Training an AI agent to run part of a factory or supply chain takes serious computing power, and most companies would rather not build a supercomputer to get it. On November 5, 2026, Viridien and AMESA announced a partnership aimed at exactly that gap.
Viridien, a Paris-based advanced technology and HPC company, will supply the high-performance computing infrastructure behind AMESA Agent Cloud. AMESA is an enterprise AI platform for what it calls operational autonomy.
What was announced
According to Viridien’s press release, AMESA already works with some of the world’s largest companies to build and train AI agents for critical operational uses such as industrial process control and supply chain and logistics. Under the partnership, AMESA customers can use enterprise-scale HPC and AI infrastructure and expertise without building, managing or maintaining their own high-performance environments.
Viridien’s dedicated, managed AI training infrastructure becomes the hosted foundation of AMESA Agent Cloud. It will give customers compute for training agents and support future hosted autonomous AI services, with managed operational support included.
The performance claim
The headline number comes from AMESA’s own testing: Viridien’s managed HPC environment delivered a 10x to 30x improvement in agent training performance compared with AMESA’s previous hosted environment. The release says that lets customers train, test and iterate agents significantly faster and shorten the path from development to production.
It is worth reading the number carefully. It is a vendor-reported comparison against AMESA’s earlier hosted setup, not an independent benchmark, and the release does not publish the workloads behind it.
Teaching agents through simulation
AMESA’s approach is built around simulation and machine teaching. An organization’s experts teach AI agents how to make complex operational decisions, then validate how the agents perform before they touch real-world environments. Because that cycle repeats across processes, facilities and business units, fast training loops matter more as deployments grow.
AMESA CEO Alistair Fulton said enterprises are ready to move beyond AI that only gives answers and recommendations to AI trusted to make complex operational decisions. The managed infrastructure, he said, lets customers train and scale such systems much faster without becoming HPC experts.
Deployment choice stays open
The partnership does not lock customers in. Organizations can run AMESA in their own cloud or on-premises, or use AMESA Agent Cloud powered by Viridien. The release says this lets enterprises choose the model that fits their security, data governance, performance and infrastructure needs while keeping one consistent approach to developing and training autonomous agents.
Gretell Otano, Viridien’s VP of Growth and Strategy for HPC and Cloud Solutions, said performance, reliability and scalability become more important as organizations move AI initiatives into production, and that Viridien brings managed infrastructure, optimization expertise and operational support.
What it means
The deal reflects a broader pattern: as enterprise AI moves from pilots to production, specialized infrastructure and the people who run it become a bottleneck. Outsourcing that layer to a managed HPC provider is one answer, while keeping an on-premises option is another.
For IT and operations leaders, the questions to ask are practical. What workloads were measured, how does the cost compare, and how are data governance requirements met when training moves to a hosted environment? The announcement points to flexibility on the last question but gives no pricing or customer names.
Bottom line
This is a partnership announcement rather than a product launch, and its biggest claim is self-reported. Still, it highlights where the enterprise AI race is heading: not just better models, but faster, managed ways to teach agents to do real operational work.