German AI powerhouse DeepL has taken a dramatic leap forward in language AI, announcing the deployment of an NVIDIA DGX SuperPOD powered by their latest GB200 systems. According to the company, this move allows them to translate the entire internet in just 18 days, a staggering improvement from the previous benchmark of 194 days .
🧠 The Tech Behind the Speed
At the heart of this breakthrough is the DGX SuperPOD, a large-scale system of NVIDIA GB200 Grace Blackwell superchips. Housed in Sweden’s EcoDataCenter, it constitutes one of Europe’s most advanced AI infrastructure setups—boasting over 1,000 Hopper GPUs, ultra-fast InfiniBand networking, liquid cooling, and exaflop-class performance .
DeepL reports:
- Up to 10× faster processing compared to their previous H100-based systems.
- A 30× increase in text output volume.
- Translation of the Oxford English Dictionary in just 2 seconds (previously 39 seconds), and Proust’s ‘In Search of Lost Time’ in 0.09 seconds (down from 0.95).
💬 What DeepL Has to Say
Stefan Mesken, DeepL’s chief scientist, told CNBC:
“Our new infrastructure will give our scientists much more computing power to build even more sophisticated models… Not long ago, this was technically impossible, but now it has become a reality.”
And according to a CNBC X-posted summary:
“DeepL on Wednesday said it was deploying one of the latest NVIDIA systems that would allow the German startup to translate the whole internet in just 18 days.”
🎥 See the Breakthrough in Action
A concise breakdown of this milestone is available in the CNBC explainer.
“CNBC’s Arjun Kharpal explains how this transformative computing platform would enable DeepL to translate the whole internet in just 18 days.”
🌍 Why It Matters
1. Quality and depth enhancements
Faster processing unlocks more nuanced language understanding, improved context-awareness, and potential for real-time generative AI features.
2. European leadership in AI infrastructure
DeepL becomes the first commercial European user of NVIDIA’s GB200 SuperPOD, showcasing competitive parity with global AI players.
3. Wider AI implications
The SuperPOD’s performance paves the way for future applications: DeepL Voice, Clarify, multilingual enterprise tools—and eventually multimodal LLMs .
🔭 What’s Next
- Expect sweeping improvements in translation speed, accuracy, and interactive AI tools.
- DeepL and NVIDIA will collaborate at GTC 2025, highlighting their symbiotic engineering efforts—the combination of cutting-edge hardware and software optimisation .
- Ongoing deployment and scaling across global data centres will support broader enterprise translation, voice translation, and real-time communication tools.
✅ Final Word
DeepL’s adoption of NVIDIA’s GB200 SuperPOD is a striking demonstration of AI-powered infrastructure tangibly accelerating capabilities—taking a process that once required months and condensing it into weeks. As DeepL continues to push the envelope in language AI, this leap offers a glimpse into an era of ultra-fast, ultra-accurate, and deeply interactive translation services. With global stakes rising in AI leadership, Europe’s presence in the high-performance AI arms race is now undeniably stronger.