New thermal cooling technology created by researchers increases semiconductor efficiency and boosts advanced chips for AI and supercomputing.
Heat is one of the main obstacles to the evolution of modern processors. The more powerful the chips become, the greater the difficulty in maintaining safe temperatures without increasing energy consumption. Now, researchers from the Korea Advanced Institute of Science and Technology (KAIST) have introduced a technology that could change this scenario.
The innovation uses microchannels thinner than a human hair embedded directly into the silicon. In tests, the new chip cooling solution managed to keep the temperature below 100°C even under thermal loads exceeding 2,000 watts per square centimeter (W/cm²). Additionally, it achieved a coefficient of performance (COP) of approximately 106,000, about 10 times higher than the previous record set in 2020.
According to a publication by Science Direct on June 15, 2026, the advancement could benefit areas such as artificial intelligence, data centers, supercomputers, and high-performance computing systems, reducing one of the biggest bottlenecks currently faced by the semiconductor industry.
-
Scientists pumped seawater to the surface of the Arctic during the third warmest winter ever recorded in the region and thickened the ice by up to 32 centimeters in the tested areas, a gain comparable to the thickness lost in 50 years.
-
At the age of 3, little Eduardo Amaral became the first Brazilian diagnosed with an ultra-rare genetic disease that affects fewer than 100 people worldwide, and he has just returned from the United States with news that seemed impossible.
-
James Webb Telescope finds possible black hole in object observed 1.8 billion years after the Big Bang and reveals more than 40 spectral lines in GLIMPSE-17775
-
NASA monitors research on plants that grow in the dark with CO₂ and electricity, a technology being studied to produce food for future missions to Mars.
How researchers reinvented water circulation inside chips
The project led by Young Lee and his team relies on a strategy different from conventional systems. Instead of having the coolant travel long distances inside the processor, the new architecture distributes water through various entry and exit points spread across the chip.
In practice, this significantly reduces the resistance to the liquid’s flow. As a result, less energy is needed to move the water and remove the heat generated by the electronic components.
The idea resembles a modern logistics network. Instead of relying on a single distribution center, multiple points work simultaneously to reduce distances and increase efficiency.
Chip cooling uses only water and dispenses with expensive materials
One of the most important differentials of the technology lies in the simplicity of the system.
In recent years, various research efforts have tried to improve chip cooling using advanced materials, nanoscale modified surfaces, and complex thermal exchange methods. Many of these alternatives present high costs and challenges for commercial application.
The solution created by the researchers follows a different path. The system uses only room temperature water circulating through carefully designed microchannels.
This can facilitate the adoption of the technology by the industry, especially in applications that require scale and controlled costs.
Advanced chips face increasingly severe thermal limits
The race for more performance has increased the pressure on processor manufacturers.
With the expansion of artificial intelligence and big data analysis, advanced chips need to perform more and more operations in reduced spaces. This increase in computational density generates significant amounts of heat.
When the temperature rises beyond the ideal, several problems can arise:
- Reduced performance;
- Increased energy consumption;
- Accelerated wear of components;
- Risk of operational failures;
- Shorter equipment lifespan.
For this reason, efficient heat dissipation has become a priority for the entire semiconductor industry.
Intelligent optimization boosted semiconductor performance
The success of the research does not depend solely on the presence of the microchannels. The KAIST team conducted extensive optimization work to find the most efficient configuration possible.
The researchers analyzed factors such as:
- Channel width;
- Channel height;
- Number of microchannels;
- Internal distribution;
- Coolant flow rate;
- Pressure required for circulation.
To arrive at the final model, multifidelity optimization methods were used, allowing the exploration of thousands of combinations before the prototype was manufactured.
After the simulations, the structure was produced on a real silicon semiconductor chip and subjected to experimental tests.
Thermal cooling surpasses world record set in 2020
The results obtained caught the attention of the scientific community.
According to the data released by the team, the system achieved a performance coefficient of approximately 106,000. The indicator measures the efficiency of heat removal in relation to the energy consumed by the pumping system.
For comparison, the best technology previously presented in 2020 recorded a COP close to 10,000.
In practice, this means that the new method requires only about one-tenth of the pumping power to remove the same amount of heat.
Besides improving energy efficiency, the system also provides a more uniform thermal distribution across the processor’s surface.
High-performance computing increasingly depends on thermal solutions
The advancement comes at a strategic time for the industry.
High-performance computing applications require processors capable of operating continuously under intense loads. This scenario includes supercomputers, advanced artificial intelligence models, scientific simulations, and engineering systems.
In these platforms, heat is already considered one of the main limiting factors for new performance gains.
The more efficient the thermal control, the greater the potential to increase processing capacity without proportionally increasing energy consumption.
Researchers see applications in AI, data centers, and advanced electronics
Professor Sung Kim, a member of the team responsible for the study, highlighted that the technology was developed with future demands of semiconductors aimed at artificial intelligence and next-generation electronic systems in mind.
The expectation is that the solution will contribute to the creation of faster, more compact, and efficient equipment.
Among the sectors that could benefit are:
- Generative artificial intelligence;
- Large-scale data centers;
- Scientific computing;
- Advanced graphics processors;
- Cloud infrastructure;
- High-performance research systems.
As the demand for processing grows, technologies of this type tend to gain strategic relevance.
An important step for the next generation of processors
The work developed by researchers at KAIST shows that there is still room for significant advances in the thermal management of electronic components.
By maintaining temperatures below 100°C even under loads exceeding 2,000 W/cm² and achieving an efficiency about 10 times greater than the previous record, the technology demonstrates potential to redefine the standards of chip cooling.
Combining smart engineering, low energy consumption, and the use of water as a cooling fluid, the innovation can help remove one of the biggest barriers to the evolution of advanced chips, modern semiconductors, and high-performance computing in the coming decades.

Be the first to react!