AMD Advances AI with Rack-Scale 'Helios' and MI355X GPU

AMD's AI Future is Rack Scale 'Helios'

At the recent Advancing AI 2025 conference, AMD unveiled several key announcements that are set to reshape the landscape of artificial intelligence (AI) hardware and software. The company’s focus on rack-scale solutions, enhanced GPUs, and robust software support underscores its commitment to providing high-performance, efficient, and scalable AI infrastructure.

New MI355X GPU: 2x AI FLOPs, More HBM, and Better Cost Efficiency

The standout announcement was the introduction of the new MI355X GPU. This latest addition to AMD’s lineup boasts:

• 2x AI FLOPs (Floating Point Operations per Second) compared to its predecessor, making it a powerhouse for compute-intensive tasks like deep learning and scientific simulations.
• Increased HBM (High Bandwidth Memory) capacity, which is crucial for handling large datasets and complex models without performance bottlenecks.
• A significant 40% better tokens/$ ratio than NVIDIA’s offerings, providing a compelling value proposition for businesses looking to optimize their AI investments.

The MI355X GPU is designed to deliver not just raw performance but also cost efficiency, making it an attractive option for data centers and cloud providers.

Software Wins: ROCm 7 with Major Performance Boosts

AMD’s commitment to software support was evident with the release of ROCM 7 (Radeon Open Compute). This latest version brings:

• Significant performance improvements across a range of AI workloads, ensuring that developers can take full advantage of the hardware capabilities.
• Day-0 support for the MI355X GPU, meaning that users can start leveraging the new hardware immediately without waiting for software updates.

ROCM 7 is a critical component of AMD’s strategy to provide a comprehensive and integrated AI ecosystem, from hardware to software.

Rack-Scale Wins: Turnkey Solutions with CPU + GPU + Network

One of the most innovative aspects of AMD’s announcement was its focus on rack-scale solutions. The company introduced new turnkey offerings that combine:

• AMD CPUs: Providing robust processing power for general-purpose computing tasks.
• AMD GPUs: Delivering high-performance AI acceleration.
• Network components: Ensuring efficient data transfer and communication within the rack.

These integrated solutions are designed to simplify deployment and management, making it easier for organizations to scale their AI infrastructure without the complexity of piecing together disparate components.

Roadmap Wins: Next-Gen in 2026 with 4x Performance and HBM4

Looking ahead, AMD outlined its roadmap for future advancements:

• 4x performance improvement over current offerings by 2026, driven by ongoing optimizations in both hardware and software.
• The introduction of HBM4, the next generation of high-bandwidth memory, which will further enhance data throughput and reduce latency.
• Continued focus on scalability, ensuring that AMD’s solutions can grow alongside the evolving needs of AI workloads.

This roadmap indicates AMD’s long-term commitment to staying at the forefront of AI technology and providing sustainable performance gains.

Efficiency Wins: Energy-Efficient Designs for Sustainable Computing

In addition to performance and cost efficiency, AMD emphasized its focus on energy efficiency. The new designs are optimized to deliver high performance while minimizing power consumption, which is crucial for data centers looking to reduce their environmental impact and operational costs.

Conclusion

AMD’s announcements at the Advancing AI 2025 conference highlight a comprehensive approach to advancing AI technology. From powerful GPUs to integrated rack-scale solutions and robust software support, AMD is positioning itself as a key player in the rapidly evolving AI landscape. With a strong roadmap for future innovations, AMD is well-equipped to meet the growing demands of AI applications.