AWS and Cerebras' collaboration aims to set a new standard for AI inference speed and performance in the cloud

AWS and Cerebras Systems Inc. have partnered to advance AI inference capabilities through the Cerebras Fast Inference Cloud, a service available on AWS Marketplace. This collaboration leverages Cerebras' Wafer-Scale Engine (WSE) and CS-3 systems to deliver low-latency, high-throughput inference for open-source models such as Llama, Qwen, and OpenAI GPT-OSS. According to customer reviews and benchmarks, the platform achieves up to 70 times faster performance than traditional GPU-based systems, with throughput exceeding 2,500 tokens per second.

The service is designed for real-time applications, including multi-step reasoning and agentic workflows, enabling users to deploy models in under 30 seconds via an OpenAI-compatible API. Pricing is usage-based, with costs tied to consumption units at $0.0000001 per unit, though additional AWS infrastructure fees may apply. Early adopters in sectors like quantitative finance and software development have reported significant efficiency gains, with one user noting a 50-fold improvement in decision-making pipelines.

Cerebras' wafer-scale architecture has also enabled partnerships with companies like NinjaTech AI, which utilizes the technology to accelerate deep research tasks by up to 5x while maintaining accuracy comparable to leading models. The company's infrastructure expansion, including new data centers and a Series G funding round, underscores growing demand for high-speed inference solutions. As AWS and Cerebras continue to integrate their offerings, the partnership highlights a shift toward performance-driven AI deployment in cloud environments.