Meta today published concrete plans for the company's AI infrastructure.
Own AI chips for enterprise data centers; Image: Meta
Meta today published concrete plans for the development of a new AI infrastructure in the company. A key part of these plans is the introduction of the MTIA (Meta Training and Inference Accelerator) chip, designed specifically for Meta. The MTIA chip promises improved performance and efficiency.
According to Santosh Janardhan, a company representative, the combination of MTIA chips and GPUs will increase performance, reduce latency and increase efficiency for various workloads. However, the chip is not expected to be available until after 2025, according to a report by TechCrunch.
Meta also plans to introduce a new ASIC (application-specific integrated circuit) that will improve video transcoding. This chip aims to improve the quality and latency of live streaming on the platform.
Meta aims to keep up with companies like NVIDIA and Microsoft. It was recently announced that Microsoft is also working on developing its own AI chips to reduce its dependence on NVIDIA graphics cards.
With these plans, Meta clearly demonstrates its commitment to the advancement of AI technology and its ambition to be a leader in this space. The chips mentioned are expected to bring important improvements in terms of performance, efficiency and areas of application. With these developments, Meta could strengthen its position as an influential company in the AI industry.