AI workloads are evolving fast, and AMD’s clearly ready to keep up. With the launch of the Instinct MI350 Series, they’re pushing performance to a whole new level, boasting up to 4x the AI compute power and 35x better inference compared to the previous gen.

Powered by CDNA 4, the MI350X and MI355X GPUs come with 288GB HBM3E, 8TB/s bandwidth, and support scalable deployment up to 128 GPUs per rack delivering up to 2.6 exaFLOPS FP4. This isn’t just raw power, it’s optimized for real AI work, from GenAI to training to inference.

The software side is no slouch either. ROCm 7 now supports major models like LLaMA, brings over 3.5x inference gains, and even powers 1.8M+ Hugging Face models out of the box. It’s open, fast, and built for scale.

And AMD’s just getting started. Coming in 2026, the next-gen MI400 Series promises a 10x performance leap, paired with 432GB HBM4 and the new “Helios” AI rack, a full-blown rack-scale system designed for serious compute density.

Bottom line? If you’re looking at building or scaling real AI infrastructure, AMD’s new Instinct lineup isn’t just promising, it’s delivering.

Instinct MI350 Series Quick Specs (Peak Theoretical)

SPECIFICATIONS   (PEAK THEORETICAL)	AMD INSTINCT™ MI350X GPU	AMD INSTINCT™ MI350X PLATFORM	AMD INSTINCT™ MI355X GPU	AMD INSTINCT™ MI355X PLATFORM
GPUs	Instinct MI350X OAM	8 x Instinct MI350X OAM	Instinct MI355X OAM	8 x Instinct MI355X OAM
GPU Architecture	CDNA 4	CDNA 4	CDNA 4	CDNA 4
Dedicated Memory Size	288 GB HBM3E	2.3 TB HBM3E	288 GB HBM3E	2.3 TB HBM3E
Memory Bandwidth	8 TB/s	8 TB/s per OAM	8 TB/s	8 TB/s per OAM
FP64 Performance	72 TFLOPs	577 TFLOPs	79 TFLOPS	1.1 PFLOPS
FP16 Performance*	4.6 PFLOPS	36.8 PFLOPS	5 PFLOPS	40.2 PFLOPS
FP8 Performance*	9.2 PFLOPs	72 PFLOPs	10.1 PFLOPs	80.5 PFLOPs
FP6 Performance*	18.45 PFLOPS	148 PFLOPS	20.1 PFLOPS	161 PFLOPS
FP4 Performance*	18.45 PFLOPS	148 PFLOPS	20.1 PFLOPS	161 PFLOPS