# AMD expands Meta partnership and invests $1B in its stock sale

> Helios racks are running at Meta, a custom Instinct GPU is coming, and AMD is buying $1 billion of Meta's equity offering.

URL: https://claudegpt.si/news/amd-meta-partnership-helios-instinct  
Published: 2026-10-27T14:56:12.090Z  
Author: ClaudeGPT Si Newsroom  
Section: AI News  
Topics: AMD, Meta, AI chips, Data centers  
Publisher: ClaudeGPT Si

## Key takeaways

- AMD and Meta expanded their partnership across CPUs, GPUs, networking, racks and software on October 27.
- AMD will invest $1 billion in Meta's equity offering; AMD Helios racks are installed and under testing at Meta.
- Meta will deploy a custom Instinct GPU, but no chip counts, dates or gigawatt figures were given.

AMD and Meta said on October 27 that they are widening their chip partnership, and AMD will put $1 billion into Meta's stock offering. According to [AMD's press release](https://ir.amd.com/news-events/press-releases/detail/1302/amd-and-meta-advance-strategic-partnership-to-power-metas-superintelligence-infrastructure), the AMD Meta partnership now spans processors, GPUs, networking, full server racks and software.

It matters because Meta is paying for its AI buildout by selling a lot of stock, and its suppliers are putting money in alongside it. Earlier this week, [Nvidia said it would invest $3 billion in Meta](/news/nvidia-and-meta-muse-3b-investment). Meta now has both of the big AI chip makers as shareholders and suppliers.

## What AMD says is already running at Meta

- **Helios racks.** AMD's Helios rack-scale systems are installed and being tested at Meta. AMD says broader deployment is expected "in the near future" and gives no date or size.
- **EPYC CPUs.** Meta has deployed millions of AMD EPYC processors across several generations and is a lead customer for the 6th generation. The two companies are also working on a later generation codenamed Verano.
- **Instinct GPUs.** AMD Instinct GPUs are in production at Meta for training and for ranking and recommendation workloads. Meta helped define the MI450 architecture and will deploy a custom Instinct GPU tuned to its own workloads. AMD says the work also feeds its MI500 series.

## The software side

AMD and Meta are working on ROCm, AMD's GPU software stack, through PyTorch. The release says AMD's GPUs support Meta's Muse Spark Open models from day one, and that Muse Glimmer runs on Ryzen AI PCs and Radeon GPUs. They are also tuning recommendation models for the MI350X and contributing to PyTorch, Triton and ROCm.

The companies describe the buildout as gigawatt scale. The release does not give a gigawatt figure, a chip count, a price or a delivery schedule.

## What we don't know

The release does not say how many GPUs or racks Meta is buying, what the custom GPU's specifications are, or what AMD gets in return for its $1 billion beyond the shares in Meta's offering. Both quotes are from the executives: AMD chief executive Lisa Su said "Helios now running at Meta" lets the companies take the partnership "to a new scale in the superintelligence era."

AMD's forward-looking statements also flag supply limits, including memory, and customer concentration as risks. Watch AMD's next earnings report for any figure tied to Meta.

## Sources

- [AMD and Meta Advance Strategic Partnership to Power Meta's Superintelligence Infrastructure](https://ir.amd.com/news-events/press-releases/detail/1302/amd-and-meta-advance-strategic-partnership-to-power-metas-superintelligence-infrastructure), AMD
