First crack in Nvidia AI wall? Meta set to use own AI chips in its servers in 2024 — but it will still use Nvidia H100 GPUs as well in its datacenters for now

Key Takeaways:

– Meta Platforms plans to deploy its own custom-designed AI chips, called Artemis, into its data centers this year.
– This move aims to reduce Meta’s reliance on Nvidia’s H100 chips and control escalating costs.
– Meta has been investing billions of dollars to boost its computing capacity for generative AI products integrated into Facebook, Instagram, and WhatsApp.
– Dylan Patel, founder of SemiAnalysis, suggests that Meta’s own chips could save the company hundreds of millions of dollars in energy costs and billions in chip purchasing costs.
– Despite this, Meta will continue to use Nvidia’s H100 GPUs in its data centers in the near future, with plans to have 350,000 processors in service by the end of the year.
– Meta’s in-house AI silicon project had previously discontinued its first iteration in favor of Nvidia’s GPUs but has now experienced a positive turn with the deployment of Artemis.
– Artemis is designed for AI inference, ranking judgments, and generating responses to user prompts.
– Meta sees its internally developed accelerators as complementary to commercially available GPUs for Meta-specific workloads.
– While Meta’s move may indicate a crack in Nvidia’s AI dominance, Nvidia’s GPUs will still play a significant role in Meta’s AI infrastructure for now.

TechRadar:

Meta Platforms, the parent company of Facebook, plans to deploy its own custom-designed artificial intelligence chips, codenamed Artemis, into its data centers this year, according to an internal document seen by Reuters. This move could potentially reduce Meta’s dependence on Nvidia‘s market dominant H100 chips and control the escalating costs of running AI workloads.

Meta has been investing billions of dollars to boost its computing capacity for the power-hungry generative AI products that it is integrating into services such as Facebook, Instagram, and WhatsApp.  This involves acquiring specialized chips and reconfiguring data centers to accommodate them.

Source link

AI Eclipse TLDR:

Meta Platforms, the parent company of Facebook, is planning to deploy its own custom-designed artificial intelligence chips called Artemis into its data centers this year. The move is intended to reduce Meta’s reliance on Nvidia’s H100 chips and control the increasing costs of running AI workloads. Meta has been investing billions of dollars to enhance its computing capacity for integrating power-hungry generative AI products into services like Facebook, Instagram, and WhatsApp. By deploying its own chip, Meta could potentially save hundreds of millions of dollars in annual energy costs and billions in chip purchasing costs. However, Meta will still continue to use Nvidia’s H100 GPUs for now, with plans to have around 350,000 processors in service by the end of the year. The deployment of Artemis marks a positive development for Meta’s in-house AI silicon project after discontinuing its initial chip in favor of Nvidia’s GPUs in 2022. The new chip, Artemis, is designed for AI inference and is expected to deliver a balance of performance and efficiency on Meta-specific workloads. While this move may signal a crack in Nvidia’s dominance in AI, Nvidia’s GPUs will still play a significant role in Meta’s AI infrastructure for now.