News

AWS Unveils Rainer Supercomputer and New AI Tools at Re:Invent

• By John K. Waters
• 12/03/2024

Amazon Web Services (AWS) unveiled plans to construct one of the world’s most powerful artificial intelligence (AI) supercomputers today. Announced at Amazon's Re:Invent conference in Las Vegas, the project—dubbed Project Rainer—will be built in collaboration with OpenAI rival Anthropic. When completed, the supercomputer will be five times larger than the cluster used to train Anthropic’s most advanced models to date, the company says.

During his conference keynote, AWS CEO Matt Garman revealed that the Rainer supercomputer will utilize hundreds of thousands of Amazon’s latest AI training chips, Trainium 2, making it the largest reported AI machine globally. Amazon also announced that Trainium 2 will soon be available in specialized UltraServer clusters for training advanced AI, offering a cost-effective alternative to Nvidia GPUs with savings of 30–40%.

"This is a transformative moment for AI infrastructure," Garman said, emphasizing Amazon’s commitment to making AI development more accessible and affordable.

Once seen as lagging behind rivals like Microsoft and Google in the generative AI race, Amazon has rapidly closed the gap. Over the past year, it has invested $8 billion into Anthropic and expanded its AWS Bedrock platform with tools tailored to generative AI needs.

At the event, Amazon introduced Trainium 3, the next iteration of its custom AI chips, promising four times the performance of its predecessor. Trainium 3 is set for release in late 2025.

AWS also rolled out tools designed to address the challenges of deploying generative AI models, which Garman described as "too expensive, unreliable, and unpredictable." Among these tools is Model Distillation, which compresses larger models into smaller, cost-effective versions without significant loss of functionality. Another service, Bedrock Agents, provides a system for creating and managing AI agents capable of automating complex tasks like customer support and data analytics.

Bryon Cook, AWS vice president of the Autonomous Reasoning Group, highlighted the tool’s versatility, noting its potential applications in industries ranging from insurance to human resources. "Reasoning will become a very important thing," Cook said. "It allows for sophisticated, reliable communication between AI systems."

Unlike competitors promoting consumer-facing AI like ChatGPT, Amazon has focused on providing the infrastructure and tools to empower other companies’ AI ambitions. Analysts believe this strategy leverages AWS’s unmatched cloud ecosystem.

AWS’s efforts to make generative AI accessible, reliable, and cost-effective appear to be paying off. According to Garman, a growing number of customers are transitioning from experimental AI projects to building commercially viable applications. "We’re seeing customers move from proofs of concept to scalable deployments," he said.