Saudi Arabia has moved another major AI infrastructure project from announcement to reality.
AMD, Cisco and Saudi AI company HUMAIN have switched on a large-scale AI computing cluster in the Kingdom. It uses AMD Instinct MI355X GPUs alongside Cisco networking technology. The infrastructure is already serving HUMAIN customers in Saudi Arabia and other markets.
AMD CEO Lisa Su described the site as the company’s largest inference cluster outside the United States. More capacity is coming too. The partners plan to add up to 250 megawatts of AI infrastructure starting in 2027.
AMD’s Largest Inference Cluster Outside the US Is Now Live
The important word here is live. Saudi Arabia has announced several huge AI infrastructure projects over the past two years. This one has reached production. HUMAIN is already offering computing capacity to customers through the new infrastructure. The system uses AMD Instinct MI355X accelerators and AMD EPYC processors. Cisco provides the high-speed networking that connects the hardware.
HUMAIN Plans to Deploy 13,000 AMD GPUs This Year
The deployment is already substantial, but HUMAIN intends to go much further. CEO Tareq Amin said the company expects to deploy 13,000 AMD GPUs this year. That gives Saudi Arabia a significant pool of computing power for AI workloads. It also strengthens HUMAIN’s position as a regional provider of AI infrastructure. The company wants to serve customers inside the Kingdom and beyond.
AMD MI355X GPUs Sit at the Center of the Cluster
HUMAIN is using AMD’s Instinct MI355X accelerators for the live deployment. These chips target demanding AI training and inference workloads. Each MI355X GPU includes 288GB of HBM3E memory and offers 8TB/s of memory bandwidth. Eight GPUs can operate together on a single baseboard. That creates 2.3TB of shared high-bandwidth memory. Large AI models can therefore process more information without constantly moving data between memory systems.
Cisco Is Building the Network Around the GPUs
Powerful GPUs alone do not create an effective AI cluster. Thousands of accelerators need to exchange enormous amounts of data quickly. Cisco handles that part of the deployment. The infrastructure uses Cisco Silicon One networking and 800G optics. Cisco N9000 Series systems connect the AMD accelerators through an AI-optimized network fabric. The companies designed it for low latency, resilience and large-scale operation.
HUMAIN Can Now Sell GPU Capacity as a Service
The new infrastructure gives HUMAIN something commercially important: working capacity it can sell today. The company can provide GPU-as-a-service for different AI workloads. Customers can use the platform for model training and inference without building their own massive data centers. That could appeal to enterprises, AI developers and government organizations. It also turns Saudi Arabia’s AI investment into an operating service rather than another future infrastructure promise.
Another 250MW of AI Infrastructure Is Coming
AMD, Cisco and HUMAIN are already planning the next stage. The companies want to deploy up to 250 megawatts of additional AI infrastructure beginning in 2027. The first capacity should come online during the second half of that year. This phase will use AMD’s newer Instinct MI400 Series GPUs. AMD EPYC processors and ROCm software will also form part of the platform. Cisco will continue supplying networking and critical infrastructure.
The Bigger Target Is 1GW by 2030
The 250MW expansion sits inside a much larger plan. AMD, Cisco and HUMAIN previously announced a joint venture targeting up to 1 gigawatt of AI infrastructure by 2030. The companies say that project remains on track. Their original joint venture announcement outlined an initial 100MW deployment in Saudi Arabia. It also positioned AMD and Cisco as key technology partners for the buildout.
Saudi Arabia Wants More Control Over AI Infrastructure
There is another reason this project matters beyond raw computing power. The platform is designed around sovereign AI. Organizations can gain more control over where their data stays. They can also decide how models are customized, deployed and governed. Open models and AMD’s ROCm software form part of that approach. Cisco says its network architecture also supports interoperability. That could appeal to governments and regulated industries that cannot simply send sensitive workloads anywhere.
AMD and HUMAIN Have a Much Bigger $10 Billion Plan
The relationship between AMD and HUMAIN extends well beyond this cluster. In May 2025, the companies announced plans to invest up to $10 billion in AI infrastructure. Their collaboration targeted 500MW of AI compute capacity over five years. The broader plan includes computing centers stretching from Saudi Arabia to the United States. HUMAIN handles areas such as data centers, power systems and connectivity. AMD supplies its AI computing portfolio and ROCm software ecosystem.
HUMAIN Is Also Offering Smaller Companies an ‘AI in a Box’
Not every Saudi business needs thousands of GPUs. HUMAIN and AMD are addressing that part of the market too. At LEAP 2026, they announced an “AI in a Box” offering for enterprises beginning their AI journey. The package uses AMD Instinct GPUs and EPYC CPUs. It targets industries such as manufacturing, financial services, and oil and gas. Companies can test AI applications without immediately committing to hyperscale infrastructure.
Saudi Arabia’s AI Push Is Moving From Spending to Delivery
Saudi Arabia has spent heavily on AI ambitions, partnerships and infrastructure plans. The HUMAIN deployment shows a different stage of that strategy. Computing capacity is now operating and serving customers.
That matters because demand for AI infrastructure is rising quickly. It is especially important for organizations that want local control over their models and data. HUMAIN is trying to capture that demand before the market becomes even more crowded.
AMD gains something too. Nvidia still dominates much of the AI accelerator market. A major operational deployment gives AMD another chance to prove its hardware at scale.
And Saudi Arabia gets what it has been chasing: more AI computing power running inside the Kingdom.
The next test is no longer whether the hardware can be installed. It is whether enough customers arrive to fill it. The companies appear confident. Their next 250MW expansion is already scheduled to begin in 2027.
Sources
Middle East AI News — Largest AMD inference cluster outside the US goes live in KSA
Read the original article
AMD — AMD, Cisco and HUMAIN Expand Saudi Arabia’s AI Infrastructure as AMD Instinct Systems Go Live
Read the AMD announcement
