sophisticated custom chip factory. Their main customers, logically, would be current and future European chip producers, who would complement their product with chips made by themselves and those made by TSMC. In reality, it is a task that TSMC has been doing since its inception, more than thirty years ago: its customers provide the design of the chips, conveniently protected by patents and confidentiality guarantee, and TSMC goes from the design to fully functional physical chips . The “downside” is that a plant of these characteristics requires an investment of the order of 20,000 million dollars and about four years to build it, apart from the appropriate location, qualified workforce and expert subcontractors, the latter two scarce.
Numerous suitors
Implementing a functional semiconductor manufacturing “ecosystem” is complex, but not impossible. Actually, what TSMC has left over are suitors for it to set up a large manufacturing complex off the island of Taiwan. With Donald Trump, TSMC agreed to set up a plant in the United States, specifically in Arizona, with an investment of 12,000 million dollars and generous American subsidies. But TSMC is also in negotiations to build other plants in Japan, China, Sngapur and, a few days ago, it received the go-ahead from its government to create a huge factory in Taiwan, with the capacity to make two-nanometer chips. TSMC president Mark Liu however confirmed on July 26 that he “is considering the construction of its first European plant in Germany.”
It is evident that TSMC finds it much more convenient to have all its factories grouped together on the island of Taiwan, if only because of the economies of scale that are generated, but it must be taken into account that the island is exposed to earthquakes, typhoons and , lately, water shortage. Therefore, a suitable strategy would be to diversify the implantation of chip factories close to its main customers. Logically, the most sophisticated and strategic plants would continue to be located in Taiwan and the secondary and expendable plants in other countries. This is what Intel has traditionally done: its laboratory plant is at its headquarters in Santa Clara and later replicates it in other states of the United States, as well as China, Israel or Ireland, where it built a plant decades ago and has been updating it.
The key to success is, in the end, that all the parties involved agree on a good business plan and that everyone wins in the long term, because it is a business at least ten years from now. There is another not inconsiderable aspect that also enters the equation: the pressure from the different governments and, especially, from China, to have a TSMC plant. If the United States, Japan, the European Union or even Singapore press for a TSMC plant, in the case of China it is a neighbor that is less than 200 kilometers away, with armed frigates that prowl its coasts without no dissimulation and that cannot be underestimated in any case, although the United States and Japan ensure the extreme defense of the island territory. China’s patience is not infinite and it has a plan B, to manufacture much of the sophisticated chips. Several consulting firms, such as Strategy Analytics, say that next year they will be fully ready to manufacture 14-nanometer chips on an industrial scale.
It is clear that for Europe the TSMC option is not the only one possible; Samsung’s is obvious, although the Korean company is much less used to making chips to order, much less outside of its main plants. What would be surprising is that the European Union put 20,000 million dollars or whatever it was to get Intel to install a factory of microprocessors for its own consumption, which is the best it knows how to do, although now it is trying hard to ensure that it wants to do chips on request. This would mean, among other things, making chips based on ARM rules and not x86 as it has been doing for more than half a century, since its inception. That Emmanuel Macron greeted Pat Gelsinger with deference to the Palace of Versailles doesn’t mean much in the devilish chip business.
In any case, this fall they could bear fruit or, at least specify, some of these agreements that are being developed at the highest level to have the manufacture of chips closer to the main customers. A system that will probably not be so profitable globally, but with the undoubted advantage of greater control of orders by customers and avoiding that some of them are not out of supply for months, as has been the case for more than a year and it will continue at least well into 2022 or even 2023 for relative normalization. Because the mismatch between supply and demand and the variability of prices is part of the half century of semiconductor history, especially memories.