Many companies are already ramping up their developments for their in-house technology, and the next that reportedly joined this trend is Meta as it is now testing its self-made AI training chip.
It is reported that Meta is also looking to reduce its reliance on NVIDIA's AI training chips, the technology that helped the company reached unimaginable heights last year when it briefly took the top spot as the world's most valuable company.
Meta Reportedly Testing its Own AI Training Chip
According to the latest report by Reuters (via Engadget), Meta has massive plans already underway that look to expand its artificial intelligence development.
The report reveals that the company has already developed an in-house AI training chip. Meta is now reportedly in the process of testing these new AI training chips that will help save on costs.
The testing is being done in a small-scale deployment which makes use of this dedicated accelerator chip, and it was designed to specialize in AI-related tasks. The report adds that the company already completed its initial tape-out', which means that a complete design is already delivered for its manufacturing test run.
The goal for Meta, according to the report, is to ace this training period so that it will be ready to train its future AI models, and the company plans to start using these custom processors by 2026.
Read Also: Meta Expands AI Integration: WhatsApp to Get Dedicated Meta AI Widget That You Can Turn Off
Meta Looks to Reduce Reliance on NVIDIA
The larger goal for Meta is to reportedly reduce its reliance on NVIDIA, making it a more cost-effective plan for them as having its own chip can help lower its infrastructure costs and be more self-sufficient.
The report claims that this in-house Meta AI training chip is more power-efficient as it focuses, unlike the general-purpose NVIDIA GPUs that they have been using as part of the Meta Training and Inference Accelerator (MTIA).
Meta's AI Development and NVIDIA
Meta is one of the many customers of NVIDIA that highly relies on their AI training chips, which helps process loads of data and train different types of large language models. With its massive focus on custom-made AI training chips, NVIDIA grew to be the largest company in the world last year, valued at $3.43 trillion, before losing the spot again to Apple.
Meta has relied on NVIDIA when it comes to heavily expanding their AI developments, especially for social media. Recently, there were reports that CEO Mark Zuckerberg plans to focus more on expanding their AI developments at the cost of laying off 5% of its low-performing employees.
The company is already planning to showcase more of what its AI has to offer later this year via the upcoming Llamacon, a dedicated event that will share more information about their open-source AI.
