AMD's new Efficiency Targets for AI Nodes by 2030

Growth in the areas of artificial intelligence and data-intensive applications is leading to a drastic increase in energy consumption. This requires a rethink in the technology industry. AMD is meeting this challenge with ambitious efficiency targets and innovative technology.

Studies by the International Energy Agency (IEA) assume that the electricity requirements of data centers will more than double to around 945 terawatt hours (TWh) by 2030 due to the increasing demand for computing power worldwide. This will result in a larger_{carbon footprint}, changing patterns of electricity demand and accelerated depletion of natural resources.

For this reason, AMD will continue to incorporate energy efficiency as a key design principle in its roadmap and product strategy. A concrete example is AMD's 2021 long-term 30x25 goal to improve the energy efficiency of AI training and high-performance computing (HPC) nodes by 30x from 2020 to 2025.

Not only has this ambitious goal been surpassed with a 38x increase, but it has also led AMD to commit to a bold new goal: a 20x improvement in the energy efficiency of AI training and inference nodes by 2030 (base year 2024). Part of this commitment is that by 2030, a typical AI model that requires more than 275 racks today can be trained in a single rack that consumes 95 percent less power than a comparable 2024 system.

AMD will be working hard to achieve this goal on time. The technical basis for this are AMD's EPYC processors and Instinct accelerators. These enable companies to invest in AI and minimize their ecological footprint at the same time. In combination with software and algorithmic advances, the new target could enable an up to 100-fold improvement in overall energy efficiency.

Back to topic page

You might also be interested in

Deutsche Messe

Extended three-member Management Board for Deutsche Messe

Deutsche Messe AG is reorganizing its personnel and consistently focusing its Management Board on the further development of the company and the exhibition grounds in particular. With an expanded three-member Management Board, the company is...

Federation Architecture | Part 1

Rethink IT-OT Convergence in an AI-Driven World

AI is shifting threat models for industrial infrastructures. Federation Architecture proposes a disciplined middle ground between isolation and full integration - with edge autonomy, one-way data flow, and human approval.

Durabook

AI-enabled rugged mobile workstation

With 'Z14I-HG' Durabook presents an AI-ready Fully Rugged Mobile Workstation with Nvidia RTX 5000/3500 Ada graphics.

Cooperation

Siemens and Nvidia are accelerating Chip Testing

Siemens and Nvidia have significantly accelerated the verification of AI chips. With the help of FPGA-based prototypes, tens of trillions of design cycles could be tested within a few days before production. This should improve the development and...

Study on AI and the World of Work

Human-Machine Teamwork becomes central

A C-level survey by valantic and the Handelsblatt Research Institute examines how AI will transform the world of work by 2030. Managers expect creativity, empathy and ethical judgment to remain crucial skills in the future.

Industrial AI

IFS relies on Asset-based Pricing Models

IFS is introducing a new pricing model designed to simplify access to 'Industrial AI'. Instead of paying per user, companies will in future pay per asset operated, enabling company-wide use without cost increases.

Congatec

Robust AI Performance at Temperatures from -40 °C to +85 °C

Congatec is expanding its portfolio of Computer-on-Modules (COMs) to include variants with Intel Core Ultra Series 3 processors for the industrial temperature range from -40 °C to +85 °C.

Personnel

Q.ANT hires Michael Krüger for Sales

Photonics specialist Q.ANT is expanding its management team. Michael Krüger is taking on the newly created position of Vice President Commercials and will be responsible for driving forward the marketing of the processor technology.