(de-news.net) – As Germany expands its collaboration with the company Nvidia to advance industrial AI, German Chancellor Friedrich Merz (CDU) is looking to draw in foreign investment. With Siemens showcasing possible uses in production planning and factory automation, the alliance seeks to boost manufacturing competitiveness, technical sovereignty, and AI adoption. With Nemotron 3.5 Super VL and Bolt, Nvidia has also broadened its open AI model portfolio, focusing on local deployment and enterprise workflows while increasing prospects for developers and industrial users.

Chancellor Merz is seeking to strengthen Germany’s appeal to foreign investors amid economic and geopolitical uncertainty, emphasizing the importance of open markets, multilateralism, and international cooperation. Speaking at the inaugural Invest in Germany summit in Berlin, he presented the country as a stable and reliable economic partner while identifying infrastructure, green energy, and defense as key priorities in its broader modernization effort.

The German government and U.S. technology company Nvidia have signed a memorandum of understanding aimed at accelerating the development and industrial application of artificial intelligence (AI). The initiative will draw on Germany’s manufacturing expertise, engineering capabilities, and industrial data to strengthen technological sovereignty and support the development of a European industrial AI ecosystem. The government’s Industry 4.0 platform is expected to serve as a central strategic network, while Nvidia plans to expand its cooperation with German businesses, including small and medium-sized enterprises.

Merz and Nvidia CEO Jensen Huang have emphasized that Germany should use its domestic industrial expertise to develop AI capabilities rather than become dependent on externally developed solutions. Nvidia reportedly invested nearly $1 billion in Germany last year. Huang has encouraged businesses to regard industrial data as a strategic national resource, arguing that AI should build on and extend the capabilities of German engineers rather than replace their expertise.

Siemens and Nvidia demonstrate AI-driven manufacturing

Siemens and Nvidia are developing AI applications for industrial production, with Siemens’ electronics facility in Erlangen serving as a potential model for wider implementation across the company and among its customers. The partnership seeks to increase automation, improve production flexibility, and help preserve manufacturing competitiveness in high-wage economies. Cedrik Neike, a Siemens executive, estimated that AI could raise the proportion of factory operations suitable for automation from 30 to 40 percent to between 70 and 80 percent, including processes that must accommodate changing production requirements.

Automated optical inspection of circuit boards provides one example of how the technology could change manufacturing. Traditionally, inspection systems have required manual training whenever product configurations change. By combining product and production data with digital twins and Nvidia’s simulation technology, manufacturers can now generate realistic representations of both acceptable and defective circuit boards before production begins. This approach is intended to reduce the need for repeated manual preparation and production interruptions.

Digital twins and simulations can also support production planning, allowing human operators to work with AI agents to determine how manufacturing processes should be organized and implemented. The approach reflects a broader effort to apply AI to complex industrial environments in which flexibility is as important as the ability to repeat standardized operations.

Nvidia expands its open AI model portfolio

Nvidia has expanded its AI offerings with the release of Nemotron 3.5 Super VL, an open model designed for enterprise AI agents, and the Nemotron 3.5 Bolt family, which is intended for local deployment on compatible hardware. Super VL has 120 billion parameters, of which 12 billion are active for each request. It can process text, images, websites, PDFs, tables, and scanned documents, while its one-million-token context window is designed to support complex business workflows involving document analysis and software tools.

According to Artificial Analysis, Super VL scored 26 on its Intelligence Index, which combines assessments of programming, AI agents, knowledge work, and scientific capabilities. Its predecessor scored 13, while Nvidia’s previous flagship, Nemotron 3 Ultra, achieved 23. Super VL also generates substantially fewer output tokens per task than its predecessor. The model is available through Hugging Face, OpenRouter, and Nvidia’s microservice offering.

Nvidia has also released training datasets and training methods alongside the model weights under the OpenMDW-1.1 license, providing developers with access to more of the resources used to build the system. The model was trained on approximately 26 trillion tokens and subsequently refined using about 3.8 million examples and reinforcement learning. According to Artificial Analysis, API usage costs approximately $0.23 per million input tokens and $0.57 per million output tokens.

Company expands open AI portfolio with locally-deployable Bolt model

The second release, Nemotron 3.5 Bolt, is designed to operate locally on compatible consumer hardware. Its initial model contains 74 billion parameters, with nine billion active, and can fit into 24 GB of memory through two-bit quantization. This makes it suitable for systems including gaming PCs equipped with Nvidia RTX graphics cards and the company’s DGX Spark and RTX-Spark devices. The model also supports commonly used tools such as Ollama, LM Studio, and llama.cpp.

Bolt scored 21 on the Intelligence Index, placing it among the stronger models with fewer than 100 billion parameters. Despite its smaller overall size, it approaches the performance of Nemotron 3 Ultra, which contains 550 billion parameters. However, Bolt generates more output tokens per task than Super VL, indicating a trade-off in computational efficiency. Nvidia has also announced plans for a smaller Bolt E4B model intended for applications such as voice assistants and robots.

Nvidia is positioning the Nemotron family as a foundation for AI systems developed by other companies. Its partners include Salesforce, Atlassian, CrowdStrike, Perplexity, Synopsys, and Dassault Systèmes. EY has also used a fine-tuned version of Super VL to develop a due-diligence agent through Scale AI. In parallel, reported discussions about a potential acquisition of open-weight AI startup Reflection AI point to Nvidia’s broader interest in expanding its role in the AI model ecosystem.