ASUS and Poesis partnered for a week-long experiment in which autonomous AI agents conducted market research, managed risk, and performed real trades. The agents ran on the ASUS ExpertCenter Pro ET900N G3 laptop with NVIDIA technology.
The experiment marks a shift in how AI operates in financial markets. Rather than providing analysis and recommendations, these agents collected data, developed trading strategies, and executed trades autonomously. The 24/7 nature of cryptocurrency markets and their programmable infrastructure made them an ideal testing ground for this approach.
Local Deployment vs. Cloud-Based Models
ASUS advocated for running agents locally rather than in the cloud. The company identified what it calls The Token Trap: continuous agent activity generates constant API calls in the background, leading to unpredictable billing and usage caps imposed by cloud providers.
ASUS compared the cost profiles, noting that an upfront hardware investment of approximately $4,000 plus roughly $384 in annual electricity costs could be more economical than $6,000 to $8,000 in annual API fees for equivalent cloud-based workloads.
The Poesis test itself was a short proof-of-concept rather than long-term performance testing and was not specifically designed for cryptocurrency trading.
The Evolution of Financial AI
The shift from advisory AI to autonomous trading represents a broader evolution in financial technology. The stages include data perception, reasoning, strategy generation, and execution with control. According to the Global AI in Financial Services Report compiled by Cambridge Judge Business School, 52% of surveyed companies across 628 organizations in 151 countries have experimented with or deployed agentic AI.
Exchange Infrastructure for Autonomous Trading
Major cryptocurrency exchanges have built infrastructure to support autonomous agents throughout 2026. Binance launched Agent OS in August, connecting AI tools to dedicated Agentic subaccounts with restricted withdrawal capabilities. Daily limits include $50,000 for swaps and $100,000 for DeFi transactions by default.
Other exchanges expanded access to autonomous trading: Kraken and OKX in March, Bitget in April, Gemini in April, and Coinbase in June. The common approach across platforms involves permissioned autonomy—agents can operate within account separation, permission controls, and defined limits.
Risks and Market Implications
Autonomous agents could enable broader market coverage, faster execution, and deeper electronic trading. However, they also carry concentrated risks. The CFA Institute warns that similar AI models reacting to similar market signals could create crowded, reflexive markets and an algorithmic monoculture.
The Financial Markets Standards Board emphasizes that market-facing AI must operate under human supervision and that accountability must remain clear. Additionally, cryptocurrency markets face a capacity constraint. If agentic activity scales significantly, increased settlement activity could move onto blockchains, potentially straining available block space.
