Published February 11, 2026

Oracle Cools AI Servers Using Water That Never Needs Changing

Oracle's AI data centers utilize a closed-loop cooling system where water circulates without evaporation or refills: it is filled just once.

Infrastructure
Event Source: Oracle Reading Time: 4 – 5 minutes

Challenges of Cooling Modern AI Infrastructure

The Problem Few Talk About

When discussing AI infrastructure, people usually talk about chip performance, model training speeds, or compute costs. But there is one topic that remains in the shadows, even though it directly impacts the scalability of the entire industry – cooling.

Modern AI accelerators generate huge amounts of heat. If not removed effectively, equipment overheats, throttles performance, or fails. Classic air cooling can no longer keep up with this: too much energy goes into running fans, and efficiency drops as server density increases.

That is why the industry is gradually shifting to liquid cooling. Water conducts heat much better than air, but this approach has its own complications: you have to constantly replenish the system due to evaporation, filter the water, and monitor its quality. All this increases operating expenses and complicates maintenance.

Oracle Closed-Loop Liquid Cooling Technology

What Oracle Did

Oracle has implemented a cooling system in its AI data centers that operates on a closed-loop principle. Simply put, water circulates inside the system, directly cooling the chips, and does not evaporate. It is filled once, after which it simply circulates, carrying away heat.

Technically, this is called «direct-to-chip closed-loop non-evaporative liquid cooling». The gist is that fluid is delivered right to the processors and accelerators, picks up heat, and returns to the system without contacting the outside environment. No evaporation means no need to add water.

Benefits of Non-Evaporative Liquid Cooling

Why It Matters

First and most obvious is resource conservation. Traditional cooling systems in large data centers can consume millions of liters of water a year. A closed system doesn't require this: fill it once, and the water works for years.

Second is lower operating costs. There is no need to constantly monitor fluid levels, filter it, or treat it with chemicals. The system becomes easier to maintain.

Third is relevance for regions with limited access to water. Data centers are often built where electricity is cheap, but there may be issues with water resources. A closed system removes this constraint.

And finally, the question of equipment density. Liquid cooling allows you to pack more servers per square meter since it removes heat more effectively. And if the system also doesn't require constant replenishment, it significantly simplifies design and operation.

Is This Technology Unique?

No, Oracle is not the only company using liquid cooling. Microsoft, Google, Meta, and other major players are experimenting with different variants. Some submerge servers in dielectric fluid, while others use water loops with external chillers.

But the closed non-evaporative system is a specific solution that emphasizes autonomy and minimizing maintenance. Oracle highlights that their approach requires filling the water only once, and this is a key difference from systems needing regular top-ups.

Impact on AI Infrastructure and Operating Costs

What This Means for the Industry

The cooling technology itself isn't a breakthrough in AI. It doesn't make models smarter or directly speed up training. But it solves an infrastructure problem that is becoming increasingly critical.

As the computing power needed to train and run large models grows, cooling is turning from a secondary issue into a key factor determining the cost and availability of AI services. If a data center consumes less water and energy for cooling, it lowers the final cost of compute.

For developers and companies renting cloud capacity, this can mean more predictable pricing and fewer restrictions tied to resource availability in specific regions.

Technical Challenges and Scaling of Closed-Loop Systems

Open Questions

Oracle hasn't revealed all the details of its system. It is unclear exactly what heat transfer fluids are used, how the loop is organized, or the longevity of the components. This is important, as any closed system can eventually face fluid degradation, corrosion, or impurity buildup.

It is also unknown how effectively this technology scales. Implementing it in a single data center is one thing; rolling it out globally, accounting for different climatic conditions and regulatory requirements, is another.

However, the very fact that corporations are investing so actively in alternative cooling methods shows that the question of infrastructure is becoming just as important as the question of algorithms. And solutions that seem like technical details today might determine tomorrow who can afford to train and launch the next generation of models.

Original Title: Closed-loop cooling in Oracle AI data centers
Publication Date: Feb 11, 2026
Oracle www.oracle.com Global technology corporation developing cloud infrastructure, databases, and AI services for enterprise use.
Previous Article Oracle Integrates AI Agents into Supply Chain Management System Next Article Managing AI Agent Prompts: Alibaba Cloud Unveils a Tool to Handle Them as Configurations

From Source to Analysis

How This Text Was Created

This material is not a direct retelling of the original publication. First, the news item itself was selected as an event important for understanding AI development. Then a processing framework was set: what needs clarification, what context to add, and where to place emphasis. This allowed us to turn a single announcement or update into a coherent and meaningful analysis.

Neural Networks Involved in the Process

We openly show which models were used at different stages of processing. Each performed its own role — analyzing the source, rewriting, fact-checking, and visual interpretation. This approach maintains transparency and clearly demonstrates how technologies participated in creating the material.

1.
Claude Sonnet 4.5 Anthropic Analyzing the Original Publication and Writing the Text The neural network studies the original material and generates a coherent text

1. Analyzing the Original Publication and Writing the Text

The neural network studies the original material and generates a coherent text

Claude Sonnet 4.5 Anthropic
2.
Gemini 3 Pro Google DeepMind step.translate-en.title

2. step.translate-en.title

Gemini 3 Pro Google DeepMind
3.
Gemini 3 Flash Preview Google DeepMind Text Review and Editing Correction of errors, inaccuracies, and ambiguous phrasing

3. Text Review and Editing

Correction of errors, inaccuracies, and ambiguous phrasing

Gemini 3 Flash Preview Google DeepMind
4.
DeepSeek-V3.2 DeepSeek Preparing the Illustration Description Generating a textual prompt for the visual model

4. Preparing the Illustration Description

Generating a textual prompt for the visual model

DeepSeek-V3.2 DeepSeek
5.
FLUX.2 Pro Black Forest Labs Creating the Illustration Generating an image based on the prepared prompt

5. Creating the Illustration

Generating an image based on the prepared prompt

FLUX.2 Pro Black Forest Labs

Related Publications

You May Also Like

Explore Other Events

Events are only part of the bigger picture. These materials help you see more broadly: the context, the consequences, and the ideas behind the news.

Want to dive deeper into the world
of neuro-creativity?

Be the first to learn about new books, articles, and AI experiments
on our Telegram channel!

Subscribe