AM ThermaLink — Turn AI Heat Into Zero-Cost Heating

The System

One sealed system that generates AI revenue — and captures every watt of heat it creates

Inside AM ThermalLink, liquid-cooled CPU/GPUs run AI workloads while recovering their waste heat as hot water. The same infrastructure delivers premium compute services and usable building heat from a single energy input.

AM ThermaLink liquid-cooled GPU unit with side panels open

8×Liquid-cooled RTX Pro Blackwell 6000 GPUs per stack

~85%Of server heat captured & reused

2Income streams — compute + heat

~60°CHot water delivered to the building

The Problem

The current data center model wastes value by paying to generate heat — then paying again to cool it down.

AI demand is exploding

AI is exploding — and so is the compute behind it. The global stock of AI computing power is growing roughly 3.4× per year, doubling about every 7 months (Epoch AI), and demand keeps climbing higher.

~100% becomes heat

Nearly all power a server draws is converted straight into heat.

Pay twice, gain nothing

Centers pay to make the heat, then pay again to cool it down.

Buildings still pay too

Nearby buildings and pools buy gas or electric heat they already need.

And new capacity is only getting harder to build. Power grids are maxing out, permits stall for years, and communities increasingly push back — so adding data-center capacity can't keep pace with demand.

The Solution

Capture the heat at the source — and put it to work

Our system captures heat from liquid-cooled GPU/CPU servers and transfers it directly into existing water-based heating systems.

AI Servers

GPU / CPU compute

Liquid Cooling

Direct-to-chip

Heat Exchanger

Loop separation

Hot-Water Tank

Thermal storage

Building / Pool

Useful heating

Heat becomes a product, not a problem

Instead of rejecting server heat with expensive chillers, we transfer it into water at up to ~60°C — ready for pools, domestic hot water, and hydronic building heating.

Up to ~60°C water ~85% heat recovery Water-based loops

Decentralized by design

The system is placed directly inside or beside the building or pool it heats. Generating heat at the point of demand eliminates long-distance distribution and minimizes thermal losses — something centralized mega data centers physically can't do.

On-site install Minimal heat loss Closer to the grid edge

Key Numbers

What a representative 50 kW AI compute system delivers

Engineering and economic modeling for a reference deployment. Figures are model-based assumptions, not guaranteed outcomes.

297,840 kWh/yr

Usable recovered heat

Annual thermal energy delivered to the host site.

~85%

Heat-recovery efficiency

Modeling assumption for captured vs. generated heat.

~60°C

Hot water output

Suitable for pools, domestic hot water & hydronic heat.

~65 t CO₂

Avoided per year

By displacing gas / electric heating demand.

5–10

Apartments fully heated

Full space heating for typical ~80 m² apartments (≈3–6 kW each).

50 kW

Reference compute block

Representative AI compute system size modeled.

Potential compute revenue

~$897K – $1.35M / yr

From renting dedicated GPU capacity (figure includes electricity cost).

Estimated installed cost

~$955K – $1.39M

One-time capital for compute hardware, cooling, and heat-recovery integration.

Compute revenue figures do not include variable operating expenses — maintenance, networking, software licensing, insurance, staffing/engineering, taxes, permitting, and hardware-replacement reserves — estimated at roughly $150K–$300K/year.

Business Model

One system. Two revenue logics. Lower costs on both sides.

AI compute revenue

Rent dedicated GPU capacity to companies, researchers, and AI developers needing private, high-performance compute.

Cooling & infrastructure savings

Capture heat directly through liquid cooling instead of relying only on expensive HVAC, chillers, and traditional cooling.

Heating-cost reduction

Host sites reuse recovered heat for pools, domestic hot water, or building heating — cutting their energy bills.

Better total efficiency

The same electricity produces both AI compute and usable thermal energy — two outputs from one input.

Lower infrastructure cost

The compute is placed directly at the host site, so there's no need for large cooling plants, long heat-transport pipes, or dedicated data-center buildings.

→

Net effect

Compute pays the bills; recovered heat becomes a second, sellable product the host actually wants.

Instead of paying large cooling and HVAC costs to reject server heat, our system captures that heat at the source and redirects it into useful heating infrastructure — reducing cooling needs while lowering pool or building energy expenses.

Best Use Cases

Where dual-use heat creates the most value

The ideal host has steady, year-round heat demand sitting next to a need for compute.

Best early market

Swimming pools

Pools need steady heat year-round, making them the strongest, most consistent thermal off-taker.

High volume

Apartment buildings

Reliable demand for domestic hot water and hydronic heating across many units.

Dual demand

Universities & hospitals

Strong fit — they need both high-performance computing and continuous heating.

Heating-led

Cold cities

Higher heating loads make recovered building heat especially valuable through long winters.

Expansion market

Hotels & resorts

Pools, spas, and constant hot-water demand make hospitality a natural expansion market.

Technology

Plugs into the infrastructure you already have

AM ThermaLink captures heat at the servers and hands it off — through one heat exchanger — into the building's existing 60 °C tank, radiators and pool.

Swipe to explore the full diagram

The computer loop and the building/pool water loop remain physically separated through a heat exchanger — for safety, water quality, and reliability. Heat crosses; fluids never mix.

Market Opportunity

Sitting at the intersection of five growing markets

AI infrastructure demand is growing rapidly

Companies, universities, hospitals, and research organizations increasingly need high-performance computing for AI, ML, automation, and data processing — driving strong demand for dedicated GPU infrastructure and private AI computing services.

Large data centers are getting harder to build

New facilities require huge electricity, cooling, grid access, land, and permitting. Power scarcity and local opposition are now major barriers — opening the door for smaller, distributed AI infrastructure placed where energy and heat can be reused.

Energy cost is a major pressure point

GPU-intensive workloads consume enormous power, and cooling adds even more cost because nearly all of it becomes heat. Reusing that heat improves overall efficiency and cuts wasted energy.

Sustainability pressure is increasing

Governments, investors, and customers demand lower carbon. Traditional data centers reject heat into the environment; heat recovery lets AI infrastructure support decarbonization by replacing gas or electric heating demand.

AI InfrastructureClean EnergyBuilding HeatingUrban Energy SystemsDecarbonization

“We are not only building a data center. We are building useful urban energy infrastructure powered by AI demand.”

Competitive Advantage

Where AI Compute Becomes a Local Energy Asset

Dual-use infrastructure: compute + heat

Lower wasted energy

Better ESG positioning

Local install close to heat demand

Lower cooling cost

Lower infrastructure cost

Fits pools, apartments, hotels, hospitals

Heat becomes revenue, not waste

⚙️ Traditional data center

Compute Sells raw GPU/CPU capacity
Waste heat Rejected via chillers & HVAC
Cost Pays twice — to make heat, then to remove it
Location Remote, far from heat demand

♻️ The AM ThermaLink model

Compute Sells dedicated GPU capacity
Recovered heat Captured at source, sold as heating
Local value Cuts host heating + our cooling bills
Location On-site, beside the heat demand

Traction & Status

Early-stage, but the groundwork is done

Completed

Research

Technical foundation and literature analysis documented.

Completed

Thermodynamic model

Heat-recovery efficiency, flow, and temperature behavior modeled.

Completed

Economic model

Capital cost, compute revenue, and payback scenarios built.

Completed

Target pilot sites identified

Candidate residential / pool sites in Southern California shortlisted.

In progress

Seeking pilot partners & investors

Raising the pre-seed round to build and validate the first system.

Currently seeking funding for prototype / pilot deployment.

Investor Ask

$350,000

Pre-Seed Funding

To design, build, and validate the first heat-recovery-enabled AI computing pilot system.

1 Pilot Objectives

Acquire AI computing hardware & supporting infrastructure
Develop & integrate the liquid-cooling and heat-recovery system
Deploy the first commercial pilot installation
Establish the first customer partnership (pool in Los Angeles)
Collect real-world operational, thermal & economic data
Demonstrate commercial viability of AI-powered heat recovery

2 The Pilot Validates

Continuous AI compute operation
Heat-recovery efficiency
Pool-heating performance
Energy-cost reduction
Customer demand & commercial scalability

Pilot demonstration: for a 6 kW system (8 GPUs), supply continuous thermal energy to a typical 32 m² (344 ft²) residential pool in Los Angeles — reducing the pool's heating costs by up to 95% while keeping it at a near temperature of 32°C (89.6°F) year-round.

$350K

Estimated pilot system cost

1.5–2 yrs

Estimated pilot payback period

0.7–1.5 yrs

Projected payback at commercial scale

AI ComputingClean EnergyBuilding & Pool HeatingUrban Decarbonization

We are not only building a data center. We are building useful energy infrastructure powered by AI demand.

One sealed system that generates AI revenue — and captures every watt of heat it creates

The current data center model wastes value by paying to generate heat — then paying again to cool it down.

AI demand is exploding

~100% becomes heat

Pay twice, gain nothing

Buildings still pay too

Capture the heat at the source — and put it to work

AI Servers

Liquid Cooling

Heat Exchanger

Hot-Water Tank

Building / Pool

Heat becomes a product, not a problem

Decentralized by design

What a representative 50 kW AI compute system delivers

Potential compute revenue

Estimated installed cost

One system. Two revenue logics. Lower costs on both sides.

AI compute revenue

Cooling & infrastructure savings

Heating-cost reduction

Better total efficiency

Lower infrastructure cost

Net effect

Where dual-use heat creates the most value

Swimming pools

Apartment buildings

Universities & hospitals

Cold cities

Hotels & resorts

Plugs into the infrastructure you already have

Sitting at the intersection of five growing markets

AI infrastructure demand is growing rapidly

Large data centers are getting harder to build

Energy cost is a major pressure point

Sustainability pressure is increasing

Where AI Compute Becomes a Local Energy Asset

⚙️ Traditional data center

♻️ The AM ThermaLink model

Early-stage, but the groundwork is done

Research

Thermodynamic model

Economic model

Target pilot sites identified

Seeking pilot partners & investors

Pre-Seed Funding

1 Pilot Objectives

2 The Pilot Validates

Let's build the first pilot together