Parisian startup Arlequin AI SAS announced a €28 million ($32 million) Series A funding round to advance a proprietary model architecture based on topological neural networks (TNNs). The round was co-led by European venture firms Redalpine and OTB Ventures, alongside Bpifrance’s Defense Innovation Fund, with participation from Vsquared Ventures, 10x Founders, Xavier Niel, and Zebox.
Unlike traditional graph-based neural networks or transformer architectures used in LLMs, Arlequin’s TNN framework is designed to capture complex, multi-element relationships across diverse data types such as transactions, documents, and operational data. Developed in partnership with leading European and U.S. research institutions, the architecture aims to pinpoint root causes in massive datasets while using significantly less computational power and energy.
Arlequin targets deployment across defense, counterterrorism, financial fraud detection, and cybersecurity—sectors requiring verifiable evidence paths. The startup plans to use the capital to scale its international engineering team and expand commercial operations across Europe and global markets.
Why it matters
Offers a alternative model architecture (TNNs) to standard transformers, targeting high-efficiency, multi-modal relational data analysis.
Reduces compute and token costs, addressing growing enterprise and government concerns around AI energy consumption.
Highlights continued venture backing for European sovereign AI capabilities targeting security, defense, and fraud applications.
Source: siliconangle.com
