June 29 (Mon) "Trends and Challenges in Power and Cooling Infrastructure and Energy Storage Systems for Next-Generation AI Data Centers ~Changes in Requirements with the Evolution of NVIDIA GPU Architecture~" Zoom Seminar to be Held

AndTech Co., Ltd. (Headquarters: Kawasaki City, Kanagawa Prefecture; President: Masao Suyama; hereinafter AndTech) will explain the "Power and Cooling Infrastructure and Energy Storage Systems" for "AI Data Centers" as part of its R&D development support.

We will explain the cooling systems and power infrastructure required for AI data centers, focusing on the evolution of NVIDIA GPU architecture!

NVIDIA's GPU architecture continues to evolve from H100 to Blackwell, and the latest Vera Rubin. In this seminar, we define AI data centers as "giant energy conversion factories that convert electricity into intelligence" and explain specific guidelines for understanding "where to apply your technology."

Live Streaming / Web Seminar Outline

──────────────────

Theme: Trends and Challenges in Power and Cooling Infrastructure and Energy Storage Systems for Next-Generation AI Data Centers ~Changes in Requirements with the Evolution of NVIDIA GPU Architecture~

Date and Time: Monday, June 29, 2026, 13:30-16:30

Participation Fee: 38,500 yen (tax included) *Materials will be distributed electronically.

URL: https://andtech.co.jp/seminars/1f1347d0-4370-6980-92de-064fb9a95405

Web Streaming Format:

This will be a live streaming seminar using the web conferencing tool "Zoom."

Details will be provided after application.

Seminar Content Structure

────────────

Daisuke Imaizumi, Representative Director, Researcher, AI x Management Strategist, Infracommons Co., Ltd.

Knowledge and Technical Challenges to be Learned/Solved in this Seminar

───────────────────────

① Understanding the NVIDIA GPU roadmap and "intelligence production efficiency"

② "Simulation-First" design method using Digital Twin (DSX)

③ Intersection of next-generation "liquid cooling systems" and materials science

④ 800VDC DC power distribution and physical countermeasures against power spikes

⑤ Strategic utilization of grid-interactive data centers and storage batteries

All program items below (please see if you are interested in details)

──────────────────────────────

[About the Speaker]

He has experience in forming natural gas (LNG) and woody biomass power plants, allowing him to view AI data centers from the upstream of power supply, and covers the latest NVIDIA trends.

[Program]

∽∽────────────────────────────∽∽

Part 1: Historical Transformation of NVIDIA GPU Architecture and "Competitive Axes"

~From "single server" performance to "AI Factory" production efficiency~

∽∽────────────────────────────∽∽

We will organize how the competitive axis has shifted from "chip performance" to "rack integration performance" and "AI processing efficiency per watt" with the evolution of GPUs. We will explain the transition to an era where performance is evaluated not by a single server, but by the entire AI infrastructure (AI Factory), including power and cooling.

1. Hopper Generation: The "solo run" period of absolute performance (2022-2023)

Competitive axis: Absolute values of computing power (FLOPS) and HBM memory capacity

2. Blackwell Generation: Emergence of pursuers and shift to "rack unit" (2024-2025)

Competitive axis: Rack-unit integrated performance (NVL72) and "inference throughput" with FP4 precision

3. Vera Rubin Generation: "Redefinition" of competitive axes — Intelligence production efficiency (2026~)

Competitive axis: "Intelligence output per 1MW (megawatt)"

∽∽────────────────────────────∽∽_

Part 2: AI Factory Design Revolution

~From "still images" of 3D CAD to "living digital twins" with OpenUSD~

∽∽────────────────────────────∽∽

In AI data centers, conventional designs are reaching their limits due to high heat generation and high density, making pre-design using digital twins and AI simulations indispensable. Therefore, we will specifically explain design methods (OpenUSD, AI-CFD) that integrally optimize wiring, piping, and thermal fluids, and their practical application.

1. The end of "trial and error": Transition to simulation-first

2. OpenUSD connecting "integrated infrastructure design"

3. Evolution of thermal design: From equipment combinations to AI-accelerated CFD

∽∽────────────────────────────