KT Cloud, the cloud subsidiary of KT, announced on Monday the launch of AI SERV, a new cloud service that provides high-performance graphic processing units (GPUs) for artificial intelligence (AI) inference at a reasonable cost. AI SERV is a dedicated infrastructure for AI inference services that applies slicing technology to GPUs, allowing users to divide a single GPU into multiple units as needed. This reduces the cost and increases the efficiency of using GPU resources for AI tasks.
What is AI inference and why is it important?
AI inference is the process of using trained AI models to make predictions or decisions based on new data. For example, an AI model that can recognize faces in images can be used to infer the identity of a person in a new photo. AI inference requires the use of GPUs, which are specialized hardware devices that can perform parallel computations faster than CPUs.
AI inference is important because it enables various AI applications and services that can benefit various industries and sectors, such as healthcare, education, entertainment, security, and more. However, AI inference also poses some challenges, such as the high cost of GPU infrastructure, the low utilization rate of GPU resources, and the performance degradation of sliced GPU units.
How does AI SERV solve these challenges?
AI SERV is a cloud service that solves these challenges by providing a flexible and scalable GPU infrastructure for AI inference. AI SERV uses slicing technology to divide a single GPU into five 0.2-GPU units, which can be used by different users or services simultaneously. This allows users to use only the amount of GPU resources they need, without wasting any excess capacity or paying for unnecessary costs.
AI SERV also ensures that the sliced GPU units do not suffer from any performance degradation, unlike previous slicing technologies. This means that users can enjoy the full performance of the GPU infrastructure, even when it is divided into smaller units. AI SERV also supports various AI frameworks and libraries, such as TensorFlow, PyTorch, MXNet, and more, making it compatible with different AI models and applications.
What are the benefits of using AI SERV?
AI SERV offers several benefits for users who need GPU infrastructure for AI inference, such as:
- Cost reduction: By using slicing technology, AI SERV reduces the effective price of its GPU cloud service by 70%, compared to using the same infrastructure for learning and inference.
- Performance improvement: By using high-performance GPUs and avoiding performance degradation in sliced units, AI SERV achieves a 1.4x performance increase for AI tasks, compared to using conventional GPUs.
- Flexibility and scalability: By allowing users to adjust the amount of GPU resources they need according to their demand and situation, AI SERV provides a flexible and scalable solution for various AI scenarios and workloads.
What are the future plans of KT Cloud?
KT Cloud is a leading cloud provider in South Korea that offers various cloud solutions and services for different industries and sectors. KT Cloud has been focusing on developing its AI infrastructure business since 2022, when it launched its first pay-as-you-go infrastructure service, HAC (Hyperscale AI Computing), which provides high-performance GPUs for AI learning.
KT Cloud plans to continue expanding its AI infrastructure portfolio by collaborating with domestic and foreign partners, such as Rebellions, Moreh, AMD, NVIDIA, and Google Cloud. KT Cloud aims to promote the growth and innovation of the hyperscale AI industry in South Korea and beyond, by providing affordable and accessible AI infrastructure for various AI companies and startups.