ASUS and Poesis Successfully Deploy Agentic AI for Autonomous Financial Trading

Quick Summary

ASUS and Poesis have successfully demonstrated that agentic AI trading can operate autonomously in live financial markets using local hardware. The week-long experiment utilized the ASUS ExpertCenter Pro ET900N G3, proving that complex investment workflows no longer require cloud-based infrastructure for secure, real-time execution.

How Does Agentic AI Trading Work on Local Hardware?

The experiment proved that agentic AI trading can function entirely on-premise, bypassing the latency and security concerns of cloud environments. By utilizing the ASUS ExpertCenter Pro ET900N G3, Poesis deployed a multi-agent AI workflow on a single deskside supercomputer. This setup provided the computational power necessary to handle high-frequency data processing and decision-making in real-time.

• 20 petaFlops of AI performance delivered by the NVIDIA GB300 Superchip.
• 748GB coherent memory allowing for massive local model processing.
• Zero cloud reliance, ensuring data privacy and reduced execution lag.

What Results Did the Poesis Autonomous Trading Experiment Deliver?

During the week-long test, the system managed live capital autonomously, handling the full lifecycle of an investment. The agents were responsible for automated investment research, risk management, and trade execution. This demonstrates the potential of agentic AI trading to move from theoretical experimentation into practical development phases for institutional asset managers.

"While this was an early-stage experiment, it showed that agents can operate continuously in live markets while remaining within clearly defined constraints," said Alex Popa, Founder and CEO of Poesis.

Why is On-Premise AI Critical for Asset Management?

For financial institutions, data sovereignty and security are paramount. The ability to run agentic AI trading workflows locally means that proprietary strategies and sensitive market data never leave the firm's physical infrastructure. This edge-to-enterprise approach allows for continuous market operation without the risk of third-party cloud outages or data breaches.

“Working alongside Poesis at the intersection of AI and financial markets gave us valuable insights,” said Yen Hoang, Director of Marketing, B2B, ASUS North America. “As agentic AI continues to evolve, projects like this are essential to helping us better understand how autonomous investment workflows can be developed and deployed in real-world trading environments.”

FF NEWS TAKE:

This experiment marks a pivotal shift in the fintech landscape. By successfully running agentic AI trading on a deskside supercomputer, ASUS and Poesis are challenging the assumption that advanced AI requires massive cloud clusters. For the industry, this moves the needle toward decentralized institutional trading, where speed, security, and autonomy are handled at the hardware level. It is a bold step toward truly autonomous asset management.