Amazon Web Services (AWS) is in discussions for a potential deal worth USD 475 million over five years with IBM to provide the tech giant with access to Nvidia’s AI chips via AWS's cloud platform, as per an article posted on Business Insider. The deal would allow IBM to use AWS EC2 servers, which are equipped with Nvidia’s GPUs, for AI model training and other machine learning tasks.
According to an internal Amazon document obtained by Business Insider, IBM Research has already begun training some AI models using AWS’s EC2 servers and machine learning platform, SageMaker. The ongoing talks are centered on expanding IBM's AI capabilities, with the potential to solidify the relationship between AWS and IBM, particularly within the realm of AI.
IBM has already integrated AWS services into its Watson AI platform, as declared in May. The new agreement could deepen the collaboration, allowing IBM to leverage AWS’s cloud infrastructure for more advanced AI tasks. The move also depicts the continuing demand for Nvidia’s GPUs, which are a crucial component in AI training and processing. While AWS has developed its own AI chips, Trainium and Inferentia, it remains to be seen whether the deal with IBM would include access to these Amazon chips, or if Nvidia’s GPUs would remain the primary choice.
The potential deal shows AWS's growing presence in the AI cloud market, which Amazon CEO Andy Jassy recently noted is on track to generate ‘multi-billion dollars’ in revenue this year, with triple-digit growth year over year. Despite AWS’s focus on its own AI chips, Nvidia's GPUs continue to be in high demand. In the latest quarter, AWS reported USD 27.5 billion in revenue, reflecting a 19 per cent year-over-year increase. Amazon is also ramping up its capital expenditures, with plans to invest more than USD 75 billion this year in cloud infrastructure, including GPUs, networking and data centers. Spokespeople from AWS and IBM have not commented on the ongoing negotiations.