Flexible GPU-as-a-Service: Ideal for AI Teams

ASI is thrilled to announce the launch of our GPU-as-a-Service offering in ASI Cloud, marking a significant step forward in our commitment to making AI infrastructure accessible and flexible for all. Starting today, customers can leverage the power of NVIDIA L40S GPUs, with additional options, including NVIDIA L4 GPUs, coming online in the next few weeks. This new service is designed to cater to the diverse needs of teams across industries, offering unparalleled flexibility and control.

Your Trusted Partner for Secure AI Solutions

At ASI, we prioritise not only performance but also security, privacy, and local support. Much like our ASI Secure LLM, this new service is built with a focus on flexibility and dependability to empower teams across Aotearoa and beyond. We understand that every organisation’s AI journey is unique, which is why we’ve structured our GPU-as-a-Service offering to suit a variety of needs.

Service Tiers: Your Path, Your Pace

1. Bare Metal Servers: For teams looking to architect their AI infrastructure from scratch, our bare metal servers provide dedicated hardware with IPMI access. This foundational tier is ideal for organisations needing total control over their environment.
2. Pre-Configured Options: Streamline your setup with pre-installed operating systems and NVIDIA NIMs. These servers eliminate the initial configuration overhead, letting you focus on your projects.
3. Model-as-a-Service: This tier offers deployable open-source models ready for inference or fine-tuning, accelerating the process of building AI applications.
4. RAG Pipelines: At the highest abstraction level, we provide complete Retrieval-Augmented Generation (RAG) pipelines. These include embeddings and hosted vector databases, offering a turnkey solution for deploying AI-powered applications. You can bring your data and make it available to LLMs with our RAG options. This includes consuming hosted vector databases or even leveraging an entire RAG pipeline with connectors to your data and tools like SharePoint.

Built to Scale with Kiwi Ingenuity

Our GPU-as-a-Service is launching with the following specifications:

Medium Instance Tier:

• 64-Core CPU
• 256GB RAM
• Dual NVIDIA L40S GPUs (48GB each)
• 2x 1.9TB NVMe Storage

Small Instance Tier:

• 16-Core CPU
• 64GB RAM
• Single NVIDIA L4 GPU (24GB)
• 1.9TB NVMe Storage

These options provide robust computational power for diverse use cases, from training complex models to running cost-effective inference workloads. With our Kiwi-based infrastructure, your data stays local, ensuring compliance and fostering trust.

Use Cases

A few examples of what’s possible with the medium instance size:

• Inference: Securely run State Of The Art (SOTA) models like Llama 70B 3.3.
• Text to Video: Privately deploy text-to-video models like Hunyuan Video.
• Fine Tuning: Fine-tune a moderately sized AI model in 48-72 hours.
• Create Your Own AI: Re-create a GPT class model from scratch.
• Research and Machine Learning: Perform massively parallel computations for research and ML projects.

You can take control of the bare metal servers and deploy them as needed for your requirements. We’re also available to help get you up and running with end-to-end services at every layer of the AI stack.

Partnering for Success

The addition of NVIDIA L4 GPUs in the coming weeks will further expand our offerings, particularly for teams seeking budget-friendly inference solutions or smaller-scale development environments. With these enhancements, we’re confident that our platform can support a wide range of AI workloads efficiently and cost-effectively.

Looking Ahead

This is just the beginning of ASI’s journey to democratise AI infrastructure. As we continue to enhance our GPU-as-a-Service offerings, we’re excited to see how our customers harness this technology to drive innovation and create groundbreaking solutions.

For pricing details and to get started, contact our sales team today. Together, let’s build the future of AI, powered by ASI Cloud.