Key Takeaways
- Amazon Web Services will integrate Cerebras Wafer-Scale Engine processors into its cloud infrastructure for AI inference workloads.
- The partnership spans multiple years, though neither party disclosed financial specifics.
- According to Cerebras, its hardware executes inference operations up to 25 times quicker than competing Nvidia graphics processors.
- OpenAI previously committed to a Cerebras agreement exceeding $10 billion in value during January 2026.
- The chip startup secured $1 billion in fresh capital during February 2026, pushing its valuation to around $23 billion.
Amazon Web Services has forged a multiyear collaboration with semiconductor newcomer Cerebras Systems, bringing its Wafer-Scale Engine technology into AWS cloud facilities. These specialized processors will focus exclusively on AI inference operations — the stage where trained models generate answers to user prompts.
As the dominant force in global cloud computing, AWS has traditionally relied on proprietary silicon branded as Trainium, engineered by its Annapurna Labs division. This fresh arrangement will see AWS merge Trainium capabilities with Cerebras processors to deliver superior inference performance.
According to Cerebras, its Wafer-Scale Engine excels during the “decode” stage of inference — the actual generation of model outputs — achieving speeds up to 25 times greater than Nvidia’s GPU alternatives.
AWS plans to market this capability as a high-performance tier. “Those seeking slower inference will find more economical alternatives,” stated Cerebras CEO Andrew Feldman. AWS representatives confirmed they’ll continue providing budget-friendly inference options powered exclusively by Trainium.
Cerebras Secures Multiple Major Wins
This AWS announcement follows closely behind OpenAI‘s January 2026 partnership with Cerebras, a transaction valued above $10 billion according to industry reports. That arrangement will supply computing power for OpenAI’s ChatGPT platform using Cerebras technology, with OpenAI targeting deployment of as much as 750 megawatts in processing infrastructure.
One month later in February 2026, Cerebras completed a $1 billion fundraising round that elevated total capital raised to $2.6 billion while establishing a company valuation near $23 billion. Notable investors include Fidelity Management, Benchmark, Tiger Global, and Coatue.
The company had initially pursued a public offering by filing IPO paperwork in September 2024, only to abandon those plans approximately twelve months afterward.
Mounting Competition for Nvidia
The collaboration between AWS and Cerebras represents another headwind for Nvidia within the inference segment. Industry momentum has been transitioning from model training — where Nvidia maintains commanding market share — toward inference applications that prioritize execution velocity.
Nvidia hasn’t remained passive. The company finalized a $20 billion licensing arrangement with chip designer Groq in December 2025. Furthermore, Nvidia has indicated plans to introduce an innovative processing architecture incorporating Groq’s intellectual property in coming months.
Nafea Bshara, Annapurna Labs co-founder and AWS executive, characterized the Cerebras collaboration as centered on maximizing throughput while minimizing expenses. “Our objective is accelerating performance and reducing costs,” he explained.
Shares of Amazon AMZN traded down 0.44% at publication time.
Get 3 Free Stock Ebooks
Discover top-performing stocks in AI, Crypto, and Technology with expert analysis.
- Top 10 AI Stocks - Leading AI companies
- Top 10 Crypto Stocks - Blockchain leaders
- Top 10 Tech Stocks - Tech giants
