In an escalation of its legal defense over supercomputing patents, Munich-based ParTec AG has sued Nvidia for allegedly infringing upon its intellectual property essential to AI-focused supercomputing.
The lawsuit, lodged in Munich’s Unified Patent Court (UPC) on October 27, accuses Nvidia of using technology central to ParTec’s dynamic Modular System Architecture (dMSA), which the German firm says is protected by its patents across 18 European countries. Filed only months after ParTec’s initial lawsuit against Microsoft, this case underscores the company’s effort to safeguard its AI-compatible supercomputing designs as demand rises globally.
In June, ParTec first filed its U.S. lawsuit against Microsoft, accusing it of integrating ParTec’s patented dynamic resource allocation system into the Azure AI platform without authorization. ParTec’s technology, central to optimizing how processors and GPUs interact for AI tasks, allegedly enables Microsoft’s platform to handle large-scale models more efficiently. Represented by Susman Godfrey LLP, ParTec seeks damages and an injunction to prevent Microsoft from using the disputed technology within its cloud services.
The Nvidia Lawsuit: Extending Patent Claims to Europe
The new lawsuit from ParTec against Nvidia is seeking an injunction across 18 European countries via the UPC. According to ParTec, Nvidia’s GPUs infringe upon two key patents—EP2628080 and EP3743812. These patents enable real-time, flexible resource allocation within computing clusters, a feature ParTec asserts is indispensable for modern AI applications. ParTec hopes the UPC will restrict Nvidia’s GPU distribution in countries covered by Europe’s Unitary Patent System, which permits a single court order for multi-nation enforcement.
In response to ParTec’s claims, Nvidia declined further negotiations, citing the ongoing Microsoft case as justification, given its customer relationship with Microsoft. Despite multiple approaches from ParTec for a licensing arrangement, Nvidia, a global leader in GPUs, has refused to engage formally.
Defining ParTec’s Patented Technology
ParTec’s dynamic Modular System Architecture (dMSA) underpins its patents and is pivotal to many supercomputing applications. Designed for efficient resource management, the dMSA allows supercomputers to dynamically shift CPU and GPU allocations based on workload demands, especially useful in AI applications that require varying computational resources. Unlike traditional setups, dMSA’s design supports flexible connections, reducing bottlenecks and improving performance as resource requirements shift.
The ParaStation Modulo software, which facilitates this modular design, powers many of Europe’s high-performance computing systems, including the JUWELS and JUPITER supercomputers at Germany’s Jülich Research Center. These systems have been crucial for Europe’s AI research, capable of handling complex tasks that demand efficient, high-capacity computing.
Patent Overview and Broader Market Implications
The two patents central to the Nvidia lawsuit—EP2628080 and EP3743812—were filed in 2011 and 2019, respectively, covering methods that optimize resource allocation across a computer cluster. The latter patent, enforceable across the Unitary Patent System, could significantly impact Nvidia’s GPU sales in Europe if ParTec’s claims are upheld. These patents cover dynamically assigned resources, ideal for AI supercomputers where different processor types, such as CPUs and GPUs, must interact seamlessly based on task demands.
ParTec has a substantial portfolio of around 150 patents, including several forming the basis of its U.S. lawsuit against Microsoft. U.S. patents 10,142,156, 11,934,883, and 11,537,442 protect similar resource management capabilities within computing clusters. According to ParTec, Microsoft uses these capabilities within Azure AI, allowing it to manage large-scale AI model training, which ParTec asserts is unauthorized.
Potential Fallout for Nvidia’s GPU Business in Europe
If the UPC rules in favor of ParTec, Nvidia could face restrictions on its DGX product line, which is widely used in AI research and model training globally. The DGX systems, designed for flexible resource distribution, would require redesign or adaptation to comply with ParTec’s patents in Europe. As Nvidia’s market influence continues to grow, a ruling against the company may lead to costly changes to retain its European market position.
This case also brings broader implications for Europe’s tech independence. As ParTec emphasizes, protecting its IP represents an opportunity for Europe to maintain control over essential AI technology, which currently remains highly concentrated among U.S. firms. In an era where the AI landscape is evolving rapidly, ParTec’s assertive legal stance could drive more European autonomy in AI infrastructure development.
Positioning Europe in the AI Race
The lawsuit has drawn interest from both technology and policy circles, as it represents one of the few major IP disputes over AI technology that aligns with European interests in technological sovereignty. ParTec’s ParaStation Modulo software, part of many leading European supercomputing projects, has enabled a modular approach that matches the demanding requirements of AI, which traditional, static architectures struggle to meet.
In the U.S., ParTec’s lawsuit against Microsoft is scheduled to reach trial in April 2026 under Judge Rodney Gilstrap in Texas, signaling a continued, cross-Atlantic effort to enforce its intellectual property in AI. With the Microsoft and Nvidia cases both in motion, ParTec’s strategy underscores the competitive tension between small innovators and large U.S. tech firms, who are increasingly targeted by IP claims in a fast-evolving AI market.
Last Updated on November 7, 2024 2:18 pm CET
