Amazon’s competition with Nvidia in the field of artificial intelligence

Sedaye Sama News Agency

A new phase of competition between Amazon and Nvidia in the AI sector has begun with the introduction of the Trainium3 chip, which AWS claims can reduce the cost of training and running AI models by up to 50%, making it a serious alternative to Nvidia’s GPUs. Developed by Annapurna Labs, the chip has already shown strong performance: early adopters such as the AI video startup Decart report significant improvements in real-time video processing, highlighting Trainium3’s competitive computing power. Meanwhile, a wave of new market deals suggests major AI companies are seeking to diversify their suppliers and reduce dependence on Nvidia.

Meta and Google are negotiating multi-billion-dollar purchases of TPU processors, while OpenAI has signed agreements with AMD and Broadcom to secure the massive computational capacity required for models like GPT and Sora. However, this shift does not signal the abandonment of Nvidia; the company’s latest quarterly results show record revenue, rising profits, and continued dominance in the GPU market. Many AWS customers continue to rely on Nvidia in parallel, but the launch of Trainium3—combined with broader supplier diversification—marks a more competitive AI compute landscape.

Alongside its new chip, AWS also introduced a suite of advanced AI tools—most notably the Frontier Agents, capable of running complex workflows for hours or even days without interruption. According to AWS’s CEO, these agents are built with a mix of advanced memory architecture, multiple integrated models, and a robust cloud infrastructure, representing a new generation of enterprise AI. These innovations reflect Amazon’s push to regain a leading position in a market where some had viewed it as lagging behind.

AWS also unveiled Nova Forge, a service enabling organizations to train private versions of Amazon’s Nova models using their own proprietary data, creating deeply customized AI systems. Companies testing the beta version have reportedly achieved 40–60% performance improvements over methods such as fine-tuning or RAG. The service is intended for businesses seeking models that can understand their workflows and operational needs at a deep and tailored level.

Together, these announcements—made at AWS’s annual conference and supported by AWS’s 20% revenue growth—signal Amazon’s new strategy to place itself back at the center of the AI race. This competition now extends beyond GPUs into enterprise infrastructure, data ecosystems, and native model development—areas in which AWS aims to strengthen its leadership.

source: donyaye eghtesd