About
Madrone builds chillerless, dew-point cooling systems that provide cool facility water for AI data centers. It sells to data-center owners and operators, differentiating through claimed reductions of approximately 30% in power and water usage and by avoiding mechanical chillers.
Market
Madrone competes in data-center thermal-management and cooling infrastructure, particularly for AI/HPC workloads and hyperscale facilities where cooling consumes substantial power and water. It positions itself as a modular, chiller-less alternative that cools facility water to the dew point and can be fabricated and installed quickly, claiming roughly 30% lower power and water use than a chiller-plus-cooling-tower approach. Its competitors span immersion, direct-liquid, and broader thermal-management systems, whereas Madrone emphasizes facility-level cooling outside the data center and avoids mechanical chillers.
Madrone targets hyperscale and high-density AI/data-center developers and operators, especially those building new facilities in hot, arid regions such as Texas. Its buyers are likely data-center infrastructure, facilities, and thermal-engineering teams seeking faster deployment, higher compute density, and lower cooling power and water consumption; the company also targets operators retrofitting existing facilities.
At a Glance
Problem
Madrone addresses the power, water, and deployment constraints of cooling modern data centers, particularly AI facilities being built in hot, arid regions such as Texas. Cooling can consume up to 40% of a data center’s power that could otherwise run GPUs, while limited grid capacity and years-long interconnection or permitting backlogs make additional compute difficult to deploy. The economics are substantial: Madrone says that every 1% improvement in cooling efficiency can generate $2.6 million of additional GPU revenue per 100 MW per year, implying roughly $78 million annually for a 30% efficiency improvement.
The primary use case is new hyperscale or AI data centers in desert climates, where conventional cooling can constrain capacity and cause thermal throttling. Madrone also targets retrofits of existing facilities, allowing operators to increase compute density inside existing buildings without waiting for entirely new sites.
Product / Service
Madrone provides an external facility-water cooling system that supplies data centers with approximately 20°C water using a dew-point-based thermodynamic process rather than mechanical chillers. In Texas, the company claims its approach uses 30% less water and electricity than a conventional chiller-and-cooling-tower setup, while avoiding the installation of a chiller and potentially saving tens of millions of dollars per facility.
The system is modular and designed for rapid fabrication and installation. Madrone says it develops the electronics, heat exchangers, and software in-house, and positions the product for both new hyperscale builds and retrofit capacity expansions. The intended benefit is more compute from the same grid connection, lower operating and infrastructure costs, and better thermal performance in extreme ambient conditions.
Market
Madrone competes in data-center facility cooling, with a specific focus on chillerless, dew-point, modular cooling for high-density AI and hyperscale infrastructure. Its most direct incumbent alternative is the conventional mechanical chiller combined with a cooling tower; adjacent alternatives include liquid-cooling, evaporative-cooling, and other modular data-center cooling systems. The evidence reviewed does not identify a specific named direct competitor, so Madrone’s differentiation is best understood as its claimed combination of dew-point cooling, lower water and power consumption, and faster deployment.
The company appears to be at an early, pre-commercial or pre-revenue stage rather than having demonstrated public customer traction. Madrone was founded in 2025, is listed as an active Spring 2026 Y Combinator company with a two-person team, and says it is backed by Y Combinator and the NVIDIA Inception Program. Its public materials describe a 1 kW prototype and a 100 kW prototype under construction, while no named customers, commercial deployments, or revenue are disclosed in the available evidence.
Founders & Leadership
Funding History
Y Combinator
Recent News
Y Combinator’s 2026 Industrials startup list identifies Madrone as a P2026 company. It describes Madrone as building data-center cooling systems using 30% less power and water through dew-point cooling technology.
An AngelsRound profile describes Madrone’s system as replacing traditional chillers for AI data centers while using less power and water. The profile reports that Madrone was raising a $5 million seed round and had rapid prototype development, a paying pilot, and an early customer.
Extruct’s Y Combinator P26 batch overview describes Madrone’s external data-center cooling system as delivering 20°C supply water without installing a chiller.
The YC Tier List reports that Madrone builds data-center cooling systems capable of using 30% less power and water through novel dew-point cooling technology.
Y Combinator’s launch coverage says Madrone builds modular cooling systems that chill water to the dew point, reducing power usage and enabling fast deployment. It also says the chiller-less approach can save tens of millions of dollars per facility, with a cited potential savings figure of $78 million.
Active Roles
0No active roles right now.
Get notified when they postBusiness Model
Madrone makes money by selling and deploying cooling hardware and system installations to AI data-center owners and operators. Its public materials do not disclose specific pricing or subscription terms; the offering is positioned as facility infrastructure that can save customers substantial operating and construction costs.