TLDR
- AWS will deploy Cerebras’s Wafer-Scale Engine chips in its data centers for AI inference.
- The deal is multiyear; financial terms were not disclosed.
- Cerebras claims its chips process inference tasks up to 25 times faster than Nvidia GPUs.
- OpenAI signed a separate deal worth over $10 billion with Cerebras in January 2026.
- Cerebras raised $1 billion in February 2026, valuing the startup at roughly $23 billion.
Amazon Web Services has signed a multiyear partnership with chip startup Cerebras Systems to deploy its Wafer-Scale Engine processors inside AWS data centers. The chips will be used specifically for AI inference — the process by which an AI model responds to a user query.
$AMZN's AWS and Cerebras are partnering to bring faster AI inference to Amazon Bedrock in the coming months.
The setup will combine Trainium for prompt processing with Cerebras CS-3 for token generation, with Bedrock becoming the first cloud service to offer Cerebras’… pic.twitter.com/LT6yoDEmCB
— Wall St Engine (@wallstengine) March 13, 2026
AWS is the largest cloud provider in the world. It has historically leaned on its own in-house chips, known as Trainium, developed by its semiconductor unit Annapurna Labs. Under the new deal, AWS plans to combine Trainium with Cerebras chips to build a faster inference offering.
Cerebras says its Wafer-Scale Engine can handle the “decode” phase of inference — when the model actually generates its response — up to 25 times faster than Nvidia’s GPUs.
The service will be positioned as a premium offering. “If you want slow inference, there will be cheaper ways to go,” Cerebras CEO Andrew Feldman said. AWS confirmed it will still offer lower-cost inference solutions using Trainium alone.
A Big Week for Cerebras
This deal comes just months after OpenAI struck a separate agreement with Cerebras in January 2026, reportedly worth more than $10 billion. That deal is designed to power OpenAI’s ChatGPT with Cerebras processors, with OpenAI seeking to deploy up to 750 megawatts of computing capacity.
In February 2026, Cerebras raised $1 billion in a new funding round, bringing total fundraising to $2.6 billion and valuing the company at approximately $23 billion. Backers include Fidelity Management, Benchmark, Tiger Global, and Coatue.
Cerebras had filed for an IPO in September 2024 but withdrew that filing roughly a year later.
Pressure on Nvidia
The AWS-Cerebras tie-up adds to a growing list of challenges for Nvidia in the inference market. The AI industry has been shifting away from model training, where Nvidia’s GPUs dominate, toward inference workloads that demand more speed.
Nvidia is not sitting still. In December 2025, it signed a $20 billion licensing deal with chip startup Groq. The company also plans to unveil a new processing system built around Groq’s technology in the near term.
AWS’s Nafea Bshara, co-founder of Annapurna Labs, said the Cerebras partnership is focused on pushing speed and bringing down costs. “Our job is to push the speed and lower the price,” he said.
Amazon’s AMZN stock was down 0.44% at the time of reporting.





