WASHINGTON: Meta Platforms has announced plans to develop a new series of in-house artificial intelligence chips as part of its efforts to expand data-centre capacity and strengthen its AI infrastructure. The company revealed a roadmap for four new processors under its Meta Training and Inference Accelerator programme. The first chip in the lineup, MTIA 300, is already being used to power Meta’s ranking and recommendation systems across its platforms. Three additional chips will follow in stages through 2026 and 2027, including: MTIA 400 MTIA 450 MTIA 500 According to Meta engineering executive Yee Jiun Song, the upcoming chips will primarily focus on AI inference, the stage when an AI model processes data and generates responses for users. The demand for inference computing has surged as AI applications become more widely used. Meta plans to release the new chips every six months, reflecting the rapid expansion of its infrastructure. The company is currently building large-scale data centres to support its major platforms, including Instagram and Facebook. The new chip systems are designed with advanced cooling technologies, including liquid cooling, and are expected to occupy multiple server racks within data centres. Meta is joining other major technology companies such as Alphabet and Microsoft, which have also developed in-house processors to improve efficiency and reduce reliance on external chip suppliers. Custom chips can help companies lower energy consumption and operating costs while optimizing performance for AI workloads. Despite building its own designs, Meta continues to work with industry partners. Broadcom assists with certain design components. Taiwan Semiconductor Manufacturing Company fabricates the processors. The company also continues to purchase chips from Nvidia and Advanced Micro Devices to support its AI development. Meta expects its capital expenditure for 2026 to reach between $115 billion and $135 billion, largely driven by investment in AI infrastructure and data centres. The announcement highlights the growing global race among technology companies to build powerful AI systems and specialised hardware.
Meta Unveils Roadmap for New In-House AI Chips
