What if there were fewer AI data centers in the world?
PR Newswire
33m ago
Ai Focus
Frore Systems stated in a new analysis that by optimizing the cooling system and allowing AI chips to operate at lower temperatures, AI data centers can increase AI output by over 40% with the same amount of power, which may reduce the need for new facilities and alleviate pressure on the power grid and community water supply.
Helpful
No.Help

What if there were fewer AI data centers in the world?

A new analysis released by Silicon Valley unicorn Frore Systems has found that better cooling technologies can enable AI data centers to produce up to 40% more token using the same amount of electricity, which may help alleviate the pressure on power grids and water supply systems faced by the communities that host these facilities.

San Jose, California, October 13, 2026 / PRNewswire / -- The competition to build a AI data center is facing a major constraint: there is not enough available power. Frore Systems explains in its white paper " Cooling Is the Next Revolution " that one of the biggest solutions is actually right in front of us. Research has found that by optimizing the thermal management stack ( Thermal Stack ) and running AI chips at lower temperatures, data centers can produce over 40% more AI output with the same amount of electricity used. This means that to meet the global demand for AI, it may not be necessary to build as many new facilities.

Every AI answer, image, or line of code is composed of “token”, which are the intelligent basic units produced and sold by the AI system. AI data centers – which are increasingly referred to as AI factories ( AI Factories ) – exist for the purpose of producing token. Their output is not only limited by chips or funds but is also becoming more and more constrained by electricity. AI factories often have to wait for several years to obtain new utility connections, regulatory agencies are limiting their power consumption, and communities are opposing the electricity and water resources consumed by these facilities.

Research papers with Frore Systems indicate that heat is quietly eroding the effective work output that can be converted from each watt of electricity. As the temperature of AI chips increases, more electricity is lost as waste heat rather than being used for computation. This wasted electricity roughly doubles with every 10-degree Celsius increase in chip temperature. According to the analysis in the papers, for every 10-degree Celsius decrease in chip temperature, each watt of electricity can produce approximately 15% more token.

Frore Systems Founder and CEO Seshu Madhavapeddy stated: "The AI industry hopes to build more data centers. We believe that it should also ask how much additional capacity can be extracted from existing data centers. In a world constrained by electricity, not even a single watt of power can be wasted. If cooling is considered part of the performance system, rather than just a supporting component, then the same buildings, the same chips, and the same available electricity could produce significantly more token."

This paper tracks the complete path of heat throughout the thermal management stack: from the silicon chip, to the packaging, to the thermal interface materials that bond the chip to the cold plate, to the coolant, and finally to the cooling tower. Its main findings include:

  • Under the same power supply, the output of token can be increased by over 40%. A well-designed cooling system can increase the output of token per watt of power supplied to the AI factory by over 40%. Since this additional output does not require an increase in production costs, it is directly converted into profit.
  • Chips that operate at lower temperatures have a longer lifespan and fewer errors. Reducing the chip operating temperature from 95 degrees Celsius to 65 degrees Celsius can double the lifespan of GPU and reduce the risk of silent computing errors.
  • Cold plate manufacturing has always been a bottleneck. Traditional cold plates are machined from a single piece of copper, which limits their design. The LiquidJet cold plates of Frore are manufactured using semiconductor manufacturing processes, allowing for a 3D ultra-short circuit multi-level structure that can be customized for the specific power distribution of each GPU. This is something that traditional machining methods cannot achieve. The LiquidJet cold plates alone can increase power output per watt by 25%. Their latest version, LiquidJet Diamond, integrates diamond material to dissipate the heat generated by the extreme hot spots of the chip, resulting in a 35% improvement.

The impact goes far beyond the engineering aspect. If this 40% improvement is applied to the construction plans of ultra-large-scale cloud service providers ( Hyperscaler ), it could mean that what previously required the construction of 10 new AI data centers to achieve, can now be accomplished with just 7. This implies that three construction projects costing billions of dollars each would not need to be undertaken, three new grid connection points in areas with tight power supply could be avoided, and three communities would not have to bear the additional burden of electricity and water consumption associated with such large-scale facilities.

Madhavapeddy added, "Thermal management stack design has always been considered a component of facility costs in the past. However, in the context of the current large-scale construction at the AI factory, it has become a strategic decision at the board of directors level. It determines how many facilities a hyperscale cloud service provider needs to build, how quickly they can launch AI production capacity, and how to respond to the surrounding communities."

Obtaining Method

The white paper Frore Systems " Cooling Is the Next Revolution " can be downloaded for free at the following URL: https :// www.froresystems.com / cooling-next-revolution

A series of 14 accompanying blogs, which will be released soon, will be published over the next few weeks to explain the findings of the paper in plain language. The URLs are: https :// www.froresystems.com / blog / tokens-the-currency-of-ai.

About Frore Systems

Frore Systems is a pioneer in the field of advanced thermal management technology, dedicated to unleashing the performance of data centers and edge devices. The company's flagship solutions include: AirJet ®, the world's first solid-state active air cooling chip, which has been used in consumer, industrial, and IoT markets to provide higher performance in ultra-compact, silent, lightweight, dust-proof, and waterproof edge devices; LiquidJet ®, a 3D ultra-short circuit multi-stage liquid cooling cold plate for data centers; as well as LiquidJet ® and Diamond which integrate diamond diffusion layers to enhance GPU performance, increase AI throughput, improve PUE, and reduce TCO; there is also LiquidJet ® and Nexus, which is a lightweight integrated cold plate system that integrates multiple LiquidJet, eliminating all hoses, connectors, and manifolds, thus supporting 1/2U computing trays. The patented cooling technology of Frore Systems has been integrated into products of major OEM manufacturers and system builders worldwide. Headquartered in Silicon Valley with manufacturing operations in Taiwan, the company is redefining thermal architecture for the AI era. For more information, please visit: www.froresystems.com.

Media contact (interview by CEO and more information)

Sue Ryan - Frore Systems

[ email protected ]

Tip
$0
Like
0
Save
0
Views 21
WalletJYS reminds readers to view blockchain rationally, stay aware of risks, and beware of virtual token issuance and speculation. All content on this site represents market information or related viewpoints only and does not constitute any form of investment advice. If you find sensitive content, please click“Report”,and we will handle it promptly。
Submit
Comment 0
Hot
Latest
No comments yet. Be the first!
Related
U.S. stocks at a critical watershed; Wall Street generally bets that it will ultimately benefit the bulls
This CNBC column states that there is a rare divergence in the U.S. stock market: the S&P 500 is approaching record highs, but the median stocks in the index are still experiencing a significant pullback. Many strategists on Wall Street believe that the recent improvement in market breadth, the correction of tech stocks, and the performance during the earnings season may ultimately benefit more stocks.
CNBC
·2026-10-13 19:12:35
4
The Ruth Porat of Alphabet brings a pragmatic optimism to the AI competition, while also reminding us of potential downside risks.
During the Fortune Most Influential Women Summit, Alphabet President and Chief Investment Officer Ruth Porat stated that the company will invest approximately $200 billion this year in AI capacity building, with even more planned for 2027. She emphasized that without jointly mitigating the downward risks of AI, it will be impossible to seize the upward opportunities. The article also touched on topics such as wealth inheritance, female leadership, and McKinsey's predictions regarding the automation of working hours in the United States.
Fortune
·2026-10-13 18:52:58
15
Hello, Robotaxi has released the Generative Driving World Model HelloWorld, which integrates scene control, continuous generation, multi-sensor observation, and efficient reasoning capabilities.
News from the IT community on October 13th: Hello, Robotaxi announced yesterday the launch of the Generative Driving World Model HelloWorld. This model focuses on building driving data on demand and interactive strategy testing, integrating scene control, continuous generation, multi-sensor observation, and efficient reasoning capabilities to explore a specific path for world models to serve autonomous driving research and development.
The Block
·2026-10-13 18:42:11
17
Microsoft Launches a New Self-Developed AI Chip, Plans to Achieve Gigawatt-Level Deployment Next Year
Microsoft releases a new generation of its own-developed AI chips, with 300 units produced, and plans to deploy several gigawatts of computing power starting next year. Nadella stated that the efficiency of these chips is 40% higher than that of "mainstream GPU systems." Microsoft hopes to use this to reduce its dependence on NVIDIA GPU and lower the operating costs of Copilot and other AI products.
The Block
·2026-10-13 18:42:08
16
View More