Arm and Meta are still arm-in-arm and their AI plans can seemingly do no harm.
The dynamic duo has announced a long-term strategic partnership to advance AI efficiency across every level of computing — from software to large-scale data center infrastructure — aiming to bring more powerful and efficient AI experiences to billions of users worldwide.
The collaboration spans everything from milliwatt-scale devices that power on-device intelligence to megawatt-scale systems that train large AI models. It builds on years of hardware and software co-design between the two companies, combining Arm’s leadership in power-efficient computing with Meta’s strength in AI innovation, infrastructure, and open-source technologies.
“From the experiences on our platforms to the devices we build, AI is transforming how people connect and create. Partnering with Arm enables us to efficiently scale that innovation to the more than three billion people who use Meta’s apps and technologies,” Santosh Janardhan, Head of Infrastructure at Meta, said in the announcement.
AI in the cloud
Meta’s AI ranking and recommendation systems — which underpin discovery and personalization across platforms such as Facebook and Instagram — will use Arm’s Neoverse-based data center platforms.
This shift promises higher performance and lower power consumption compared to traditional x86 systems. Arm Neoverse technology will also help Meta reach performance-per-watt parity at hyperscale, reinforcing Arm’s position as a leader in efficient computing.
To achieve these improvements, both companies collaborated closely to optimize Meta’s AI infrastructure software stack, from compilers and libraries to AI frameworks, for Arm architectures.
Their work includes tuning open-source components such as Facebook GEneral Matrix Multiplication (FBGEMM) and PyTorch, leveraging Arm’s vector extensions and performance libraries.
These joint optimizations have already shown measurable gains in inference efficiency and throughput and are being contributed back to the open-source community to benefit the wider AI ecosystem.
Cloud to edge
The partnership extends beyond data centers, deepening cooperation across AI software such as the PyTorch machine learning framework, the ExecuTorch edge-inference engine, and the vLLM datacenter-inference engine.
The companies are working to enhance efficiency with Arm KleidiAI, allowing for faster inference on billions of devices. Together, these efforts aim to simplify AI model deployment and boost performance from the cloud to the edge.
These open-source initiatives are central to Meta’s AI strategy, enabling the development of tools and applications that power everything from personalized recommendations to conversational systems.
Both companies plan to continue contributing optimizations to the open-source community, allowing developers to build AI solutions on Arm architecture.
AI may be the technology of the decade, but a new $500 million initiative aims to keep humans at the heart of it.