Philippe Notton stamps his feet in his chair. Guest of a round table on electronic chips, at the Ministry of the Economy and Finance, Monday June 17, the boss of the start-up SiPearl listens for long minutes to managers from STMicroelectronics and Soitec, two major French companies of the sector, list the major investments from which they have benefited in recent years. The European Chips Act, this 43 billion euro investment program, announced in February 2022 then adopted last summer, allowed them to finance new factories and increase their production, essential to major industries on the continent. , automobiles and aeronautics in the first place, victims of shortages during the Covid crisis. When Philippe Notton finally takes the microphone, he gets exasperated: “These are perhaps 10 to 20% of current needs. Can we finally pay attention to everything else?” In its sights, among other subjects, artificial intelligence (AI) and its latest generative evolution, in full swing and currently absent from the European chip strategy. A handful of applause and even a few clamors pierce the atmosphere which has been so polite until now.
Certainly, the man preaches for his parish: SiPearl designs chips dedicated to supercomputers or machines allowing AI to function on a daily basis, this is what we call inference. But the scene illustrates the burning debate around the European approach to electronic chips: secure its industry or innovate?
This question will be at the center of the next Chips Act 2, a text “already in gestation in Brussels”, assures Estelle Prin, founder of the European Semiconductor Observatory. “Generative AI did not exist when the European Chips Act was imagined,” deciphers the expert. And the geopolitical context, with the war in Ukraine and Chinese pressure on Taiwan, which produces high-level chips, has changed the stakes. Both camps are now refining their arguments.
Quarrel of objectives
For some, continuity is good. Chip Europe is very specialized. “One of Europe’s strengths is its vertical integration, from chip manufacturers to automobile manufacturers and their software designers. We must continue this effort,” says Arrian Ebrahimi, researcher specializing in semiconductors and author of the Chipcapitols newsletter. Prioritizing support for existing players means strengthening them. In other words, making yourself indispensable in the very globalized chip industry. “There persists, in semiconductors, the notion of interdependence. Everyone needs everyone. We have to choose our battles”, judges the director of strategy of Soitec, Steve Babureck, heard in Bercy. It is also the best way to consolidate hegemonic positions in essential niches, such as in power electronics with STMicro, or industrial gases with Air Liquide.
The second camp calls for opening your chakras. Philippe Notton, from SiPearl, hopes for a Chips Act 2 more favorable to innovative start-ups like his, dedicated to computing power, AI, cloud computing, and many other areas of the future in which Europe is lagging behind the American and Chinese blocs. “However, to date, 90% of public and private funds go to very large established companies, as well as to research laboratories,” he laments. The hope of financing the production and design of cutting-edge chips, below seven nanometers, has its virtues. The main one: market shares to conquer.
Thanks to AI, the semiconductor market, which is currently around 600 billion dollars, is expected to reach 1,000 billion by 2030. But it is also a question of supply and competitiveness of European companies . “Today, current projects do not cover all technologies. This is a risk in the evolution and growth of our aeronautics and automobile industries,” underlined Jean-Noël Mahieu, director of operations at Safran. , present at the ministry.
The interest of fabless
This argument struck a chord in the political world. Emmanuel Macron recently confirmed in our columns that he wants to equip himself with “a category of more advanced semiconductors”, in order to meet the growing needs in this area. The President of the Republic is thinking, in this, of the production of advanced nodes on European soil. “Several Taiwanese manufacturers were present at Choose France to discuss the production of sophisticated chips in France,” he said. The neighboring archipelago of China is in fact the largest designer of cutting-edge semiconductors, with its juggernaut TSMC. Europe nevertheless has assets. Imec, a Belgian nanoelectronics research institute, received 2.5 billion euros to develop chips of less than two nanometers, the dream of European Commissioner Thierry Breton. The EU can also count on the leader in manufacturing the most sophisticated machine tools in the world, the Dutch ASML.
But these ambitions to compete with the best are not unanimous. “This is absolutely not consistent with the reality of the European industry. There are neither the funds, nor the ecosystem, nor the talents, nor the end customers to hope to produce the most advanced chips on the Old Continent,” snaps Arrian Ebrahimi. Factories – a fortiori, foreign ones – are not the only way for Europe to influence AI. The Nvidia case proves it. The American firm is the new figurehead of artificial intelligence. It recently rose to first place among the largest stock market valuations in the world, at more than 3,000 billion dollars, including 1,400 billion acquired over the last year.
Nvidia owes a large part of its success to design, this phase upstream of chip production, both extremely lucrative, but also fabless: without factories at several billion dollars per unit. In addition to SiPearl, which is more of a competitor to Intel on CPU-type processors, start-ups like Scalinx or Menta are positioning themselves in this area from which Europe is generally absent, and where the United States is crushing all competition with Nvidia therefore, but also AMD or Qualcomm. Failing to build them, Europe could imagine its own chips.
The success of others
While it is difficult to say which option will prevail, one thing is certain: there is urgency for the EU. A recent study by the Boston consulting group, in partnership with the Semiconductor Industry Association (SIA), the sector’s leading association, predicts a share of European production in global chips, all categories combined, of 8% by 2032. Or more or less its current level. In its first Chips Act, Europe nevertheless hoped to reach 20% by this date. It was a bit quick to forget that all the countries involved in the very globalized semiconductor industry were also going to step up their game. The 43 billion euros invested by Europe were matched by 53 billion dollars from the United States, the multiple plans of China, the total of which exceeds 100 billion dollars, and a multitude of similar initiatives in Taiwan, Japan, Malaysia, Vietnam… As in poker, Europe thought it had a good hand and won setting. But everyone paid to see what happened next.
Some are doing quite well. The United States has already generated, thanks to its own Chips Act, more than $300 billion in private investment, three times more than in Europe. China, for its part, produces increasingly advanced semiconductors, despite American sanctions, as evidenced by the rebirth of its flagship Huawei. “The two main blocs, the United States and China, have clear goals: the first, to maintain its technological lead over its Asian rival. The second, to move towards autonomy throughout the semiconductor value chain. Europe is the only one that has not clearly defined its objectives,” explains Arrian Ebrahimi.
Good news, all the same: the European chip industry also agrees on certain points. It is seriously lacking in talent in microelectronics. And it fails to stimulate risk capital which remains insufficient, given the immense barriers to entry into the semiconductor market. Not to mention the cost of factories, with the simple design of a chip costing several hundred million euros. In fact, “while Europe brings together 37% of global start-ups, it only has 8% in semiconductors, compared to 18% for the United States and 60% for China”, indicates Arrian Ebrahimi. Initiatives are trying to reverse the trend. Asset manager Ardian launched the very first private fund dedicated to chips at the end of 2023 and plans to raise more than a billion euros to fund it. With the hope of financing the Nvidia or the European TSMC of tomorrow.
.