- AMD says new Instinct MI350 Series GPUs are 35x better at inferencing than their predecessors
- AMD claims it has exceeded its energy efficiency goals, lays out bolder goals
- MI400 GPUs will power future Helios AI Racks
AMD has unveiled its Instinct MI350 Series GPUs, promising a staggering 4x improvement to AI performance compared with the previous generation chips – enough to have Nvidia worried about the market dominance of its Blackwell chips.
Company CEO Lisa Su also revealed details of the Helios AI Rack, which is to be built on next-generation Instinct MI400 Series GPUs as well as AMD EPYV Venice CPUs and AMD Pensando Vulcano NICs.
The news came at AMD’s Advancing AI 2025 conference, together with a series of other hardware, software and AI announcements.
AMD’s latest chips put up a fight against Nvidia’s Blackwell
Besides the 4x improvement to AI performance, AMD also boasts an eyewatering 35x generational improvement in inferencing as well as price-performance gains, unlocking 40% more tokens-per-dollar compared to its key like-for-like rival, the Nvidia B200.
Despite Nvidia’s market dominance, AMD proudly claims that seven in 10 of the largest model builders and Al companies use its Instinct accelerators, including Meta, OpenAI, Microsoft and xAI.
MI300X has been deployed for Llama 3 and 4 inferencing with Meta and proprietary and open-source models with Azure, among others.
Besides performance, AMD is also honing in on its environmental goals, claiming that its MI350 Series GPUs exceeded the five-year organizational goal to improve the energy efficiency of AI training and high-performance computing nodes by 30x – by reaching a figure of 38x.
By 2030, the company also wants to increase rack-scale energy efficiency by 20x compared with 2024, and it already predicts a 95% reduction in electricity for typical AI model training.
Looking ahead, Instinct MI400 Series GPUs are expected to deliver up to 10x more performance running inference on Mixture of Experts models.
Despite the bold claims, AMD’s market cap remains considerably lower than Nvidia’s, reaching $192.14 billion at press time.
“AMD is driving AI innovation at an unprecedented pace, highlighted by the launch of our AMD Instinct MI350 series accelerators, advances in our next generation AMD ‘Helios’ rack-scale solutions, and growing momentum for our ROCm open software stack,” said Su.
“We are entering the next phase of AI, driven by open standards, shared innovation and AMD’s expanding leadership across a broad ecosystem of hardware and software partners who are collaborating to define the future of AI.”
Leave a comment