Article · AI translation
Thanks to IT Home reader HH_KK for the tip!
IT Home reported on October 27 that, following Nvidia's $3 billion investment, AMD has also announced a $1 billion investment in Meta's equity offering (IT Home note: approximately 6.684 billion yuan at the current exchange rate), further deepening its strategic partnership with Meta.
The two companies will collaborate on CPUs, GPUs, networking, systems and software to provide computing power for Meta's artificial intelligence infrastructure and AI services.
According to the official announcement, AMD Helios rack-scale systems have been installed at Meta and entered testing. The two companies plan to expand deployment further. AMD said they are aligning product roadmaps and engineering designs around Meta's current and future workloads, with the goal of jointly building and deploying gigawatt-scale AI infrastructure.
On the CPU side, AMD EPYC processors are already supporting the dedicated computing environment behind Meta's Muse personal AI assistant. AMD said Meta has deployed millions of EPYC processors across multiple generations and is also a major customer for sixth-generation EPYC CPUs. The two companies are also collaborating on future products, including a processor code-named “Verano.”
On the GPU side, AMD Instinct GPUs are already used in Meta's production environment for workloads including AI model training, ranking and recommendation engines. AMD also said Meta is an early design customer for the Instinct MI450 architecture and will deploy an AMD Instinct GPU customized and optimized for Meta's workloads. Their collaboration will also inform future Instinct MI500-series GPUs and next-generation rack-scale systems.
On the software side, AMD and Meta are optimizing the ROCm software ecosystem through PyTorch to make AMD GPUs easier for developers to use. AMD said it has provided Day-0 support for Meta's Muse Spark Open model on Instinct GPUs, as well as Day-0 support for Muse Glimmer on PCs equipped with Ryzen AI processors and on Radeon GPUs.
The two companies previously expanded their software collaboration on large-scale AI training, using AMD Instinct MI350X GPUs to optimize recommendation models and jointly advancing improvements to PyTorch, Triton and ROCm. AMD said the resulting work will also be available to other developers.
AMD Chair and CEO Lisa Su said that, after years of technical collaboration, the partnership has expanded from EPYC processors to Instinct GPUs, Pensando networking and ROCm software. With Helios systems entering Meta's infrastructure, the two companies will expand their collaboration further.
Meta founder and CEO Mark Zuckerberg said AMD is an important partner in building the computing infrastructure for Meta's frontier AI models. Meta is continuing to invest in computing capacity to advance personal superintelligence services.
AMD to invest $1 billion in Meta's equity offering
Related reading:
“$3 Billion Investment: Nvidia and Meta Expand Collaboration on Muse AI Assistant”