On April 10, Meta Platforms, the parent company of Facebook, introduced its latest proprietary chip, the "Meta Training and Inference Accelerator" (MTIA). This chip is designed to enhance artificial intelligence (AI) services, particularly in content ranking and recommendations on Facebook and Instagram. By leveraging the MTIA, Meta aims to reduce its dependency on external semiconductor providers such as NVIDIA. This announcement follows Meta's earlier release of the first-generation MTIA product last year.
The new MTIA chip utilizes TSMC's 5-nanometer process, offering a significant efficiency boost compared to its predecessor, with three times the performance. Currently deployed in data centers, the MTIA chip powers various AI applications and is undergoing further development to expand its capabilities, including support for generative AI operations.
Meta's strategic focus on AI services underscores the increasing demand for computational power in the digital landscape. Last year, the company launched its own AI models, competing with industry-leading solutions like OpenAI's ChatGPT. Additionally, Meta integrated new generative AI features into its social media applications, enhancing user experiences with customized stickers and chatbot characters resembling celebrities.
In October of the previous year, Meta announced plans to invest up to $35 billion in AI-supporting infrastructure, including data centers and hardware. CEO Zuckerberg emphasized the significance of AI investment, stating that it would be the company's largest area of focus by 2024.
While Meta's internal chip development efforts aim to reduce reliance on external providers, significant investment may still flow to companies like NVIDIA, which produce GPUs powering AI models. Earlier this year, Zuckerberg announced the purchase of 350,000 H100 chips, each priced at a substantial amount.
The tech industry's shift towards internal chip development is evident, with Meta joining other major players such as Amazon's AWS, Microsoft, and Google in this endeavor. Despite these efforts, demand for NVIDIA's AI accelerators remains high, reflecting the ongoing growth of the AI market.
NVIDIA's strong position in the global tech landscape is evident, as it ranks as the third most valuable tech company worldwide, trailing only Microsoft and Apple. The company's sales to data center operators in the fiscal year 2024 saw a significant increase, totaling $47.5 billion, up from $15 billion the previous year. Analysts predict continued growth, with sales expected to more than double by the fiscal year 2025.