%41%49%20%61%67%65%6e%74
-
KAIST Identifies the “Hidden Energy Cost” of AI Agents for the First Time
As the era of AI agents—systems that can reason and act autonomously—begins, the power consumption of data centers is emerging as a critical challenge. A KAIST research team has, for the first time, analyzed the computational cost and energy consumption of AI agents, finding that they can consume up to 136.5 times energy per query than conventional generative AI. The study shows that competitiveness in the AI era is expanding beyond model performance to include the efficiency of data centers and power infrastructure.
KAIST announced that a research team led by Professor Minsoo Rhu of the School of Electrical Engineering has systematically analyzed, for the first time, how much computational resources and power AI agents require in real-world service environments.
Large language model (LLMs) powered applications such as ChatGPT have rapidly evolved beyond simply answering questions. They are now developing into AI agents: next-generation AI systems that can plan, use external tools such as web search, calculators, and code execution environments, and solve complex tasks by coordinating multiple steps on their own.
Although AI agents are increasingly being adopted in areas such as software development, research, and workplace automation, little has been known about the amount of electricity and operational cost required to run them in practice.
The research team defined AI agents not merely as software programs, but as a new type of workload that must be continuously processed by data-center servers and graphics processing units, or GPUs—high-performance chips used for large-scale AI computation. The team then analyzed the computational load and energy consumption incurred during actual AI agent execution.
The analysis found that AI agents perform, far higher volumes of LLM invocations than conventional chain-of-thought reasoning. Chain-of-thought, or CoT, refers to a method in which an AI model breaks down its reasoning process step by step to reach an answer, while an LLM invocation refers to each computational request made to a language model to generate a new judgment or response.
Because AI agents repeatedly call language models during execution, their response latency also increases significantly. The team found that response time can increase by up to 153.7 times, while GPUs remain idle for as much as 54.5 percent of the total execution time as external tools perform their tasks. In other words, as AI systems take on more complex tasks, a new form of inefficiency emerges in which expensive GPUs cannot be fully utilized.
The research team also analyzed the power consumption of AI agents at data-center scale. An AI agent using a 70-billion-parameter LLM—a scale comparable to current commercial AI services—consumed an average of 348.41 watt-hours per query. This is 136.5 times higher than the energy consumed by a conventional generative AI system performing simple question answering.
In addition, the team projected a future scenario in which 13.7 billion AI agent requests are generated per day — a volume equivalent to current Google search traffic. Under this scenario, data-center power demand would reach approximately 198.9 gigawatts, a level far exceeding the scale of AI data centers currently under development (which are in the range of a few gigawatts) and equivalent to roughly half of the average power consumption of the United States.
This study demonstrates that the focus of competition in the AI era is shifting from “smarter AI” to “optimally efficient AI.” Going forward, it will be essential not only to advance AI models, but also to jointly optimize AI semiconductors, data centers, and power infrastructure through co-design. Such an approach is expected to become a key strategy for reducing the operating cost of AI services and building sustainable AI infrastructure.
“This study is the first to quantitatively show not only how AI is becoming more intelligent, but also how much electricity and cost are required to implement and sustain that intelligence,” said Professor Rhu. “As AI agents become widespread, it will become increasingly important to take an integrated co-design approach that optimizes not only AI data-center infrastructure, but also AI agent models and power infrastructure.” He added, “Research and investment in this direction will be essential to dramatically reduce the cost for end users to access AI services while building sustainable AI infrastructure.”
The study was conducted with Jiin Kim, a Ph.D. student in the KAIST School of Electrical Engineering, as the first author. The paper was presented in February at the 32nd IEEE International Symposium on High-Performance Computer Architecture, or HPCA, one of the most prestigious international conferences in computer system design. The research team has also released the AI agent implementations and benchmarks used in the paper as open source to support follow-up studies by researchers worldwide.
Paper title: “The Cost of Dynamic Reasoning: Demystifying AI Agents and Test-Time Scaling from an AI Infrastructure Perspective”
Open-source repository: 10.1109/HPCA68181.2026.11408569
This research was supported by the Institute of Information & Communications Technology Planning & Evaluation (IITP) through the SW Starlab program, the K-Cloud Technology Development Program using AI semiconductors, and the Leading Technology Development Program for Advancing AI-Semiconductor-Based Data Centers, as well as by the Samsung Electronics Future Technology Incubation Center.
2026.07.07 View 412 -
“Entrepreneurial Mutual Growth Fair 2026” to be Held... KAIST Super Star Companies Gather for AI Solopreneurship, Tech Commercialization, Investment, and Youth Job Fair
KAIST announced that it will host the ‘AI Agent-Based Solopreneurship Program Information Session’ and the ‘Entrepreneurial Mutual Growth Fair 2026’ for two days from May 18th to 19th.
In this event, KAIST’s new AI-based solopreneurship model, which utilizes AI not merely as an operational tool but as a ‘Co-founder,’ will be introduced in depth. The university will hold an information session for the ‘AI Solopreneur Support Project,’ which enables a single individual to carry out the entire entrepreneurial process—including planning, development, marketing, and fundraising—using AI agents prepared by the university.
In this program, 100 prospective entrepreneurs will be selected nationwide, and faculty from the KAIST Institute for Entrepreneurship and the KAIST Graduate School of AI will provide eight weeks of intensive training. Additionally, a network of top-tier domestic and global mentors will be established to support business optimization and linkage with overseas investments.
In particular, outstanding teams will be provided with seed investment of up to 100 million KRW, prototype production support, and infrastructure for GPU and AI services. By fostering world-class AI utilization skills in prospective entrepreneurs with diverse domain knowledge, KAIST plans to accelerate the introduction of AI into various domestic industries while nurturing AI business models with global competitiveness.
This event is organized as a venue to introduce the KAIST-style full-cycle entrepreneurial ecosystem, encompassing artificial intelligence (AI)-based entrepreneurship, technology commercialization, industry-academic cooperation, investment linkage, and youth job creation. In particular, it will showcase the competitiveness of the deep-tech (advanced technology-based) startup ecosystem from multiple perspectives, focusing on the technological prowess and industrial application cases of KAIST startup companies.
Global big tech companies' choice of AI solution providers will also participate to reveal various technologies reflecting the AX (AI Transformation) trend across industries. Actual application cases that supported the digital transformation of major domestic corporations through factory and office automation solutions will also be announced.
In the field of robotics, Lion Robotics will introduce field-application technologies based on quadruped robots and leading R&D cases for humanoid robots. In addition, next-generation AI semiconductor startups such as Panmnesia and HyperAccel will present next-generation chip design technologies for implementing On-Device AI. These companies will showcase technologies and business models that can run Large Language Model (LLM)-based AI services faster while reducing dependence on GPUs (Graphics Processing Units). In the deep-tech bio and healthcare AI field, Barreleye will introduce an innovative solution that complements the limitations of traditional MRI (Magnetic Resonance Imaging)-centered diagnosis through AI-based quantitative ultrasound analysis technology. In the bio and medical robot field, Roen Surgical will present next-generation medical innovation cases based on precision surgical robot technology.
On the first day of the event, May 18th, the ‘Entrepreneurial Mutual Growth Fair’ will be held in the main hall on the 1st floor of the KI Building along with the ‘AI Agent-Based Solopreneurship Program Information Session.’ Representative startup companies that have led KAIST’s technology commercialization success will participate in this session to share successful technology commercialization models that connected R&D achievements to actual market results.
Through this, they plan to present a virtuous cycle for the KAIST startup ecosystem leading from ‘Research → Startup → Investment → Growth.’ Furthermore, KAIST startup companies will operate recruitment sessions alongside technology exhibitions. Participating companies will conduct direct recruitment consultations and talent discovery on-site, providing youth with high-quality, technology-based job opportunities. Through this event, the university plans to support scientific and technological talents so they can advance into startups and industrial fields rather than staying in research, and to lead technology-based entrepreneurship and employment creation. On the second day, May 19th, an ‘Open Innovation Information Session’ will be held to connect KAIST’s research capabilities with industrial demand.
At the event, the ‘1 Lab N Startup’ model, which connects KAIST faculty’s technology with corporate R&D needs to promote joint research and commercialization, will be introduced. Industry-academic cooperation strategies that expand beyond technology transfer to joint entrepreneurship and new business creation will also be announced. Following this, in the ‘KAIST Startup Investment Linkage IR Pitching Session,’ the investment attraction program ‘Tech Plaza’ will be operated, featuring five Korean deep-tech bio companies. Companies selected based on the KAIST Startup Platform (KSTP) will present their business models and technological prowess to investors, and tangible investment results are expected through linkage with venture capital (VC) and accelerators. Bae Hyeon-min, Dean of the KAIST Institute for Entrepreneurship, said, “This Entrepreneurial Mutual Growth Fair is an integrated startup platform that connects the entire process from AI-based individual entrepreneurship to technology commercialization, industry-academic cooperation, investment, and job creation.
We expect it to serve as an opportunity to present a new direction for the domestic deep-tech startup ecosystem through the success stories of KAIST’s representative startups.” This event is open to students, the general public, corporations, and investment institutions interested in entrepreneurship, and is prepared as a place to directly confirm the innovative achievements and expansion possibilities of the KAIST startup ecosystem. Information regarding the KAIST AI Solopreneurship Program information session and participation applications can be found on the website (https://www.kaist-overedge.com/).
By accessing the website, people can watch the information session on YouTube and apply for participation.
2026.05.13 View 2588