Amazon is all set to launch its self-developed AI chip to reduce reliance on NVIDIA

Amazon is preparing to introduce its latest artificial intelligence chips, marking a significant step in its multibillion-dollar investment in semiconductor technology.

The objective is to reduce its reliance on NVIDIA, which currently dominates the AI processor market. Amazon’s cloud computing arm, Amazon Web Services (AWS), is banking on these custom chips to improve efficiency across its data centres, ultimately lowering operational costs and offering more affordable options to its customers.

Amazon’s in-house chips is being developed by Annapurna Labs, an Austin-based start-up, which was acquired by Amazon in 2015 for $350 million. The new AI chip, Trainium 2, is expected to be unveiled next month, promising significant improvements in training large-scale AI models.

Trainium 2 is already being tested by several companies. Companies who have the Trainium under trial in their processes include, OpenAI’s rival Anthropic, Databricks, Deutsche Telekom, and Ricoh of Japan.

More from Tech

How ChatGPT is becoming everyone’s BFF and why that’s dangerous America ready for self-driving cars, but it has a legal problem

AWS and Annapurna Labs have trained their guns on NVIDIA. While AWS aims to remain a top-tier platform for running NVIDIA technology, Amazon is keen on providing a viable alternative.

The company claims that its existing AI chip line, Inferentia, can deliver AI-generated responses at 40 per cent lower costs, a compelling incentive for enterprises managing hefty machine learning expenses.

Amazon’s ambitious investment in AI infrastructure reflects a broader trend among tech giants. The company anticipates spending roughly $75 billion on technology infrastructure in 2024, a substantial jump from the $48.4 billion spent in 2023.

Impact Shorts

More Shorts

America ready for self-driving cars, but it has a legal problem

Alibaba, Baidu begin using own AI chips as China shifts away from US tech amid Nvidia row

Chief Executive Andy Jassy has signalled that this expenditure could rise even further in 2025. This surge in spending aligns with strategies from other major players like Microsoft and Google, all engaged in an aggressive race to develop AI technology.

Analysts note that cloud providers, including Amazon, Microsoft, and Meta, remain heavy consumers of NVIDIA’s technology. Yet, they are simultaneously developing their own AI chips to achieve greater efficiency and control over their data centre operations.

Daniel Newman from The Futurum Group highlights a broader industry trend: major companies are racing to build vertical and integrated chip systems to lower costs, improve performance, and secure supply chains.

Despite Amazon’s efforts, NVIDIA’s dominance remains formidable. NVIDIA reported $26.3 billion in revenue from AI data centre chips in its second fiscal quarter of 2024, a figure that rivals Amazon’s total earnings from AWS for the same period.

Analysts remain sceptical about Amazon’s capacity to meaningfully challenge NVIDIA’s market stronghold, as AWS’s AI infrastructure has yet to make a significant impact.

While Amazon refrains from direct performance comparisons between its chips and NVIDIA’s, experts believe that offering a credible alternative could reshape the market.

Patrick Moorhead, a chip consultant, suggests that while NVIDIA’s innovation is appreciated, there is growing discomfort with its overwhelming market share. As Amazon pushes forward, the demand for competition in the AI chip sector is only intensifying.