- Amazon is accelerating the development of its own AI chips to reduce its reliance on expensive Nvidia chips and to meet customer demand for more affordable AI computing.
- The company tested new server designs at its chip lab in Austin, Texas, and deployed lots of its own AI chips to cope with traffic spikes during the Prime Day shopping festival.
OUR TAKE
Amazon is working on building its own AI chips, aiming to reduce its reliance on Nvidia chips by providing more cost-effective computing solutions, and successfully applys them to tackle large-scale online shopping activities during Prime Day, highlighting its technological prowess and innovative strategy in the cloud computing market.
-Rae Li, BTW reporter
What happened
Amazon is aggressively developing its own AI chips with the aim of reducing its reliance on expensive Nvidia chips that are currently at the heart of AI computing in Amazon’s cloud computing service. Amazon engineers have been testing a new server design at its chip lab in Austin, Texas, that comes with Amazon’s own AI chips, which are designed to compete with Nvidia’s offerings. While Amazon’s AI chip development efforts are still in their infancy, its main chip for non-AI computing, Graviton, has been in development for nearly a decade and is currently in its fourth generation. Amazon’s AI chips Trainium and Inferentia, on the other hand, are newer designs. Amazon’s customers are increasingly looking for cheaper alternatives to Nvidia, so Amazon is working to help its customers compute complex calculations and process large amounts of data more economically with its own chips.
Amazon’s AWS business saw a 17% year-on-year increase in sales to $25 billion in the first quarter of 2023, accounting for nearly one-fifth of Amazon’s total revenue. AWS controls about one-third of the cloud market, while Microsoft’s Azure accounts for about 25%. During the recent Prime Day shopping event, Amazon deployed 250,000 Graviton chips and 80,000 custom AI chips in response to the surge of activity on the platform. According to Adobe Analytics, the shopping event generated a record $14.2 billion in sales.
Also read: Nvidia supplier SK Hynix posts record profits on AI chip boom
Also read: Nvidia approves Samsung’s HBM3 for China market GPUs
Why it’s important
The development of Amazon’s own AI chips is a significant move that underscores the importance of technological innovation and competition within the cloud computing and AI sectors. By reducing dependency on external suppliers, Amazon can better control costs, performance, and supply chain reliability, potentially maintaining a competitive edge. This initiative also highlights the need for cost-effective solutions in the market which could influence pricing strategies and competitive dynamics across the cloud computing industry. For businesses relying on cloud services, lower costs can translate into increased profit margins or more efficient resource utilisation.
Moreover, Amazon’s foray into chip development showcases the industry’s demand for diversified chip solutions in that AI and machine learning applications become more widespread. The growing need for high-performance computing has prompted companies to seek more efficient and cost-effective computational resources.
