Extending a comprehensive lead in the AI race does not necessarily mean that you continue luring talent from other companies by offering them massive paydays, because Meta is also encountering a problem that it cannot keep a lid on; rising capital expenditure. Generative AI requires a heap of GPUs to tackle millions of queries, with NVIDIA charging a notable sum for its chips that require both storage and adequate cooling to work optimally. These skyrocketing costs leave Meta with just one alternative; develop its own AI chip codenamed ‘Arke.’ According to the latest report, it will seek the help of MediaTek to achieve its own goal, with the silicon expected to use the cutting-edge 2nm process.
Arke was never intended to be added to Meta’s custom silicon lineup, but the company wants to exercise a more cost-effective approach
With MediaTek already said to begin 2nm chip tape-out in the fourth quarter of 2025 as it looks to take advantage of TSMC’s newest lithography, the Taiwanese firm would already have garnered experience in designing the ‘state of the art’ chipsets, making it the ideal partner for Meta. The social media giant has likely wanted to ally with the fabless semiconductor manufacturer because the two have already had a deal in developing custom silicon for smart glasses.
According to DigiTimes, Arke was never included in Meta’s original plan, with the company intending to mass produce another chip, Iris, by the end of 2025, with Olympus being next and said to leverage TSMC’s 2nm ‘N2P’ architecture. However, the need to maintain cost-effectiveness and reduce dependency on NVIDIA likely forced Meta to inject Arke into its current lineup, with the chip said to be used for AI inference.
When we previously reported about Meta’s in-house chip ambitions, the company had kicked off small-scale development, with its custom chip said to be utilized for training purposes from 2026. However, no codename details were shared on the previous occasion, but it was mentioned that TSMC would be tasked with manufacturing this solution on an undisclosed lithography.
Given that Meta’s AI infrastructure costs are estimated to soar to between $114 billion and $119 billion this year, it is not surprising that the company has had to resort to developing its solution to curb these ludicrous costs. As for Arke’s, the mass production schedule based on the report is in the first half of 2027.
News Source: DigiTimes
About the author: Omar Sohail is a reporter and analyst for Wccftech's mobile section, specializing in the technology and business of the mobile industry. His expertise lies in the intricate hardware supply chain, covering developments in semiconductor manufacturing, chip lithography, and camera sensor technology.
Follow Wccftech on Google to get more of our news coverage in your feeds.
