Arm Joins Nvidia's NVLink Fusion Ecosystem, Enabling Custom AI Server Development for Hyperscalers

Partnership Announcement

Arm and Nvidia announced at the Supercomputing '25 conference that Arm has joined the NVLink Fusion ecosystem, marking a significant advancement in AI infrastructure development 1

1

. The integration enables central processing units based on Arm's technology to connect seamlessly with Nvidia's AI accelerators through NVLink Fusion technology 2

2

.

Technical Integration Details

The partnership centers on integrating NVLink IP directly into Arm's architecture portfolio, allowing licensees to build CPUs that connect natively with Nvidia's GPU ecosystem 1

1

. Custom Neoverse chips will incorporate a new protocol that enables seamless data movement between CPUs and GPUs, utilizing Arm's AMBA CHI C2C (Coherent Hub Interface Chip-to-Chip) protocol for coherent, high-bandwidth connections 4

4

.

"Arm is integrating NVLink IP so that their customers can build their CPU SoCs to connect Nvidia GPUs," said Dion Harris, head of data center product marketing at Nvidia 1

1

. This integration removes memory and bandwidth bottlenecks that traditionally limit AI system performance.

Hyperscaler Benefits

The collaboration particularly benefits hyperscalers—large-scale cloud operators including Microsoft, Amazon, and Google—who are developing or deploying Arm-based CPUs in their cloud platforms 3

3

. These companies can now pair Arm-based Neoverse CPUs directly with Nvidia's dominant graphics processing units in custom infrastructure setups, providing enhanced control over configurations while reducing costs.

With NVLink Fusion, hyperscalers can significantly reduce design complexity, save development costs, and reach the market faster 1

1

. The technology enables up to eight GPUs to be paired with a single CPU in AI server configurations, prioritizing accelerator performance for processing intensive AI models.

Strategic Implications for Both Companies

For Arm, the partnership provides a major competitive advantage in the data center market. As an IP provider, Arm gains the ability to offer licensees a ready-made pathway to build CPUs compatible with market-leading Nvidia GPUs for AI and high-performance computing applications 1

1

. This makes Arm-based designs more attractive to hyperscalers and sovereign cloud builders seeking custom CPU solutions.

For Nvidia, the partnership massively expands the pool of CPUs that can serve natively in Nvidia-centric AI systems without requiring Nvidia to build all processors itself 1

1

. By enabling Arm licensees to integrate NVLink directly into their systems-on-chips, Nvidia ensures future Arm-based processors will be architected around or compatible with Nvidia GPUs.

Market Context and Competition

The announcement represents Nvidia's strategy of opening its NVLink platform to integrate with various custom chips rather than forcing customers to use its CPUs exclusively 2

2

. Nvidia currently sells Grace Blackwell, an AI product pairing multiple GPUs with an Nvidia-branded Arm-based CPU, alongside configurations using Intel or AMD CPUs.

This development could pose significant challenges to competitors, particularly AMD and Intel, as it reduces the appeal of alternative interconnect solutions and strengthens Nvidia's position in sovereign AI projects 1

1

. In September, Nvidia invested $5 billion in Intel to enable Intel CPUs to integrate into AI servers using NVLink technology, demonstrating the strategic importance of these partnerships 3

3

.