Amazon Web Services (AMZN) continues forward with its ambitious in-house chips, this time using its first 3nm artificial intelligence chip to power its Trainium3 UltraServers.
The new Trainium3 UltraServers, which are now available, are four times more energy efficient and hold nearly four times the memory bandwidth over their prior iteration, the Trainium2 UltraServers.
Amazon said its customers utilizing Tranium, including Anthropic (ANTHRO), Karakuri, Metagenomi, NetoAI, Ricoh and Splash Music, have reduced training costs by as much as 50% over competing GPUs.
“Pioneering AI companies including Decart, an AI lab specializing in efficient, optimized generative AI video and image models that power real-time interactive experiences, are leveraging Trainium3’s capabilities for demanding workloads like real-time generative video, achieving 4x faster frame generation at half the cost of GPUs,” Amazon said.
Amazon is already working on the Trainium4 series, which is being designed to support Nvidia’s (NVDA) NVLink Fusion high-speed chip interconnect technology.
“This integration will enable Trainium4, Graviton, and Elastic Fabric Adapter to work together seamlessly within common MGX racks, providing you with a cost-effective, rack-scale AI infrastructure that supports both GPU and Trainium servers,” Amazon added.