Amazon Web Services offers Trainium
Product
Trainium is AWS's purpose-built AI accelerator chip family for training and inference of large generative models.
- Currently Offered
Trainium is AWS's purpose-built AI accelerator chip family for training and inference of large generative models.
AWS Trainium is a custom AI accelerator designed for high-performance training and inference of large language and generative models. The Trainium2 generation scales into EC2 Trn2 instances and the Trn2 UltraServer, which connects up to 64 Trainium2 chips to deliver 83.2 FP8 petaflops of compute. Chips are paired with high-bandwidth memory, including HBM3e configurations, to feed data-intensive transformer workloads.
The Trainium family is positioned as a cost-optimized cloud alternative to general-purpose GPU clusters for model training and inference. AWS also previews subsequent generations, with Trainium3 announced as a next-step accelerator using advanced process technology to push performance and efficiency for future generative AI workloads.
AMD Instinct is a family of data-center accelerators for AI and high-performance computing.
World's largest AI chip: 4T transistors, 125 petaflops on one wafer.
Cerebras Wafer-Scale Engine is a family of wafer-scale AI processors.
Broadcom designs and supplies semiconductors for networking, wireless, broadband, and industrial markets.
RDU is SambaNova's Reconfigurable Dataflow Unit, the company's AI chip family for inference and training at scale.
Tenstorrent PCIe accelerator card family for AI training and inference workloads.
PowerEdge XE9680 is Dell's 6U GPU-dense server for AI/ML supporting up to eight NVIDIA, AMD, or Intel accelerators.
Tenstorrent scale-out server family for production AI training and inference.
Wafer-scale AI supercomputer in mini-fridge form with cluster-scale power.
TPU is Google's custom AI accelerator ASIC family delivered through Google Cloud TPU Pods for ML training and inference.
AMD Instinct is a family of data-center accelerators for AI and high-performance computing.
Intel Gaudi is a data-center AI accelerator product family.
World's largest AI chip: 4T transistors, 125 petaflops on one wafer.
NVIDIA H100 is a data-center GPU for large AI model training and inference.
Cerebras Wafer-Scale Engine is a family of wafer-scale AI processors.
Broadcom designs and supplies semiconductors for networking, wireless, broadband, and industrial markets.
RDU is SambaNova's Reconfigurable Dataflow Unit, the company's AI chip family for inference and training at scale.
Tenstorrent PCIe accelerator card family for AI training and inference workloads.
PowerEdge XE9680 is Dell's 6U GPU-dense server for AI/ML supporting up to eight NVIDIA, AMD, or Intel accelerators.
Groq LPU is a custom inference chip with on-chip SRAM and deterministic scheduling delivering low-latency AI inference for LLMs.
SambaNova DataScale is a data-center AI computing platform.
Tenstorrent scale-out server family for production AI training and inference.
Wafer-scale AI supercomputer in mini-fridge form with cluster-scale power.
Amazon S3
Amazon ElastiCache
Kiro
Amazon Aurora
Amazon EMR
Amazon Relational Database Service
AWS Database Migration Service
Amazon Neptune
Amazon Braket
Amazon DynamoDB
Amazon Bedrock Guardrails
AWS Enterprise Support
AWS AppSync
Amazon SageMaker AI
Amazon Quick
AWS IoT Greengrass
AWS IoT Core
AWS IoT SiteWise
AWS Clean Rooms
Amazon Connect Customer