Chennai, Tamil Nadu Feb 1, 2024 (Issuewire.com) - In this year, Facebook-owner Meta plans to deploy into data centers. They are going to launch a new version of the custom chip which can support Meta’s AI push. This has been disclosed by an internal document of this business giant this Thursday. Last year, Meta announced this chip, which is going to be a second generation of an in-house silicon line.
This could reduce Meta’s reliance on the Nvidia chips. These Nvidia chips rule this market and control the rapidly increasing expenses associated with the workload of running AI. Now, Meta will be in this race to launch their in-house powered chips to power AI-memo.
Meta is the world’s largest social media company. They have been moving quickly to boost their computing capabilities to a great extent. This could aid the AI –products which are essentially power-ravenous generative to push the efficacy of the Meta exclusive apps like Facebook, Instagram, WhatsApp, and other hardware devices like Ray-Ban smart glasses. Meta is spending billions of dollars to accumulate arsenals of dedicated chips. These specialized chips will reconfigure data centers to accommodate these applications.
Meta operates on a very large and wide scale, and with a successful deployment of this chip has the potential to save hundreds of millions for this company each year. According to the founder of the silicon research group SemiAnalysis, Dylan Patel, the chip purchasing costs are in billions which escalates their annual energy costs.
A Meta spokesperson confirmed the plan and also updated on the production aspects of this chip in 2024. That person said, "We see our internally developed accelerators to be highly complementary to commercially available GPUs in delivering the optimal mix of performance and efficiency on Meta-specific workloads," in a statement.
Meta CEO Mark Zuckerberg also mentioned this last month. According to him, the company has planned to have this chip by the end of 2024. And this could be the last year when they have purchased roughly Nvidia’s 350,000 flagship "H100" processors. From next year, Meta would amass the equivalent compute capacity of 600,000 H100s in total.
Media Contact
Daniel Martin dm3805508@gmail.com
