Glacian sells $1.91M of a $5M round for data-center cooling AI

Why it matters

AI infrastructure is running into power limits faster than utilities can add capacity. Glacian is betting that cooling software can recover usable megawatts inside existing facilities, turning university research into an infrastructure product with a short path to measurable value.

Glacian Technologies has sold $1.91 million of securities in a $5 million offering, giving co-founders Jie Zhao and Wangda Zuo their first institutional-scale financing as they turn years of university research into software for data-center cooling.

Glacian Technologies disclosed the financing in a Form D filed with the SEC on September 17th. The filing lists eight investors and $3.09 million remaining in the offering. Glacian recorded the first sale on September 10th under Rule 506(b).

The $1.91 million figure requires some unpacking. Glacian said the total includes the conversion of $150,000 in simple agreements for future equity, or SAFEs, previously disclosed in an August 4th filing. That earlier money is part of the new total, so the September filing should not be read as $1.91 million of fresh cash arriving this month.

Glacian did not name the buyers in the filing. Zuo announced an investment from Samsung Next on LinkedIn on July 22nd, alongside a collaboration with Samsung Research America. The post did not state the check size, and the Form D does not establish how Samsung Next's investment maps to the disclosed proceeds. No valuation appears in the filing.

A founder pair 13 years in the making

Zhao and Zuo formed Glacian in January after following different sides of the same problem for more than a decade: Zuo developed the cooling models, while Zhao learned how to sell research-driven building technology.

The two met at a conference about 13 years ago, when Zhao was pursuing a doctorate at Carnegie Mellon University and Zuo was teaching at the University of Miami, Zhao told Technical.ly. They stayed in contact as Zuo continued his academic work and Zhao spent a decade at Delos, a New York property-technology business focused on health and buildings.

Zhao eventually led commercialization work at Delos after earlier roles teaching at the University of Pennsylvania and working at lighting-controls manufacturer Lutron Electronics. Glacian's timing followed several years of conversations between the founders and the sharp increase in electricity demand attached to AI infrastructure.

"We think this is the right timing, so I quit my job," Zhao told Technical.ly about leaving Delos in late 2025.

Zuo brought the technical base. He is a Penn State professor of architectural engineering and a fellow of ASHRAE and the International Building Performance Simulation Association. Glacian's biography says he has led nearly 40 funded research projects and published more than 170 peer-reviewed papers. His previous posts include Lawrence Berkeley National Laboratory, the University of Miami and the University of Colorado Boulder.

Glacian says the underlying work received over $8 million in research support beginning in 2013 from organizations including the Department of Energy, Department of Defense, National Science Foundation, ASHRAE, JPMorgan Chase and Penn State. That non-dilutive funding gave the founders a long technical runway before they formed a venture-backed business.

Selling recovered electricity as compute capacity

Glacian calls its software "Physical AI." The system builds physics-based digital twins of cooling equipment, then uses machine-learning models to recommend operating settings based on conditions inside and outside a facility. Glacian says the software can run locally, integrate with existing monitoring and data-center infrastructure management systems, and work across cooling hardware vendors.

The commercial pitch has shifted from saving energy costs to recovering scarce electrical capacity. Every unit of power no longer required for cooling can potentially support additional servers, allowing an operator to add compute within an existing power allocation. That framing gives Glacian a direct route into AI infrastructure budgets, where available megawatts can matter more than modest reductions in a utility bill.

The power constraint is substantial. A Department of Energy update estimated that data centers could account for 11.8% of US electricity consumption by 2030, with scenarios ranging from 9.5% to 15.3%. Building new generation and transmission can take years, leaving efficiency software as one of the faster options available to operators with facilities already connected to the grid.

Penn State researchers reported on April 6th that a trained AI agent reduced cooling energy use by more than 24% in a simulation of a Houston data center. Glacian separately promotes Department of Energy case studies showing 53% savings at a Florida facility and 74% at a Massachusetts facility.

Those percentages come from research projects and modeled or case-study environments, rather than published operating results from Glacian's current commercial product. Facilities differ sharply in climate, hardware, cooling design and baseline efficiency, so the repeatable result across customer deployments will matter more than the highest figure on the website.

The first customer is the next test

Glacian has identified Alerify, a colocation and private-cloud provider in Harrisburg, as its first customer. Penn State Engineering reported an integration agreement with Alerify in April, while Zhao described the customer signing during a Pittsburgh event this week.

The Alerify work gives Glacian a commercial environment for proving that its models can move from simulations into a commercial facility without creating new reliability risks. Data-center operators run cooling systems conservatively because an optimization error can damage costly computing equipment or interrupt customer workloads. Glacian's physics-based approach is designed to keep operational limits inside the model, giving operators recommendations that account for temperature, humidity, equipment constraints and reliability.

Zhao has been direct about the work ahead. "The key thing right now for us is customer, customer, customer," he told Technical.ly after Glacian won a $10,000 pitch competition on September 14th.

The financing gives Zhao and Zuo room to turn one deployment into a repeatable sales and implementation process. Glacian still has $3.09 million left to sell under the offering. The immediate benchmark will come from Alerify: whether a decade of funded research can produce savings that survive the operational constraints of a commercial data center.