[Event Report] Next-Generation Infrastructure Pioneered by AI Data Centers - The Future of Computing Power Supported by Energy and Electronic Components -
AKKODiS Consulting held an event on February 26 regarding next-gen infrastructure, focusing on the massive power demands and energy strategies surrounding AI data centers.
📋 Article Processing Timeline
- 📰 Published: April 15, 2026 at 01:00
- 🔍 Collected: April 14, 2026 at 16:31
- 🤖 AI Analyzed: April 19, 2026 at 14:57 (118h 25m after Collected)
AKKODiS Consulting Ltd. (Headquarters: Minato-ku, Tokyo, President & CEO: Kenichiro Kawasaki, hereinafter "AKKODiS"), the Japanese subsidiary of AKKODiS which deploys digital engineering consulting to accelerate corporate innovation and growth in 30 countries worldwide, supporting corporate productivity improvement and AI transformation realization through on-site transformation power and digital technology, held an event titled "Thorough Explanation by Experts in Each Field! Next-Generation Infrastructure Pioneered by AI Data Centers - The Future of Computing Power Supported by Energy and Electronic Components -" on February 26.
With the rapid spread of generative AI, the importance of "data centers" as the infrastructure supporting AI is rapidly increasing. AI, which requires massive computing power, necessitates power and cooling technologies of a different scale than conventional IT infrastructure, bringing major changes to the industrial structure as well. Under this background, experts in the fields of energy, electronic components, and data center technology took the stage at this seminar. Multifaceted discussions were held on the background of the rapid expansion of AI data centers, the evolution of foundational technologies such as power and electronic components, and Japan's strategy in the era of AI infrastructure.
At the beginning of the seminar, Takuma Tanimoto (Ph.D. in Engineering), Master Instructor of the Career Development Promotion Department, People Development Division at AKKODiS, and committee member of the Component Technology Roadmap of JEITA (Japan Electronics and Information Technology Industries Association), explained the structural changes in data centers in the AI era.
After that, Mr. Yukio Sakaguchi, Representative of the Clean Energy Research Institute, and Mr. Manabu Hozoji, Manager of the Development Planning Department, Research & Development Laboratory at TAIYO YUDEN CO., LTD., gave lectures.
■ Expansion of AI Data Centers and the Direction of Energy Strategy
Yukio Sakaguchi (Representative, Clean Energy Research Institute)
Based on the latest trends centering on Silicon Valley, Mr. Sakaguchi explained the reality of the rapid expansion of AI data centers and the power and energy strategies that support it.
What Mr. Sakaguchi emphasized first was that "To understand data centers, it is first necessary to understand the structure of the cloud business." The current cloud market is dominated by hyperscalers like Amazon, Microsoft, and Google, and most AI services are provided on top of these giant cloud platforms.
In other words, the expansion of AI data centers is not just the development of IT infrastructure, but is closely related to the competitive structure of the cloud industry itself. The competition over AI is transforming from competition not only over algorithms and semiconductor performance, but into a capital-intensive competition over how much "physical infrastructure" like data centers can be secured.
Furthermore, regarding the ongoing investments in AI data centers, Mr. Sakaguchi explained, "While it cannot be denied that there is a bubble-like aspect," he drew parallels to the past IT bubble, stating, "Even if the bubble bursts, the infrastructure itself remains in society and has supported subsequent industrial development." Just as fiber optic networks and data centers are still utilized today as the foundation of the internet industry, his view is that AI data centers will likely remain as irreversible social infrastructure in the medium to long term, regardless of short-term overheating.
On the other hand, the biggest constraint emerging on the expansion of AI data centers is the issue of power supply. The performance improvement of AI models relies on "scaling laws," where injecting more computational resources improves performance, leading to fierce competition among companies to secure GPUs and power. As a result, the power demand of data centers is rapidly increasing.
In addition, the speed of evolution of AI semiconductors is extremely fast, and the time gap required to build data centers is also severe. While GPUs undergo a generation change in about 9 months, it takes several years to design and construct a data center and prepare power transmission facilities. This time difference creates a structural difficulty where the assumptions become obsolete by the time of completion.
Furthermore, looking ahead, it is expected that the power consumption of not just AI model training, but "inference" where generative AI and business AI are constantly running, will also expand. In the future, if AI becomes established as social infrastructure, the demand for power will not be temporary but will increase continuously.
Under these circumstances, various power supply methods such as renewable energy, nuclear power, and small modular reactors (SMRs) are being discussed, but there is no definitive silver bullet that can eliminate the power constraints in the short term.
Mr. Sakaguchi concluded that "The competition in AI infrastructure is not just about computing power, but power and energy."
With the rapid spread of generative AI, the importance of "data centers" as the infrastructure supporting AI is rapidly increasing. AI, which requires massive computing power, necessitates power and cooling technologies of a different scale than conventional IT infrastructure, bringing major changes to the industrial structure as well. Under this background, experts in the fields of energy, electronic components, and data center technology took the stage at this seminar. Multifaceted discussions were held on the background of the rapid expansion of AI data centers, the evolution of foundational technologies such as power and electronic components, and Japan's strategy in the era of AI infrastructure.
At the beginning of the seminar, Takuma Tanimoto (Ph.D. in Engineering), Master Instructor of the Career Development Promotion Department, People Development Division at AKKODiS, and committee member of the Component Technology Roadmap of JEITA (Japan Electronics and Information Technology Industries Association), explained the structural changes in data centers in the AI era.
After that, Mr. Yukio Sakaguchi, Representative of the Clean Energy Research Institute, and Mr. Manabu Hozoji, Manager of the Development Planning Department, Research & Development Laboratory at TAIYO YUDEN CO., LTD., gave lectures.
■ Expansion of AI Data Centers and the Direction of Energy Strategy
Yukio Sakaguchi (Representative, Clean Energy Research Institute)
Based on the latest trends centering on Silicon Valley, Mr. Sakaguchi explained the reality of the rapid expansion of AI data centers and the power and energy strategies that support it.
What Mr. Sakaguchi emphasized first was that "To understand data centers, it is first necessary to understand the structure of the cloud business." The current cloud market is dominated by hyperscalers like Amazon, Microsoft, and Google, and most AI services are provided on top of these giant cloud platforms.
In other words, the expansion of AI data centers is not just the development of IT infrastructure, but is closely related to the competitive structure of the cloud industry itself. The competition over AI is transforming from competition not only over algorithms and semiconductor performance, but into a capital-intensive competition over how much "physical infrastructure" like data centers can be secured.
Furthermore, regarding the ongoing investments in AI data centers, Mr. Sakaguchi explained, "While it cannot be denied that there is a bubble-like aspect," he drew parallels to the past IT bubble, stating, "Even if the bubble bursts, the infrastructure itself remains in society and has supported subsequent industrial development." Just as fiber optic networks and data centers are still utilized today as the foundation of the internet industry, his view is that AI data centers will likely remain as irreversible social infrastructure in the medium to long term, regardless of short-term overheating.
On the other hand, the biggest constraint emerging on the expansion of AI data centers is the issue of power supply. The performance improvement of AI models relies on "scaling laws," where injecting more computational resources improves performance, leading to fierce competition among companies to secure GPUs and power. As a result, the power demand of data centers is rapidly increasing.
In addition, the speed of evolution of AI semiconductors is extremely fast, and the time gap required to build data centers is also severe. While GPUs undergo a generation change in about 9 months, it takes several years to design and construct a data center and prepare power transmission facilities. This time difference creates a structural difficulty where the assumptions become obsolete by the time of completion.
Furthermore, looking ahead, it is expected that the power consumption of not just AI model training, but "inference" where generative AI and business AI are constantly running, will also expand. In the future, if AI becomes established as social infrastructure, the demand for power will not be temporary but will increase continuously.
Under these circumstances, various power supply methods such as renewable energy, nuclear power, and small modular reactors (SMRs) are being discussed, but there is no definitive silver bullet that can eliminate the power constraints in the short term.
Mr. Sakaguchi concluded that "The competition in AI infrastructure is not just about computing power, but power and energy."