FOR IMMEDIATE RELEASE

FLUIX AI Announces A.I.M.I.® 1.0, a Groundbreaking AI for Data Center Efficiency, Achieving Significant Energy Savings in Live Trials

San Francisco, CA – January 2025 – FLUIX AI today released a new white paper on A.I.M.I.® 1.0, a deep reinforcement learning (DRL) system for the continuous control of data center cooling.

In a real-world testbed using production-grade equipment, A.I.M.I. demonstrated:

• A 53.6%+ reduction in HVAC energy consumption

• A 28% improvement in Power Usage Effectiveness (PUE)

when compared to traditional control systems.

These results signal a significant step toward creating autonomous, energy-efficient infrastructure capable of supporting the next generation of AI workloads and meeting the energy demand for data centers. 

The Challenge: Cooling the AI Revolution

Data centers are the backbone of the digital economy, but their energy consumption is a growing concern. The U.S. Department of Energy projects that data centers could consume up to 9% of the nation’s electricity by 2030, largely driven by the explosive growth of AI factories [1]. Cooling is the single largest consumer of energy within these facilities.

Conventional cooling systems rely on static, rule-based PID controllers that are slow to react and cannot predictively adapt to the volatile thermal loads generated by modern AI and high-performance computing.

The Solution: A.I.M.I.® 1.0 Continuous Control

A.I.M.I. (Artificial Intelligence for Managing Infrastructure) is a closed-loop AI controller that learns the unique thermal dynamics of a facility and makes continuous, predictive adjustments to cooling systems in real time.

By using an off-policy actor–critic DRL framework, A.I.M.I. can safely explore control strategies and optimize for energy efficiency while operating within strict temperature and equipment safety constraints.

“Traditional systems are reactive,” said Chase Overcash, lead author of the white paper, “A.I.M.I. is predictive. It learns the physics of the data center to anticipate thermal changes and continuously fine-tunes cooling setpoints for supply air, fan speed, and humidity. This is a fundamental shift from static setpoints to dynamic, intelligent control.”

Verified Performance in a Real-World Testbed

FLUIX AI evaluated A.I.M.I.® 1.0 in a mock data center lab equipped with a production-grade Liebert PDX CRAC unit. The system was tested under various thermal loads and compared against the unit’s native PID-based “Auto” mode.

The results demonstrate significant efficiency gains in common operating conditions.

Transparency and the Path Forward

In the spirit of rigorous engineering and transparency, the white paper also details a known limitation. Under sustained, peak-load conditions, the current A.I.M.I.® 1.0 policy was less efficient than the traditional controller.

FLUIX AI is actively addressing this edge case, with improved peak-load handling being a primary focus for the next iteration of the system.

The A.I.M.I. roadmap includes expanding the model to holistically manage power and IT systems, enabling it to predict and mitigate demand spikes originating from workload schedulers. This will provide data center operators and grid managers with a powerful tool for reducing both operational costs and systemic risk.

About FLUIX AI

FLUIX AI builds A.I.M.I.® (Artificial Intelligence for Managing Infrastructure) to enable safer, more efficient autonomous operation of critical infrastructure.

By applying deep reinforcement learning to real-world physical systems, FLUIX AI is creating the autonomous control layer for the AI factories of the future.

Interested Press should reach out to: Info@fluix.ai

Refer to the article page to learn more and access the technical white paper

References

[1] A.I.M.I. 1.0: Deep Reinforcement Learning for Continuous Control of Data Center Cooling under Real-Time Thermodynamic and Operational Constraints. FLUIX AI. December 2025.