Europe bets on supercomputer to catch up in AI race

The inauguration of Europe's most advanced supercomputer, Jupiter, took place on Friday in Germany, with Chancellor Friedrich Merz highlighting its potential to elevate the continent's standing in the global artificial intelligence (AI) arena. "Germany, along with Europe, has the chance to not only catch up but also maintain parity with AI leaders such as the United States and China," Merz stated. Jupiter, located at the Juelich Supercomputing Centre in western Germany, is heralded as the first 'exascale' supercomputer in Europe, capable of executing at least one quintillion calculations per second. In comparison, the United States boasts three similar systems, all managed by the Department of Energy. The expansive facility spans approximately 3,600 square meters, housing racks filled with around 24,000 Nvidia chips, which are essential for AI applications. The development and operation of the supercomputer are backed by a budget of 500 million euros ($580 million), with funding split evenly between the European Union and Germany. The immense computational power of Jupiter will be accessible to researchers and businesses across various disciplines for endeavors such as training AI models. Thomas Lippert, head of the Juelich centre, remarked that Jupiter represents a significant advancement in Europe's computing capabilities, being 20 times more powerful than any other existing computer in Germany. Lippert emphasized that Jupiter is uniquely positioned to help Europe compete internationally in AI model training, an area where it has historically fallen behind. A recent report from Stanford University indicated that US institutions produced 40 noteworthy AI models in 2024, while China generated 15, and Europe lagged with merely three. Merz acknowledged the current tight competition between the US and China but reiterated the importance of Europe regaining its footing in this critical technology sector. "We need to establish computing capacities that are competitive on a global scale, which is essential for both our economic standing and national security," he asserted. The Jupiter supercomputer was developed by a consortium led by Eviden, a subsidiary of French tech giant Atos, and the German firm ParTec. However, the reliance on Nvidia chips highlights the ongoing dependence on US technology, raising concerns amid strained US-European relations. Beyond its AI capabilities, Jupiter has diverse applications, including enhancing climate forecasting. Researchers aspire to generate more accurate long-term predictions for extreme weather events, with the potential to extend forecasts from a decade to 30 years or even up to 100 years in some instances. Additionally, it could play a role in simulating brain processes, aiding drug development for diseases like Alzheimer's, and optimizing designs for clean energy solutions, such as wind turbines. In terms of energy consumption, Jupiter is expected to require around 11 megawatts of power, comparable to the energy demands of thousands of households or a small industrial facility. Nevertheless, its operators claim it is among the most energy-efficient supercomputers globally, utilizing cutting-edge hardware, water-cooling systems, and recycling waste heat to warm nearby buildings.