SEMIVISION @_@

SEMIVISION @_@

From 800V HVDC and Chroma’s Testing Platforms to Delta’s High-Voltage Infrastructure

Original Article By SemiVision Research [Reading time: 16 mins]

SEMIVISION's avatar
SEMIVISION
Jun 27, 2026
∙ Paid

The 72nd annual IEEE International Electron Devices Meeting (IEDM) is scheduled for December 12 - 16, 2026 at the Hilton San Francisco Union Square, with OnDemand access to recorded content after the event. Paper submission deadline is July 16, 2026, and the late-news deadline is August 18, 2026.

https://ieee-iedm.org/

The Next Bottleneck in AI Data Centers Is Not the GPU — It Is Power Architecture

Over the past two years, when the market discussed AI data centers, most attention went to GPUs, HBM, CoWoS, optical interconnects, liquid cooling, PCBs, CCLs, and server assembly.

All of these are important.

But as AI systems move from single-server architectures to rack-scale systems, and eventually to cluster-scale and data-center-scale deployment, the real bottleneck is no longer only computing power.

It is power.

More precisely, the key question is how electricity enters the data center, how it is distributed to racks, how it is converted for GPUs, how the system handles fast transient loads, how conversion losses are reduced, and how reliability is maintained under extreme power density.

AI data centers are evolving from IT infrastructure into power infrastructure.

In the past, a data center was mainly a facility that housed servers. In the AI era, a data center is increasingly becoming an AI factory — a highly electrified system that integrates power grids, transformers, rectifiers, energy storage, power distribution, liquid cooling, monitoring, testing, reliability validation, and operations.

This is why 800V HVDC is becoming one of the most important trends in next-generation AI data centers.

7 Key AI Hardware Keywords to Watch in 2026

7 Key AI Hardware Keywords to Watch in 2026

SEMIVISION
·
Jun 18
Read full story

The Nature of AI Data Centers Is Changing

Traditional cloud data centers were designed for search, storage, streaming, enterprise cloud services, and general-purpose computing. These workloads were large, but their power density was relatively manageable.

AI data centers are different.

In the past, a rack might consume 10kW, 20kW, or 30kW. Today, AI racks are moving toward 100kW and 200kW. Future rack-scale AI systems may push toward several hundred kilowatts or even megawatt-level power.

Below we will share:

  • Why 800V Is the Future Trend

  • 800V Will Create a New Supply Chain

  • Chroma: The Testing Platform Behind 800V HVDC AI Power Systems

  • Delta: From Power Supply to Grid-to-Chip Infrastructure

  • The Next AI Battleground Is Power Efficiency and Reliability

The Next Power Architecture for AI Data Centers: Solid-State Transformers and 800V HVDC

The Next Power Architecture for AI Data Centers: Solid-State Transformers and 800V HVDC

SEMIVISION
·
Jun 3
Read full story
User's avatar

Continue reading this post for free, courtesy of SEMIVISION.

Or purchase a paid subscription.
© 2026 TSPA Semiconductor · Privacy ∙ Terms ∙ Collection notice
Start your SubstackGet the app
Substack is the home for great culture