FLUIX AI Announces Results from Autonomous AI Cooling Deployment at TenHats Data Center

In a live, mission-critical data center, an AI took control of cooling on its own, and freed up power. The deployment showed 0.4 MWh/day modeled HVAC savings, 6.6% PUE improvement, and a projected path to up to ~130 kW of potential power headroom at full-site scale.


SAN FRANCISCO, CA and KNOXVILLE, TN — June 18, 2026 — Most AI in data centers only watches and advises. At TenHats, a mission-critical facility in Knoxville, Tennessee, FLUIX let its A.I.M.I.® platform do something rarer: take direct control of the site’s cooling, autonomously, inside guardrails the operators set. Over a multi-month run, A.I.M.I. held zero SLA excursions while cutting cooling waste, and in doing so surfaced something the industry is desperate for. By making cooling more efficient, FLUIX freed up power the facility was already paying for, the equivalent of finding a small power plant hiding inside an existing building. The work was supported by TVA EnergyRight as part of an effort to evaluate how AI-driven cooling optimization can help data centers reduce energy waste, improve facility efficiency, and support more flexible, sustainable growth across the Tennessee Valley.


The simplest way to describe A.I.M.I. is autopilot for data center cooling: the system flies, operators supervise, and hard limits are enforced at every step. It was deployed as an on-premise control layer integrated with TenHats’ existing building management and HVAC systems, zero new sensors, no rip-and-replace, and no cloud dependency in the control loop. A.I.M.I. learned the site’s thermal behavior in under two weeks, then began predicting cooling needs and acting earlier, cutting the lag between IT load changes and cooling response from roughly 15 minutes to roughly five.



Preliminary analysis showed 0.4 MWh/day of modeled HVAC savings during full A.I.M.I. control periods and a 6.6% PUE improvement, moving from 1.36 to 1.27. At TenHats’ current operating profile, that represents roughly 17 kW of average modeled cooling reduction. Projected to a fully loaded 2 MW facility profile, the same PUE improvement points to up to ~130 kW of potential facility power headroom.


At typical colocation economics of approximately $300/kW-month, that level of potential headroom could represent roughly $470,000 to $500,000 per year of illustrative incremental capacity value. This figure is a full-site projection based on observed PUE improvement, not directly measured incremental revenue. Full-load validation is a focus of FLUIX’s 2026 deployments.


“Data center flexibility is becoming a grid and infrastructure priority, but not every facility can safely flex compute workloads today. The industry is spending billions to build new power and chase exotic sites; our result suggests a faster path may be the capacity already stranded inside the facilities we have,” said Abhishek Sastri, Co-Founder and CEO of FLUIX AI. “Cooling is one of the most approachable ways to make data centers more flexible. Every data center has cooling systems, and most are still operated reactively. By using AI to coordinate cooling earlier and more intelligently, we can help existing facilities reduce waste, create power headroom, and support more compute without major retrofits.”


As AI workloads drive rapid growth in data center demand, operators are increasingly constrained by power, cooling, and the ability to safely support higher compute density. FLUIX’s approach uses human-in-the-loop AI control, physics-informed modeling, human-action learning, and deep reinforcement learning to optimize cooling systems within operator-approved guardrails. Reactive PID control, the decades-old standard in most facilities, chases load that has already moved; A.I.M.I. moves before it.


“This initiative allows us to reduce energy costs by curbing peak demand, strengthen compliance with industry and operational requirements, and unlock additional capacity to support new revenue opportunities,” said Wade Orloski, Data Center Manager at TenHats. “We look forward to continuing our work with FLUIX. This is only the tip of the iceberg in terms of the potential AI brings to data center efficiency, compliance, and performance.” Following the deployment, TenHats has contracted with FLUIX to keep A.I.M.I. in continuous autonomous control of its cooling systems.


TenHats is a 2 MW data center built for AI-ready, mission-critical operations, combining redundant power and cooling, 24/7 security, on-site technical support, and enterprise-grade compliance. The facility served as a live production environment to evaluate how AI control can reduce cooling inefficiency while maintaining reliability.


TVA EnergyRight supported FLUIX’s research and deployment of AI-driven cooling optimization to evaluate how data centers can reduce energy waste, improve facility efficiency, and support more flexible, sustainable growth across the Tennessee Valley.

“Cooling efficiency is no longer just a sustainability lever,” Sastri added. “For data centers, every watt wasted on avoidable cooling is a watt that cannot be used for compute. FLUIX is building the AI control layer that turns efficiency into capacity.”


FLUIX is continuing to evaluate results and expand validation of A.I.M.I. across data center environments. The company is focused on applying autonomous infrastructure intelligence to help energy-constrained data centers reduce cooling waste, improve operational responsiveness, and create potential capacity for additional IT load.


Results Summary

  • Signed continuation: TenHats has contracted to keep A.I.M.I. in continuous autonomous control of its cooling

  • 0.4 MWh/day modeled HVAC savings during full A.I.M.I. control periods 

  • 6.6% PUE improvement, moving from 1.36 to 1.27 

  • 0 new sensors required for deployment 

  • 0 SLA excursions during full A.I.M.I. control periods 

  • Up to ~130 kW potential facility power headroom projected at a fully loaded 2 MW facility profile 

  • ~$470,000-$500,000/year illustrative incremental capacity value, assuming $300/kW-month colocation economics and full-site validation



About FLUIX

FLUIX Inc. (operating as FLUIX AI) is a startup based in San Francisco, helps data centers get more compute online faster by using AI to optimize cooling, power, and token usage, unlocking stranded capacity from the infrastructure they already have.


About TenHats 

TenHats operates mission-critical data center infrastructure designed for reliability, redundancy, security, and high-performance digital operations. Its Knoxville facility supports AI-ready and enterprise workloads with redundant power and cooling, 24/7 security, on-site technical support, and enterprise-grade compliance. Website: https://tenhats.com/


Media Contact | FLUIX Inc. | info@fluix.ai | www.fluix.ai 

FLUIX is selecting data center facilities for 2026 full-load validation deployments. Operators interested in an assessment can contact the team at info@fluix.ai.