Meta will adopt hundreds of thousands of AWS Graviton chips in latest AI infrastructure grab

Meta, the giant behind popular social media applications, is taking significant steps to enhance its AI infrastructure. With around 3.6 billion daily users, the company has been operating 32 data centers, including a newly completed facility in Oklahoma. However, this expansion is not stopping there. Recently, Amazon Web Services (AWS) announced a partnership with Meta to utilize its Graviton chips in a contract that spans at least three years. This collaboration reflects Meta's commitment, under CEO Mark Zuckerberg, to invest in robust computing capabilities to meet soaring demand, joining the ranks of tech leaders like Alphabet and Microsoft. In conjunction with this deal, Meta has also signed agreements worth a staggering $48 billion with companies like CoreWeave and Nebius for access to Nvidia GPUs, which are essential for running AI models. While Amazon has not disclosed the financial details of its agreement with Meta, the implications for AI processing are significant. To balance its infrastructure expansion, Meta recently announced layoffs affecting around 8,000 employees, or 10% of its workforce. The move to adopt Arm-based Graviton processors—capable of handling an array of computing tasks similar to traditional CPUs—marks a pivotal shift. Graviton chips can be especially beneficial for post-training processes in AI, refining models after extensive data analysis. Nafea Bshara, an AWS vice president, highlighted that Graviton is increasingly popular among foundational model companies, making Meta one of its newest adopters. Since Amazon's acquisition of Annapurna Labs in 2015, the development of specialized chips for AI training has accelerated, with Graviton emerging as a preferred choice for firms like Adobe, Apple, and Snowflake. According to AWS, Graviton offers the best performance-to-price ratio among all computing options available, boasting a remarkable 60% reduction in energy consumption compared to other solutions. With this new commitment, Meta is expected to scale its usage of Graviton chips from a small scale to hundreds of thousands, positioning itself among the top five customers of this technology. Additionally, amid rising demand for Intel's Xeon server chips, Intel's CEO acknowledged that the landscape of high-performance computing is evolving, with CPUs regaining their vital role in the AI era. However, Bshara clarified that Meta's choice to expand to Graviton was not due to a lack of alternatives but rather a strategic decision to enhance CPU-intensive workloads for AI applications. As the tech landscape continues to evolve, Meta's embrace of AWS Graviton chips signifies a bold move towards the future of AI capabilities.