Nvidia wants to anchor itself in the European AI world
Six research centers are being built in Europe with Nvidia. The company also wants to build an "Industrial AI Cloud" for local companies.
Nvidia's DGX Superpod with B200 GPUs, which will be used by DeepL and others.
(Image: Nvidia)
Nvidia wants to make itself indispensable in European AI projects. To this end, the company is announcing numerous collaborations at its in-house exhibition GTC Paris. A total of six so-called Nvidia AI Technology Centers for research – are being established in Europe, one each in Germany, Sweden, Finland, Italy, Spain and the UK.
In Germany, Nvidia is cooperating with the BayernKI consortium, which includes the Center for National High Performance Computing Erlangen (NHR@FAU) and the Leibniz Supercomputing Center of the Bavarian Academy of Sciences and Humanities (LRZ). The "Bavaria AI Hub" will conduct research into medicine, stable diffusion models and robots.
Nvidia is not commenting on the scope of these technology centers. Nvidia usually provides access to hardware and software to facilitate research at universities. However, Nvidia generally keeps its hands off companies that want access.
Data center for manufacturers
Nvidia wants to build "the first Industrial AI Cloud" for them. The manufacturer has chosen Germany as the location, but has not yet named a specific region. Compared to hyperscalers and the fastest supercomputers, the planned system will be quite compact: The announcement mentions 10,000 Blackwell GPUs.
The fastest current AI accelerators from Nvidia, the B200, will be used, as well as the slower RTX Pro 6000, which is closely related to the GeForce RTX 5090 and uses GDDR7 memory instead of the super-fast High-Bandwidth Memory (HBM3e). Nvidia leaves the distribution between the two GPU types open.
For comparison: the fastest German supercomputer Jupiter Booster has 24,000 AI accelerators, albeit from the last generation Hopper (H100).
Videos by heise
The Industrial AI Cloud is intended for the development of product designs, simulations and so-called digital twins. In such twins, a factory is digitally copied in order to optimize processes if necessary. Mercedes-Benz, for example, wants to reduce downtimes in its own plants. The software developers Ansys, Cadence and Siemens are also on board.
Nvidia is also looking at investments from European AI developers who want to buy more hardware. First and foremost is DeepL from Cologne, which is putting a so-called DGX Superpod into operation. It contains 288 Grace processors and 576 B200 accelerators, pre-wired by Nvidia. DeepL develops the translation tool of the same name.
Empfohlener redaktioneller Inhalt
Mit Ihrer Zustimmung wird hier ein externer Preisvergleich (heise Preisvergleich) geladen.
Ich bin damit einverstanden, dass mir externe Inhalte angezeigt werden. Damit können personenbezogene Daten an Drittplattformen (heise Preisvergleich) übermittelt werden. Mehr dazu in unserer Datenschutzerklärung.
(mma)
