loader
Comparison of two GPU chips: AMD on the left with a red logo, Nvidia on the right with a green logo on a dark background.
A Bold Step Forward in the AI Hardware Race

New Chips, New Ambitions
At the 2025 tech conference, AMD CEO Lisa Su announced the launch of two new AI accelerators: the MI350 series, set for release in late 2024, and the MI400 family, expected in 2026. These chips are designed to rival Nvidia’s dominance in AI data centers by offering greater efficiency, scalability, and performance.

Targeting Training and Inference at Scale
The MI350, built on AMD’s advanced CDNA 4 architecture, is optimized for training large language models and other demanding AI workloads. Meanwhile, the future MI400 chips are expected to push performance boundaries even further, continuing AMD’s rapid innovation cycle.

Competitive Landscape Shifts
These launches mark a strategic move to capture a greater share of the booming AI accelerator market. AMD’s goal is to become a viable alternative to Nvidia’s H100 GPUs, particularly in cloud and enterprise computing.

Industry Reactions
Tech analysts note that AMD’s increased focus on AI hardware coincides with rising demand for more cost-effective, power-efficient alternatives. The MI350’s promise of improved throughput and better total cost of ownership could attract hyperscalers and AI startups alike.

Beyond Chips: Software and Ecosystem
AMD is also investing in ROCm (its open software platform) and ecosystem partnerships to make it easier for developers to adopt its new chips. The combined hardware-software approach could be crucial in gaining developer mindshare.