Amazon’s custom silicon chips are poised to give the e-commerce giant an advantage in generative AI, say industry analysts. The company has been producing custom chips since 2013, with its specialized hardware, Nitro, now being the highest-volume chip used by its cloud computing arm, Amazon Web Services (AWS). In recent years, Amazon has bolstered its chip capabilities through acquisitions, including chip startup Annapurna Labs in 2015. Notably, it launched its Arm-based server chip, Graviton, in 2018, as well as AI-focused chips.
While Amazon continues to push the boundaries in chip development, its biggest competitors, Microsoft and Google, have yet to make significant strides in this area. However, Microsoft is expected to announce its Athena AI chip soon, while Google already has its Tensor Processor Unit.
Amazon’s Trainium chip stands out for its ability to improve price performance by 50% when training machine learning models. This chip brings significant cost and efficiency benefits to the table. On the other hand, the Inferentia chip offers low-cost, high-throughput, and low-latency inference specifically tailored for generative AI models. Both chips position Amazon strongly in the growing generative AI market.
Although Nvidia’s GPUs currently dominate the AI training market, Amazon’s AWS cloud dominance is a major advantage. With a massive customer base that is already familiar with AWS, Amazon can easily expand its generative AI capabilities and attract even more users.
The numbers speak to AWS’ power in the cloud computing space. It is the world’s biggest provider, boasting a 40% market share in 2022. Despite declining operating income, AWS still accounted for 70% of Amazon’s overall operating profit in the second quarter.
To further solidify its position in generative AI, AWS has developed a growing portfolio of developer tools. One of these is the Bedrock service, which provides access to state-of-the-art models from various providers. By offering a wide range of tools and options, Amazon aims to give its customers tools that are tailored to their specific needs, as it believes that no single model will dominate the AI world.
With its continued investment in custom chips and a robust portfolio of developer tools, Amazon is well-positioned to make significant inroads in the ever-expanding field of generative AI.
“Zombie enthusiast. Subtly charming travel practitioner. Webaholic. Internet expert.”