Back to top

Image: Bigstock

META Unveils Next-Gen Custom-Made Chips for AI Workloads

Read MoreHide Full Article

Meta Platforms (META - Free Report) recently unveiled its next-generation custom-made chips tailored for AI. This strategic move underscores Meta’s anticipation of the growing demand for AI-driven solutions across its platforms and services.

At the heart of this endeavor lies the Meta Training and Inference Accelerator (MTIA), a custom-designed chip meticulously crafted to cater to Meta’s unique AI workloads. The first iteration, MTIA v1, debuted last year, serving as a testament to its dedication to optimizing compute efficiency for deep learning recommendation models.

Building upon the success of its predecessor, the latest iteration of MTIA promises substantial enhancements in compute and memory bandwidth. With a keen focus on striking the perfect equilibrium between compute power, memory capacity and bandwidth, this chip is poised to elevate the performance of Meta’s ranking and recommendation models, thereby enriching user experiences across its platforms.

Meta’s holistic approach to custom silicon underscores its commitment to innovation and scalability. By investing in domain-specific silicon, Meta aims to unlock unprecedented efficiency, surpassing commercially available GPUs.

Moreover, Meta’s strategic investments extend beyond compute silicon, encompassing memory bandwidth, networking and capacity, ensuring a robust foundation for future advancements.


As Meta continues to expand the scope of MTIA, including support for generative AI (gen AI) workloads, it reaffirms its long-term vision of pioneering the most powerful and efficient infrastructure for AI.

In March this year, Meta announced two 24k GPU clusters that are built on top of Grand Teton, OpenRack and PyTorch. These clusters support current and next-generation AI models, including Llama 3.

Meta expects that its infrastructure build-out will include 350,000 NVIDIA H100 GPUs as part of a portfolio that will feature compute power equivalent to nearly 600,000 H100s by the end of 2024.

Meta Leverages Gen AI to Boost User Experience

Meta has been leveraging gen AI to boost user experience across its apps. AI stickers, AI-based editing tools, advanced conversational assistant — Meta AI — and others to improve user engagement.

It has been experimenting with incorporating gen AI to help creators create different captions and thumbnails.

Meta is leveraging the large language model technology, which powers Meta AI, to give English-language users in various markets options for AI-generated post-comment suggestions and community chat topic suggestions in groups.

It has added a feature called reimagine to Meta AI on Messenger and Instagram. This is an enhancement to the most popular text-to-image generation feature — imagine — across its messaging apps.

Meta has also started to roll out Reels in Meta AI chats. It is also exploring ways to make Facebook more attractive and fun-filled for users leveraging Meta AI.

Moreover, it is collaborating with industry partners on common technical standards for identifying AI content, including video and audio. The identification helps Meta label AI-generated images that users post to Facebook, Instagram and Threads. This is expected to further improve user experience.

Zacks Rank & Other Stocks to Consider

Meta currently has a Zacks Rank #2 (Buy).

Shares of Meta have gained 47.8% year to date compared with the Zacks Computer & Technology sector’s growth of 10.9%.

Meta expects total revenues between $34.5 billion and $37 billion for the first quarter of 2024.

The Zacks Consensus Estimate for first-quarter 2024 is pegged at $36.15 billion, indicating 26.2% year-over-year growth. The consensus mark for earnings is pegged at $4.29 per share, unchanged over the past 30 days.

NVIDIA (NVDA - Free Report) , Veeva Systems (VEEV - Free Report) and Synopsys (SNPS - Free Report) are some other top-ranked stocks in the broader sector, each sporting a Zacks Rank #1 (Strong Buy) at present. You can see the complete list of today’s Zacks #1 Rank stocks here.

NVIDIA shares have returned 83% year to date. The long-term earnings growth rate is pegged at 30.93%.

Veeva shares have gained 9.7% year to date. The long-term earnings growth rate is pegged at 24.09%.

Synopsys shares have gained 12.5% year to date. The long-term earnings growth rate is pegged at 17.51%.

Published in