Nvidia, HPE to build new supercomputer in Germany

Nvidia (NVDA.O) and Hewlett Packard Enterprise (HPE.N) announced on Tuesday their collaboration with the Leibniz Supercomputing Centre to develop a new supercomputer utilizing Nvidia’s next-generation chips.

The project, named the Blue Lion supercomputer, is expected to be accessible to researchers by early 2027, employing Nvidia’s “Vera Rubin” chips. This announcement was made during a supercomputing conference held in Hamburg, Germany, and follows Nvidia’s earlier statement regarding the Lawrence Berkeley National Lab in the United States, which also intends to construct a system using these chips next year. Additionally, Nvidia reported that Jupiter, another supercomputer utilizing its chips at the German national research institute Forschungszentrum Julich, has officially been recognized as Europe’s fastest system.

These agreements signify the efforts of European institutions to maintain competitiveness against the United States in the realm of supercomputers employed for scientific disciplines ranging from biotechnology to climate research.

Long before establishing itself as a leader in artificial intelligence, Nvidia aimed to encourage scientists to adopt its chips to accelerate the resolution of complex computational challenges, such as climate change modeling. These challenges necessitated numerous precise calculations that could span months. Currently, Nvidia is advocating for the use of artificial intelligence by scientists. These AI systems can take the outcomes of a limited number of precise calculations and utilize them to generate predictions that, although not as precise as fully computed results, remain beneficial while requiring significantly less time.

On Tuesday, Nvidia introduced its “Climate in a Bottle” AI model. During a press briefing, Dion Harris, the head of data center product marketing at Nvidia, stated that scientists will have the capability to input a few initial conditions, such as sea surface temperatures, and produce a forecast for 10 to 30 years into the future, allowing them to visualize potential weather conditions at any given kilometer of the Earth’s surface.