Arm joins hands with Intel, AMD, NVIDIA, Qualcomm, Microsoft, and Meta to promote the development of AI technology applications

At the second Hsinchu stop of the Arm Tech Symposia 2023, Arm announced collaborations with industry leaders including Intel, AMD, NVIDIA, Qualcomm, Microsoft, and Meta. These partnerships aim to foster the development of artificial intelligence technology applications, empowering over 15 million developers working with Arm architecture to create an array of AI applications, underpinned by appropriate foundational frameworks, technology, and specifications.

Arm indicated that currently, up to 70% of third-party AI applications on mobile devices are executed on Arm architecture processors. This new consortium will enable a broader array of AI models to run on edge computing devices, providing developers with a wealth of technological resources.

The collaboration with NVIDIA includes the adaptation of the TAO toolkit suite, facilitating the creation of optimized visual AI models through low-code execution on Ethos-U NPUs and development within the open-source frameworks of TensorFlow and PyTorch. In partnership with Meta, PyTorch will be introduced to Arm architecture mobile and embedded platforms, utilizing Arm’s extensively invested Tensor Operator Set Architecture (TOSA) as a foundation. This allows developers to deploy advanced AI or machine learning workloads requiring neural networks across various mobile and edge devices through ExecuTorch.

Furthermore, the novel 8-bit floating-point specification FP8, proposed last year by Arm in collaboration with Intel and NVIDIA, has now gained the support of AMD, Google, and Meta, culminating in the creation of the official OCP 8-bit Floating Point Specification (OFP8). Arm’s latest Cortex-A series architecture incorporates this standard, propelling the evolution of neural network AI computation.

Additionally, Arm has recently joined the MX consortium, including AMD, Intel, Meta, Microsoft, NVIDIA, and Qualcomm. This alliance seeks to standardize narrow-bit data specifications and accelerate the scaling of AI.

Arm emphasizes that it has laid a solid foundation for global AI development and anticipates a future where a multitude of AI computational advancements will be built upon Arm’s infrastructure, ranging from sensory components and smartphones to software-defined vehicles, and extending to servers and supercomputers.