The Wafr Platform
Three proprietary products. One closed-loop thermal operating system built for the AI era.
Store · Spend · Prove — one closed loop
A thermal battery for the data center. Charges overnight on cheap grid power, releases stored cooling on demand to absorb peak GPU heat. The chiller stays idle during the most expensive hours of the day.
Wafr's proprietary software layer. Real-time ML dispatch that forecasts load, grid price, and ambient windows 24–48 hours ahead. Makes every BTU a decision. Automates charge, discharge, rejection, and standby.
Every kilowatt-hour, every cubic meter, every gram of CO₂: measured, signed, and verifiable. Turns performance savings into bankable, registry-ready records that operators, lenders, and insurers can act on.
FRDGE is Wafr's flagship, fully integrated thermal dispatch platform: a complete engineered product that connects into the secondary cooling loop of a data center. It captures cooling capacity when grid electricity is cheap and releases it on demand to absorb the peak heat coming off high-density AI compute.
This is not a supplementary component. FRDGE is a complete thermal operating system designed to store, buffer, and dispatch cooling capacity on demand, reduce peak electrical stress, and enable near-zero water consumption through a fully closed-loop, air-cooled architecture.
Global data center electricity demand is projected to grow from 415 TWh in 2024 to 945 TWh by 2030 (IEA). Cooling accounts for 30–40% of total facility power in conventional deployments. Modern AI GPU clusters run at 40–100 kW per rack, four to seven times the density of traditional compute, and existing chiller-based systems were not designed for that thermal load. Water scarcity is compounding the pressure: evaporative cooling is being restricted or priced out across key deployment regions. Operators face a hard constraint — they cannot scale AI compute without solving cooling infrastructure first.
Thermint is the decision layer that turns FRDGE hardware into a self-optimizing dispatchable asset. It ingests live grid pricing, ambient temperature forecasts, GPU workload schedules, and facility sensor telemetry, then runs continuous optimization to determine exactly when to charge, when to discharge, when to idle the chiller, and when to reject heat directly.
Without Thermint, a thermal battery is passive hardware. With it, the facility cooling loop becomes a managed resource: predictable to operators, optimized against grid economics, and continuously responsive to the variable nature of AI workloads.
AI training and inference workloads are episodic, variable, and thermally intense. Ambient temperatures shift by 15–20°C across a 24-hour cycle in many climates. A static cooling schedule captures none of this value. Thermint re-optimizes every 15 minutes across all variables simultaneously, solving a mixed-integer optimization across all variables to find the optimal charge and discharge schedule. Every BTU becomes a deliberate economic decision rather than a thermostatic reflex.
| Module | Function | Capability |
|---|---|---|
| M1 | Site overview and command | Unified monitoring dashboard with full situational awareness across all FRDGE units, sensors, and facility systems |
| M2 | Predictive dispatch engine | Core ML optimizer that ingests facility telemetry, grid pricing, weather, and workload schedules, solves a mixed-integer optimization across the full dispatch horizon every 15 minutes, and outputs charge and discharge setpoints |
| M3 | FRDGE fleet control | Direct hardware interface that sequences charge and discharge across all TEU units and manages state-of-charge envelopes and mode transitions |
| M4 | Workload-aware cooling | Matches cooling allocation to the real GPU job mix in real time. Prevents over-provisioning during partial-load periods and reduces cycling |
| M5 | TerraQuant MRV | Real-time savings computation layer that feeds the verification, reporting, and registry-submission stack |
| M6 | Digital twin | Stress-tests dispatch schedules in simulation before committing to live hardware. Reduces risk of thermal excursions under novel load profiles |
| M7 | Reports and API | Disclosure packs, BMS, EMS, and SCADA integration endpoints, and data export for operator, investor, and regulatory reporting |
| M8 | Liquid cooling (new) | CDU and manifold control, leak detection, and direct-to-chip cooling management for high-density rack environments |
TerraQuant is the independent measurement layer that makes Wafr's performance savings bankable. It applies boundary-defined energy accounting aligned to customer acceptance criteria, producing signed M&V reports for operators, regulators, PPA counterparties, and infrastructure underwriters.
Savings claims in the cooling industry are typically self-reported and unverifiable. TerraQuant closes this gap. Every kilowatt-hour, every cubic meter of water, and every gram of CO₂ avoided is metered and independently verified.
Infrastructure lenders, power purchase agreement counterparties, and carbon registries cannot act on operator-reported performance data. They require methodology-aligned, independently validated measurement: the same standard applied to utility-scale energy projects. TerraQuant is designed to meet that bar from day one, not as a compliance add-on, but as the foundation of the commercial relationship. A financeable savings stream requires a financeable measurement system.
Verified thermal performance
Whether you are exploring AI data center strategy, advanced cooling solutions, energy optimization, or infrastructure modernization, we would be glad to connect.
Tell us a little about your project, goals, or challenge. Our team will review your inquiry and respond promptly.
We work with enterprises, utilities, technology partners, and infrastructure investors who are shaping the future of AI-ready facilities.
