TDK — leaderboard ()

AMD Instinct MI400 GPUs launch with HBM4 and ROCm for AI and HPC

AMD has launched its AMD Instinct MI400 Series GPUs, a new family of data center GPUs aimed at AI and high-performance computing (HPC) deployments spanning “frontier AI,” “AI factory” environments, sovereign AI, and scientific computing. The lineup includes the AMD Instinct MI455X GPUs and the AMD Instinct MI430X GPU, and it pairs the GPUs with what AMD calls the AMD Helios rackscale solution for MI455X-based deployments.

The MI400 Series is built around HBM4 memory, and AMD is framing the family as a platform that combines compute, memory, and networking performance with advanced security features and system-scale design. For operators, that combination matters because the hard part of modern AI clusters isn’t only raw FLOPS; it’s scaling memory bandwidth, interconnect, and operational controls across large GPU domains without creating fragile, one-off architectures.

Within the lineup, AMD targets MI455X at high-volume inference, frontier-model training, and fine-tuning. MI430X is positioned for sovereign AI and HPC, and AMD quantifies its scientific-computing focus with up to 288 TFLOPS of hardware-based FP64 performance. AMD also points to “high memory bandwidth” from HBM4 as a lever for supporting larger models, reducing infrastructure complexity, and improving efficiency at scale.

Software and security details

On the software side, the MI400 Series uses AMD ROCm, including programming models, compilers, libraries, runtimes, and deployment tools for AI and HPC. AMD emphasizes open standards, interoperability, and long-term portability across software, system, and networking choices, with the explicit goal of avoiding vendor lock-in.

Security features called out for MI400 include secure boot, encrypted GPU-to-GPU links, and hardware-based protections intended to safeguard sensitive AI and HPC workloads. AMD ties those features to scaling options ranging from the Helios rackscale approach for frontier AI to more traditional mesh-based HPC deployments.

“The next generation of AI will span frontier AI, sovereign AI and scientific computing, and each requires infrastructure optimized for its unique demands,” said Vamsi Boppana, senior vice president, AI, AMD. “The AMD Instinct MI400 Series extends our AI portfolio with purpose-built solutions optimized for the full spectrum of AI and HPC deployments, built on an open software foundation that gives customers the flexibility to innovate at every scale.”

AMD introduced the MI400 Series at Advancing AI 2026.

Source: AMD

Get Data Center Engineering News In Your Inbox:

By subscribing, you agree to our Privacy Policy and Terms of Use. You can unsubscribe at any time.

Popular Posts:

Why data center power validation is moving to real-time simulation
Why data center power validation is moving to real-time simulation
CPC-TRM-Tech-Guide-Liquid-Cooling-Reliability-(1)-4
In liquid cooling reliability, tubing choice creates another layer of protection
Warm-Water-Ready-Accelsius-Whitepaper-3
Accelsius two-phase cooling runs NVIDIA B200 9° C cooler than single-phase
Vertiv-Tognana-Innovation-Lab
Vertiv to double chiller production capacity for data center cooling by 2026
Prevalon_Close_Image_solar_+_storage_Chile
Nextpower acquires Prevalon to expand BESS for grid and data centers
TDK — tower ()

Share Your Data Center Engineering News

Do you have a new product announcement, webinar, whitepaper, or article topic? 

Get Data Center Engineering News In Your Inbox:

By subscribing, you agree to our Privacy Policy and Terms of Use. You can unsubscribe at any time.