Akamai deploys NVIDIA Blackwell GPUs for AI platform

Akamai is expanding its distributed cloud infrastructure with the deployment of thousands of NVIDIA Blackwell GPUs, creating what the company describes as a unified AI platform designed to support research, model optimisation and large-scale inference workloads.

The initiative forms the foundation of the Akamai NVIDIA Blackwell GPUs AI platform, which aims to route AI inference tasks across compute resources distributed throughout Akamai’s global network. The approach is intended to reduce latency and address data transfer challenges that arise when AI workloads rely solely on centralised datacentres.

According to the company, the architecture allows AI workloads to be processed closer to users and connected devices, enabling faster response times and improved efficiency for applications that require real-time decision-making.

AI infrastructure shifting from training to inference

The deployment reflects a broader shift in how organisations are using artificial intelligence infrastructure. Earlier phases of AI development focused primarily on training models within centralised facilities. Increasingly, the challenge lies in delivering inference efficiently once those models are deployed.

TheAkamai NVIDIA Blackwell GPUs AI platform is designed to address this stage of the AI lifecycle. By distributing GPU-based compute resources globally, the platform enables inference workloads to be processed near the point of use rather than relying on distant computing hubs.

Adam Karon, Chief Operating Officer and General Manager, Cloud Technology Group, Akamai, said the company is focusing on the operational phase of AI deployment.

“While hyperscalers continue to push the boundaries of AI training, Akamai is focused on meeting the unique demands of the inference era,” Karon said.

He noted that centralised AI facilities remain important for model development but may not always support the latency requirements of real-time AI applications.

“Centralised AI factories remain essential for building models, but bringing those models to life at scale requires a decentralised nervous system,” Karon said. “By distributing inference-optimised compute across our global fabric, Akamai is providing the scale at minimal latency required to move AI from the laboratory to real-world environments.”

Distributed AI infrastructure for real-world systems

TheAkamai NVIDIA Blackwell GPUs AI platform is structured around a distributed compute grid designed to support AI workloads that interact with physical systems. These include environments where decisions must be made quickly and close to the source of data.

Examples cited for such applications include:

• Autonomous delivery systems

• Smart grid management

• Surgical robotics

• Fraud detection systems

In these scenarios, sending data to distant computing centres for processing can introduce delays. Distributed AI infrastructure attempts to minimise this issue by moving compute resources closer to where data is generated.

The company said its architecture treats the global network as a unified computing layer capable of processing AI tasks across geographically distributed infrastructure.

NVIDIA Blackwell infrastructure integrated with Akamai network

The deployment integrates NVIDIA RTX PRO Servers, equipped with NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs, along with NVIDIA BlueField-3 DPUs. These components are combined with Akamai’s distributed cloud infrastructure and global edge network.

Akamai’s network currently spans more than 4,400 locations worldwide, enabling compute resources to be positioned closer to users and connected devices.

The GPU clusters are intended to support several stages of AI workload processing, including:

• High-performance inference: Dedicated GPU clusters handle AI workloads that require rapid responses.

• Localised model fine-tuning: Large Language Models (LLMs) can be optimised on-site to address data privacy and regional regulatory requirements.

• Post-training optimisation: Foundation models can be adapted using proprietary data to improve task-specific accuracy.

This architecture allows organisations to refine and deploy models without transferring sensitive datasets across long distances.

Part of Akamai’s expanding AI infrastructure strategy

The Akamai NVIDIA Blackwell GPUs AI platform follows earlier initiatives by the company to expand its AI inference capabilities.

In October 2025, Akamai introduced Akamai Inference Cloud, a platform designed to move AI processing closer to end users and connected devices. The company positioned the service as a way to support developers building AI applications that require fast response times and efficient data processing.

According to Akamai, distributing compute resources across its infrastructure can improve performance while reducing network delays.

The company said its architecture can reduce latency by up to 2.5 times compared with conventional centralised infrastructure. It also estimates that businesses could reduce AI inference costs by as much as 86 percent when using NVIDIA-based infrastructure within its distributed network.

Demand growing for distributed GPU capacity

Akamai said early deployments of NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs have already seen strong demand from customers building AI applications and data-intensive workloads.

As part of its infrastructure strategy, the company plans to continue expanding GPU capacity across its distributed cloud network.

The Akamai NVIDIA Blackwell GPUs AI platform represents an effort to adapt cloud infrastructure for the next stage of AI adoption. By combining distributed compute resources with GPU acceleration, the company is positioning its network to support AI workloads that increasingly operate outside traditional datacentre environments.

Read More: 

Lenovo CIO Playbook 2026 highlights AI surge

Tech Mahindra and Microsoft launch Agentic AI platform

International Women’s Day: Women leaders share insights on resilience and inclusion in enterprise tech