June 29 (Monday) "Trends and Challenges of Power, 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; Representative Director and President: Masao Toyama; hereinafter AndTech) will provide an explanation on "Power, 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 "gigantic energy conversion factories that convert electricity into intelligence" and explain concrete guidelines for "where to focus your company's technology."

Live Streaming/WEB Seminar Outline

──────────────────

Theme: Trends and Challenges of Power, 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 Distribution Format:

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

Details will be provided after application.

Seminar Content Structure

────────────

Mr. Daisuke Imaizumi, Representative Director, Researcher, AI x Management Strategist, Infra Commons Inc.

Knowledge that can be learned and technical issues that can be resolved in this seminar

───────────────────────

① Understanding NVIDIA GPU roadmap and "intelligence production efficiency"

② Design methodology of "simulation-first" by digital twin (DSX)

③ Contact point between next-generation "liquid cooling system" and material science

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

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

All items in the program below (please take a look if you are interested in details)

──────────────────────────────

[About the Speaker]

Has experience in natural gas (LNG) and wood biomass power plant composition, can grasp AI data centers from the upstream of power supply, and covers the latest NVIDIA trends.

[Program]

∽∽────────────────────────────∽∽

Part 1: Historical Shift in NVIDIA GPU Architecture and "Competitive Axis"

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

∽∽────────────────────────────∽∽

We will organize how the competitive axis has shifted from "chip performance" to "rack integrated performance" and "AI processing efficiency per power" with the evolution of GPUs. We will explain the shift to an era where performance is evaluated across the entire AI infrastructure (AI Factory), including power and cooling, rather than just individual servers.

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

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

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

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

3. Vera Rubin Generation: "Redefining" the competitive axis — Intelligence production efficiency (2026~)

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

∽∽────────────────────────────∽∽_

Part 2: Design Revolution of AI Factory

~ 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 prior design using digital twins and AI simulation indispensable. Therefore, we will specifically explain design methodologies (OpenUSD, AI-CFD) that integrate and optimize wiring, piping, and fluid dynamics, and their application to practical work.

1. The End of "Trial and Error": Shift to Simulation-First

2. "Integrated Infrastructure Design" connected by OpenUSD

3. Evolution of Thermal Design: From equipment combinations to AI-accelerated CFD

∽∽────────────────────────────