Arm Stock: Nvidia Partnership Brings NVLink Integration to Custom AI Chips

TLDR

• Arm CPUs based on its technology will integrate with Nvidia’s NVLink Fusion technology for AI data centers.
• The partnership makes it easier for hyperscalers like Microsoft, Amazon, and Google to pair custom Arm-based Neoverse CPUs with Nvidia GPUs.
• Nvidia is opening its NVLink platform to various custom chips instead of requiring customers to use only Nvidia CPUs.
• Arm’s Neoverse chips will include a new protocol for seamless data transfer with GPUs in AI servers.
• The deal follows a similar September agreement between Nvidia and Intel to integrate NVLink technology into Intel server products.

Arm announced on Monday that its central processing units will support Nvidia’s NVLink Fusion technology. The integration allows data centers to connect Arm-based processors with Nvidia’s graphics processing units more easily.

The partnership targets hyperscalers building custom infrastructure. Companies like Microsoft, Amazon, and Google are developing their own Arm-based CPUs for cloud operations. These custom chips help reduce costs and give tech giants more control over their data center setups.

Arm Holdings plc American Depositary Shares, ARM

Nvidia’s decision to open its NVLink platform represents a shift in strategy. Instead of requiring customers to use Nvidia-branded CPUs, the company now allows integration with various chip designs. This flexibility gives customers more options when building AI infrastructure.

Custom Chip Ecosystem Expands

Arm licenses its instruction set technology but doesn’t manufacture chips directly. The company also sells designs that help partners build Arm-based processors faster. Under this new agreement, Arm’s Neoverse chips will feature a protocol for seamless data movement between CPUs and GPUs.

The role of CPUs in servers has changed with the rise of generative AI. While CPUs were historically the most important component, AI accelerators now sit at the center of modern infrastructure. In typical AI servers, as many as eight GPUs connect to a single CPU.

Nvidia currently sells its Grace Blackwell product, which pairs multiple GPUs with an Arm-based CPU. The company also offers configurations using CPUs from Intel or AMD. But hyperscalers prefer custom solutions that match their specific needs.

Industry Relationships and Deals

In September, Nvidia invested $5 billion in Intel. A key element of that deal enabled Intel CPUs to integrate with Nvidia’s NVLink technology. The new Arm partnership follows a similar pattern of Nvidia expanding its ecosystem.

Nvidia tried to acquire Arm for $40 billion in 2020. That deal collapsed in 2022 due to regulatory concerns in the United States and United Kingdom. Nvidia held a small stake in Arm as of February, while SoftBank remains the majority owner.

SoftBank sold its entire Nvidia stake earlier this month. The Japanese conglomerate now backs OpenAI’s Stargate project, which plans to use Arm technology alongside chips from Nvidia and AMD.

Arm is pushing deeper into the data center market. The company works with clients like Amazon Web Services and Microsoft to challenge Intel’s traditional dominance with Xeon processors. Nvidia already uses Arm technology in some of its processors.

NVLink serves as a bridge connecting chips in AI infrastructure. The technology enables high-speed data sharing between Nvidia’s AI accelerators and allows efficient task distribution. Nvidia recently launched NVLink Fusion, making the technology available to other hardware manufacturers.

Under CEO Rene Haas, Arm is offering more comprehensive chip designs to customers. The strategy aims to increase revenue and expand beyond the company’s core smartphone business into more profitable markets.

The NVLink integration into Arm’s Neoverse platform strengthens ties between the two semiconductor companies. Both firms sit at the center of the AI infrastructure boom as data centers upgrade their systems for machine learning workloads.

The post Arm Stock: Nvidia Partnership Brings NVLink Integration to Custom AI Chips appeared first on CoinCentral.