Chips, especially Nvidia’s, have been in the limelight lately, particularly since the release of the new Deep Seek R1 model in January 2025. The news is that Meta is testing its first in-house AI chip. The chip is made specifically for AI-specific tasks only. This means the chip uses power more efficiently than GPUs for AI tasks. Meta plans to increase production if the testing goes well. This will help Meta cut costs by using in-house-built, cheaper, yet AI-specific chips rather than expensive chips. Meta CFO Susan Li said earlier in the year that company expenses for the year 2025 are expected to be between $114-119 billion. This is up from $95.1 billion in fiscal 2024.
Meta works with Taiwan-based chip-making company Taiwan Semiconductor Manufacturing Co Ltd (TSMC). The test deployment for the new chip started after the tape-out phase was completed. The “tape-out” process is the final step in the chip’s design before it is sent for manufacturing. Tens of millions of dollars and 3 to 6 months of handwork are typically required to reach this stage. However, a point to be noted here is that the tape-out process doesn’t guarantee the success of the chip.
Meta has also been involved in chip-making. It was named the Meta Training and Interference Accelerator (MTIA) series. This new chip is also made under the same series. Interestingly, one of the earlier chips failed in the testing after tape-out. Meta executives have said they want to use their own chip for AI training and the compute-intensive process of feeding large amounts of data to AI to teach the AI models.
the goal for the training chip is to start with recommendation systems and later use it for generative AI products like chatbot Meta AI, the executives said. Meta’s Chief Product Officer Chris Cox said at the Morgan Stanley technology, media and telecom conference last week.
“We’re working on how would we do training for recommender systems and then eventually how do we think about training and inference for gen AI”
Nvidia losing ground?
In light of the above development, it is clear that Meta wants to eliminate their dependence on Nvidia because it is a costly option. It would be a big blow for Nvidia. It has been reeling from the stock setup since the release of the DeepSeek R1 model. Four of Nvidia’s customers are working on developing their chips. It includes large tech companies like Google, Amazon, Microsoft, and Meta.
According to the Motley Fool Stock Advisor report on Nvidia, the company was not listed among the 10 best stocks in the last quarter of 2024. The future doesn’t present a good outlook for Nvidia. They need to develop a revolutionized new chip that gives better performance and is also cost-effective. Only with such tactics can they keep ahead of the competition. Such a development is likely to kill some of the competition.
