Amazon is preparing to unveil its latest artificial intelligence chip, Trainium 2, as part of its broader multibillion-dollar investment in semiconductor technology. This move by Amazon aims to reduce the company’s reliance on NVIDIA, which currently dominates the AI processor market, and to increase the efficiency of its cloud computing arm, Amazon Web Services (AWS).
The introduction of Trainium 2 follows Amazon’s ongoing efforts to control more of its AI infrastructure. Developed by Annapurna Labs, a startup acquired by Amazon in 2015 for USD 350 million, Trainium 2 is designed to enhance the performance of AI model training in AWS data centres. By creating its own chips, Amazon hopes to lower operational costs and offer customers more affordable cloud computing options. The new chip is expected to be officially revealed next month and has already been trialed by several major companies, including OpenAI competitor Anthropic, Databricks, Deutsche Telekom and Ricoh.
Amazon’s AI chip strategy isn’t just about internal efficiency; it is also aimed at competing with NVIDIA’s near-total dominance in the AI processor market. Currently, AWS relies on NVIDIA’s technology for its AI operations, but Amazon aims to offer a viable alternative to NVIDIA’s offerings, which continue to drive the AI chip market. The company claims its existing chip, Inferentia, has already reduced AI response costs by 40 per cent, presenting an appealing alternative for enterprises grappling with high machine learning expenses.
Amazon’s decision to invest heavily in AI chips aligns with broader industry trends, where tech giants like Microsoft, Google and Meta are also developing their own semiconductor solutions. In 2024, Amazon plans to spend USD 75 billion on technology infrastructure, a substantial increase from USD 48.4 billion in 2023. This way Amazon can be building integrated systems that can lower costs, improve performance and increase control over its supply chains.
Despite Amazon’s ambitions, NVIDIA remains a dominant player in the AI chip market. NVIDIA’s second fiscal quarter of 2024 saw USD 26.3 billion in revenue from AI data centre chips, a figure that rivals the total earnings from AWS for the same period.
_20241114115759_original_image_40.webp)
