Fast Technology May 19 news,On Thursday, Meta announced that it is working on its first custom chip specifically designed to run artificial intelligence models.

Meta isn't in the business of selling cloud computing services like tech giants such as Google and Microsoft, and the company hasn't publicly discussed its internal data center chip projects in the past. But Meta executives say the global tech community is showing interest, so communication is necessary.

The chip, called Meta Training and Inference Accelerator, or MTIA,Power consumption is only 25 watts, much lower than chips from manufacturers like NVIDIA.

According to Meta, the first MTIA chip will focus on AI reasoning. Meta Software engineer Joel Coburn said Meta initially considered using Gpus for inference tasks, but later found that Gpus were not so good for inference work.

"Gpus are still inefficient in real-world model applications and expensive to deploy." "That's why we need MTIA," says Meta software engineer Coburn.

In addition,Meta also describes a chip called a Scalable Video Processor (MSVP) that can be used to process and transmit video while reducing energy demand,The company says it needs to process 4 billion videos a day.