Meta Announces Its Next Generation Of Infrastructure Specifically Built For AI

Meta, formerly Facebook, has unveiled ambitious plans to integrate artificial intelligence (AI) across its infrastructure and applications.

In a recent announcement, Meta outlined a comprehensive strategy centred on the creation of a new generation of AI infrastructure. This plan includes developing Meta’s first custom silicon chip for AI models, constructing an AI-optimised data centre, and extending the capabilities of their 16,000 GPU supercomputer for AI research.

These projects, and others in the pipeline, will enable Meta to develop and deploy larger, more sophisticated AI models. This is key to delivering better personalisation, safer products, and richer experiences for the audiences that businesses value the most.

A key part of this strategy is CodeCompose, a generative AI-based coding assistant developed in-house at Meta to boost productivity throughout the software development lifecycle. This move signifies a reinvention of traditional coding practices in favour of more efficient, AI-based alternatives.

AI is far from a new concept for Meta. Since constructing their first data centre in 2010, the company has been integrating AI systems into its infrastructure – from the Big Sur hardware introduced in 2015 to the development of PyTorch and their AI supercomputer.

Now, Meta is set to expand their infrastructure in several new ways.

MTIA (Meta Training and Inference Accelerator), a custom accelerator chip developed by Meta specifically for inference workloads. It is expected to deliver superior compute power and efficiency, outperforming traditional CPUs for internal workloads.
Next-Gen Data Centre:, the forthcoming data centre will support both current products and future generations of AI hardware. With an AI-optimised design featuring liquid-cooled AI hardware and a high-performance AI network, it promises to be faster, more cost-effective, and more efficient.
Research SuperCluster (RSC) AI Supercomputer, that Meta considers one of the fastest AI supercomputers in the world, will help develop cutting-edge tools and technologies such as augmented reality tools, content understanding systems, and real-time translation technology.

Looking ahead, Meta is prepared for an era of increasing specialisation in chip design and AI infrastructure, new systems for deployment at scale, and improved efficiency in product and design support. The company’s focus on long-term value and its track record of developing world-class infrastructure are the driving forces behind its vision for the future of AI. As Meta continues to break new ground, the world eagerly awaits the fruits of this latest endeavour.