YTL AI Cloud has received the NVIDIA Exemplar Cloud status, making it one of the first cloud infrastructure providers in Asia to achieve this, and placing it among global hyperscalers and AI-specialised clouds.
YTL AI Cloud is now one of the top global NVIDIA Exemplar Cloud providers and among the first NVIDIA Cloud Partners in the Asia-Pacific region to earn this recognition. This milestone highlights YTL AI Cloud’s role as a key provider of AI infrastructure, making advanced AI computing available to businesses, researchers, developers, and public sector organizations across Malaysia and beyond.
The NVIDIA Exemplar Cloud initiative is a strict set of standards designed to minimize problems and performance issues often found in public cloud services. It confirms that a cloud provider’s infrastructure meets or exceeds NVIDIA’s high standards for performance, efficiency, and scalability. This validation helps enterprise customers achieve consistent and reliable performance based on NVIDIA’s benchmarks, ensuring they get the most value from their infrastructure investments.
To achieve the Exemplar Cloud status, YTL AI Cloud’s infrastructure underwent NVIDIA Exemplar’s tough training tests. These tests assessed a 512-GPU setup using leading AI model workloads, including DeepSeek-V3, Llama 3.1, Qwen3, Grok-1, Nemotron 3, Nemotron-H, GPT(OSS), and similar large-model tests. The validation focused on key performance measures, such as average step time, TFLOPS per GPU, and consistency across multiple runs.
“Being one of the first NVIDIA Exemplar Clouds in Asia shows that we have the talent and infrastructure to lead in building a sovereign, world-class AI infrastructure that performs at the same level as the leading hyperscalers globally,” said Dato’ Seri Yeoh Seok Hong, Managing Director of YTL Power International.
“To push the boundaries of AI, our customers demand cloud infrastructure and operations that perform consistently under the highest standards,” said Philip Lin, Chief Executive Officer at YTL AI Cloud. “Achieving NVIDIA Exemplar Cloud status validates our organization’s proven ability to handle the most intensive, real world AI workloads, reducing operational and infrastructure risk so that our customers can focus on their core business objectives.”
“YTL AI Cloud is bringing NVIDIA accelerated computing infrastructure to customers across Asia,” said Marc Hamilton, Vice President, Solutions Architecture and Engineering, NVIDIA. “Built to NVIDIA NCP Reference Architecture standards, the platform achieves NVIDIA Exemplar Cloud status, and will help advance sovereign AI adoption, frontier model development, scientific research and innovation across the region.”
For senior cloud leaders and enterprise architects, a key challenge in running large AI models is the inconsistent performance and inefficient use of hardware caused by virtualization and network issues. YTL AI Cloud solves this problem with a streamlined AI cloud design that combines GPU computing, fast networking, storage, and management. This setup allows large GPU groups to work together efficiently as a powerful AI computing system.
This is made possible through NVIDIA NVLink technology, which offers quick and efficient connections between chips, along with NVIDIA’s Spectrum-X Ethernet Networking Platform to ensure high-speed data flow, supported by NVIDIA Blackwell GPUs.
Managed by YTL AI Cloud’s unique Cloud Orchestration Platform (COP), the infrastructure simplifies operations while maintaining strict hardware control. The COP actively manages computing power, network setup, and high-performance storage to keep hardware usage high and ensure consistent results across large clusters.
YTL AI Cloud delivers a stable, scalable, and high-performance infrastructure for large model tasks, consistently meeting or surpassing NVIDIA’s performance benchmarks. The platform demonstrates exceptional reliability, internal consistency, and hardware efficiency, ensuring smooth operations without overheating or delays, thus allowing code execution as intended by developers.