Meta Rolls Out Own AI Chip to Power Facebook and Instagram, Targets Cost Cuts and Nvidia Independence

Meta Platforms Inc META is advancing its artificial intelligence (AI) capabilities by integrating a newly developed custom chip into its data centers this year, an internal document revealed. 

This move marks the second generation of Meta’s silicon technology, initially announced last year, aiming to reduce its reliance on Nvidia Corp’s NVDA dominant market presence in AI chipsets. 

Previous reports indicated the Big Techs, including Meta, Google parent Alphabet Inc GOOG GOOGL, Microsoft Corp MSFT, Amazon.Com Inc AMZN, migrating towards customized chips to trim costs and resolve compatibility and supply crises.

Also Read: Meta Platforms Is Wall Street’s Top Internet Pick, Analysts Laud AI Investments and Advertising Strength

With Meta’s extensive push into generative AI products across its platforms, such as Facebook, Instagram, and WhatsApp, and in hardware like Ray-Ban smartglasses, the company faces escalating costs from running AI workloads.

This has propelled Meta to invest billions in acquiring specialized chips and reconfiguring its data centers to support these demands, Reuters reports.

Adopting Meta’s in-house chip could lead to significant financial savings, potentially cutting hundreds of millions in annual energy costs and reducing chip purchasing expenses by billions, Dylan Patel, founder of SemiAnalysis, points out to Reuters.

As AI applications increasingly become a significant investment area for tech companies, balancing the excitement for the technology with its substantial operational costs is crucial. Meta’s plan to supplement its custom chips with hundreds of thousands of commercial GPUs signifies a strategic approach to optimize performance and efficiency for specific Meta workloads.

Despite earlier challenges, including discontinuing its first chip iteration in favor of purchasing Nvidia’s GPUs, Meta is pursuing the deployment of its “Artemis” chip for inference tasks.

This effort is complemented by the development of a more ambitious chip capable of both training and inference, akin to the functionalities of GPUs.

According to Patel, this strategic pivot could enable Meta to significantly enhance the efficiency of its recommendation models, offering a more cost-effective alternative to the current energy-intensive solutions provided by Nvidia processors. 

Meanwhile, Nvidia supplier Taiwan Semiconductor Manufacturing Company Ltd TSM is expanding its base beyond Taiwan in the U.S. and Japan. The critical contract chipmaker’s expansion symbolizes the AI demand frenzy worldwide.

Price Action: META shares traded higher by 19.80% at $473.04 on the last check Friday.

Disclaimer: This content was partially produced with the help of AI tools and was reviewed and published by Benzinga editors.

Image Credits – Shutterstock