Meta has joined hands with Arm in a multi-year partnership to enhance AI performance across devices and data centres.
Meta and Arm announced a long-term collaboration aimed at making artificial intelligence (AI) systems faster, more efficient, and scalable for billions of users worldwide. The partnership focuses on optimising AI for both large data centres and small, low-power devices like smartphones.
Meta plans to use Arm’s Neoverse chips in its massive data centres. These chips are designed to deliver performance comparable to traditional Intel or AMD processors but use much less energy, helping Meta reduce electricity costs while speeding up AI operations. This is crucial for Meta’s platforms, including Facebook, Instagram, and the Meta AI app, which rely on AI for tasks like recommending posts, suggesting friends, and powering tools such as chatbots and image filters.
The partnership also focuses on software. Meta’s open-source AI tools, including PyTorch (for training AI models) and ExecuTorch (for running AI on devices), are being optimised to work efficiently with Arm chips. This makes AI models faster, more responsive, and less energy-consuming, whether they run in the cloud or directly on phones.
Arm’s KleidiAI technology will help AI models perform better on small battery-powered devices, improving features like voice assistants, filters, and image recognition without draining the phone’s battery. Meta and Arm said these improvements will be shared with the global developer community, allowing others to build efficient AI applications.
This partnership will enable Meta to scale AI for more than three billion users, making advanced AI features accessible across its platforms. For Arm, it represents a step beyond smartphones, expanding into high-performance servers and computers that power modern AI systems.
By combining Meta’s AI expertise with Arm’s energy-efficient chip designs, both companies aim to make AI faster, greener, and more widely available. Users can expect quicker AI responses, better app performance, and lower energy usage in future Meta services and devices.
