Intel’s domination in the CPU space meant that PC makers were at the mercy of the company. Any setbacks in new chips meant that companies were forced to use older chips, which might not sit well with customers. Perhaps the AI industry is learning from this, because, according to a report from Reuters, Meta is testing its in-house AI chips in a bid to reduce its reliance on NVIDIA.
Going for an in-house solution
Training AI isn’t cheap. A lot of data processing is required, which is why Meta has invested in a $200 billion data center. Meta is now hoping to reduce its overall infrastructure costs by developing in-house AI chips. Another benefit is that the company can reduce its reliance on NVIDIA, currently the largest supplier of AI development hardware.
Meta already spends billions of dollars on NVIDIA’s GPUs, so if this is a way to reduce its overall costs and reduce its dependence on NVIDIA, it makes sense. Meta had previously attempted to build an in-house inference silicon. However, it failed, which meant that Meta had no choice but to go back to NVIDIA.
The report also reveals some potential features of the chip. Apparently, the chip is a dedicated accelerator that will only handle AI-specific tasks. This increases power efficiency compared to GPUs, which currently handle AI training. Meta is also turning to TSMC to help them make the chip.
Expanding beyond social media
Meta is primarily known for their online services such as Facebook, Instagram, and WhatsApp. Social media helped propel the company into its current position. However, Meta knows that it cannot rely on social media forever. The company is slowly expanding beyond the social media space.
We’ve seen Meta launch other non-social media initiatives, including wearable headsets. The company teamed up with Ray-Ban to produce a pair of smart glasses, but they’re not stopping there. Last we heard, Meta plans to invest $100 billion into other headset ventures such as VR headsets and more smart glasses.
The AI space is highly-competitive at the moment. Many companies are trying their hand at developing its own AI solutions. This includes Google, Apple, and Samsung. There are also a lot of competitors from other countries, such as China, who have created similarly-capable AI models.
