Europe launches Jupiter, its supercomputer, intended to catch up in AI
German Chancellor Friedrich Merz (left) speaks with Thomas Lippert, director of the Jülich Supercomputing Center, after the inauguration of Jupiter on September 5, 2025. INA FASSBENDER / AFP
"Top priority" : German Chancellor Friedrich Merz launched Jupiter on Friday, September 5, the first giant, ultra-fast computer in Europe, intended to make up for the Europeans' delay compared to the United States and China in artificial intelligence (AI).
Based in the municipality of Jülich (Germany), west of Cologne, Jupiter is the first "exascale" supercomputer on the Old Continent, capable of performing at least a quintillion calculations per second, or a billion billion.
"It's as if 10 million conventional laptops were being used at the same time, stacked up to 300 kilometers high," the German chancellor compared it on Friday at the inauguration of the supercomputer. The United States already has three such machines, all operated by the Department of Energy.
Half a football field
Jupiter occupies an area of nearly 3,600 square meters – about half the size of a football field – with rows of processors and around 24,000 chips from the major American group Nvidia, prized by the artificial intelligence industry.
The supercomputer, developed by the French group Atos for a budget of 500 million euros, financed equally by the European Union and Germany, is the first exascale computer in Europe, and the fourth in the world according to known data.
Its colossal computing power represents "today the most powerful computing power in the world," Emmanuel Le Roux, head of Atos' advanced computing branch, told Agence France-Presse (AFP).
Jupiter is the first supercomputer that can be considered internationally competitive, according to Thomas Lippert, director of the Julier research center.
The best water bottles to replace disposable bottles
Every morning, browse the essential news of the day with the latest headlines from “Le Monde”
Every weekend, the editorial team selects the articles of the week that you shouldn't miss.
Every day new crossword, Sudoku and word search puzzles.
Le Monde's view of today's economic news Your meeting with intellectual life Le Monde's view of today's economic news
"The United States and China are in a neck-and-neck race in a global economy driven by artificial intelligence," but Germany and Europe have "every chance of catching up," according to the German Chancellor, who hailed a "historic European pioneering project."
“A question of competitiveness as much as of security”
According to a report from Stanford University (USA) published this year, American institutions produced 40 "notable" AI models in 2024, i.e. particularly influential, compared to 15 for China and three for Europe.
The new system is also "very significant" for European efforts in training AI models, says José Maria Cela, a researcher at the Supercomputing Center in Barcelona, Spain. "The performance of an AI model depends directly on the computing power of the computer used," he told AFP, adding that Europe suffers from a "deficit" of systems of this size.
Jupiter thus provides the computing power needed to efficiently train large language models (LLMs) producing huge volumes of text and used in generative chatbots, such as ChatGPT or Gemini .
In Germany and Europe, "having sovereign computing capabilities, at the level of our international competitors, is a question of competitiveness as much as of national security," Mr. Merz insisted.
However, due to its many Nvidia chips, Jupiter remains heavily dependent on American technology, even as the areas of contention between the United States and Europe have multiplied.
Applications that go far beyond AI
Researchers also want to use Jupiter to create more detailed, long-term climate forecasts to more accurately anticipate extreme events such as heat waves.
"With today's weather models, we can simulate climate changes in ten years. With Jupiter, scientists estimate that they will be able to go at least thirty years, and with some models, perhaps even up to one hundred years," explains Emmanuel Le Roux.
The machine will also be able to aid research into the energy transition, for example by simulating air flows around wind turbines to optimize their design.
In health, it could be used to simulate brain processes more realistically, for the development of drugs against diseases such as Alzheimer's.
