DeepL has announced its intention to deploy a computing system on the NVIDIA DGX SuperPOD platform with GB200 accelerators. The system will be located at the EcoDataCenter site in Sweden, and its commissioning is scheduled for mid-2025.
DeepL specializes in developing AI-based automatic translation tools. According to Jarek Kutylowski, CEO and founder of DeepL, the company has created a solution that surpasses all other services on the market in terms of translation accuracy. More than 100,000 enterprises, governments and other organizations, as well as millions of individual users around the world, use DeepL language AI tools. The company has more than 1 thousand employees. It is supported by investments from Benchmark, IVP, Index Ventures, etc.
In 2023, DeepL deployed the Mercury supercomputer based on NVIDIA DGX SuperPOD with H100 accelerators. In the June TOP500 ranking, this system ranks 41st with FP64 performance of 21.85 Pflops and theoretical peak performance of 33.85 Pflops.
Image source: NVIDIA
The NVIDIA DGX SuperPOD platform with GB200 accelerators uses liquid cooling. Scaling to tens of thousands of accelerators is possible. DeepL intends to use the new complex for research tasks – in particular, to develop advanced AI models that will further expand the capabilities of translation tools between different languages. This will “break down language barriers for businesses and professionals around the world,” the company promises.