Meta plans big for AI, announces new chipset creation
Meta announces new chipsets to improve the performance of its AI models and supercomputers, and eyes big for data centres.
Highlights
- Meta develops infrastructure for AI growth
- The tech giant announced chipset manufacturing for the creation of AI models
- Meta builds new data centre for AI hardware generations
With the manufacturing of a new processor created exclusively for running AI models, Meta is making tremendous advancements in artificial intelligence (AI). The company, which is well-known for its family of applications and its conception of the metaverse, is starting a bold project to develop the infrastructure for the next wave of artificial intelligence.
Meta plans to create special silicon chips for AI models
Santosh Janardhan, VP & Head of Infrastructure at Meta, recently shared the progress of their upcoming silicon chip. In addition to developing a special silicon chip for AI models, Meta is also developing an AI-optimised data centre design and the second stage of a gigantic 16,000 GPU supercomputer for AI research.
As per the company, such advancements are directed towards the growth of AI at scale.
According to Meta, AI already plays a crucial role in Meta's products, helping to improve personalisation, make them safer and fairer, and improve user experiences. With the introduction of CodeCompose, a generative AI-based coding helper designed to increase developer productivity throughout the software development lifecycle, the firm is even reinventing how coding is done.
The MTIA (Meta Training and Inference Accelerator), their in-house custom accelerator chip family, is the focal point of Meta's infrastructure advancements. According to Meta, MTIA is more powerful and efficient than CPUs because it was created expressly for inference tasks. As per the recent reports from the company, the combination of MTIA chips with GPUs may lead to higher performance, lower latency, and better efficiency.
Tech giant focuses on building data centres
Meta is also building a data centre that will support current products and have space for new AI hardware generations. This AI-optimised design will include liquid-cooled AI hardware and an efficient AI network, which would connect thousands of AI chips to form data centre-scale AI training clusters. According to Meta, it will be easier and less expensive to develop the new data centre.
