Meta claims custom chips will reduce costs and energy use compared to Nvidia

Meta: Meta Platforms develops social media platforms and invests heavily in AI infrastructure. The company designs custom MTIA accelerators in partnership with Broadcom and TSMC to handle recommendation, ranking, and generative AI workloads more efficiently than off-the-shelf options. These chips are positioned to deliver cost and energy advantages compared with Nvidia GPUs for Meta's specific use cases. Nvidia: Nvidia designs and supplies GPUs and related software widely used for AI training and inference. The company continues advancing its architectures amid growing competition from custom silicon developed by large tech firms. Its products remain central to many AI deployments even as clients like Meta pursue in-house alternatives for targeted efficiency gains. Efficiency: Hyperscalers like Meta are prioritizing custom AI accelerators optimized for specific workloads to achieve better energy and cost performance than general-purpose GPUs. Competition: Major technology companies continue expanding in-house chip programs to diversify supply and tailor hardware to their AI needs while maintaining spending on established GPU providers.