Do you have a question? Want to learn more about our products and solutions, the latest career opportunities, or our events? We're here to help. Get in touch with us.
Imagine your organisation having its own secure, custom-trained ChatGPT-like artificial intelligence (AI) assistant, or an in-house AI agent that is constantly checking and identifying potential cybersecurity risks. These types of AI applications are the future of business, but they take considerable processing power to train and scale.
AI needs the right infrastructure and processing power and the cost of the powerful GPUs needed for complex AI models is significant.
“Because GPUs do all the computation behind AI, a lot of investment has gone into them and the infrastructure to do AI at scale is not cheap,” explains Richard Chambers, Associate Director, Datacom Infrastructure Products.
"Data centre-grade GPUs for large language models start at $50,000 per card, with total infrastructure costs quickly escalating into hundreds of thousands of dollars when running 24/7. Whether your organisation needs three GPUs or 30, purchasing this hardware outright creates massive upfront expenses that are difficult to justify without consistent utilisation."
With such high upfront costs, Datacom has developed GPU as a service (GPUaaS), which gives customers consumption-based access to powerful GPUs, pre-integrated frameworks and high-performance compute and storage services. This eliminates the need to spend hundreds of thousands of dollars in upfront costs and provides scalability and local expertise when you need it. Pay by the minute, the hour or day. Alternatively reserve time for a year or three.
“The cost of buying AI infrastructure at scale can be prohibitive for anyone outside large organisations like the big banks, insurance companies and utilities, whose business tends to be capex-rich, maybe with deeper pockets than most, able to self-fund sizable investment. Everyone else is a bit more cost-conscious, so GPU as a service addresses the market for overhead focused businesses to evolve their AI journeys.”
Datacom’s GPUaaS delivers flexible enterprise-grade GPU resources without the cost or complexity of building and maintaining your own. Powered by Dell infrastructure, underpinned by Nvidia GPUs, the infrastructure is local so your data remains onshore.
“There’s an added benefit of your AI infrastructure being hosted locally by Datacom, your friendly local service provider,” says Chambers. “That is one of our core value propositions: being accessible to our customers is at the heart of who we are, we’re accessible to you from the leadership team all the way down. Our GPUs are hosted locally, in a data centre owned by Datacom, that’s governed by New Zealand regulation and managed by Datacom teams. GPUaaS is entirely sovereign, side-stepping the potential geopolitical complications you increasingly hear in the news."
Locally based infrastructure provides security and compliance advantages for any organisation dealing with critical infrastructure or sensitive data. It is available for Datacom customers as either a standalone offering or as part of a larger package of services.
“We have customers who are already deeply integrated into our data centres, cloud solutions, managed services, app development, fronted by our Contact Centre teams – the full gamut,” says Chambers.
“Equally, customers might just take a quarter rack and GPUaaS and say, ‘We need six months to do some proof-of-concept testing’, using their own software development team.”
GPUaaS allows organisations to trial AI in a safe, cost-effective and production ready enviroment. This can accelerate AI adoption and innovation, in a controlled environment, without the business needing to spend huge sums on infrastructure.
Datacom works hand-in-hand with Dell, which provides the hardware for the GPUaaS, including the Nvidia chips. For most organisations in Australasia, the wait time for the latest GPU technology could be long – if they could even get their name on the back order, with global demand for the most powerful GPUs snapped up by the major tech giants.
“The best infrastructure is expensive and hard to come by,” says Chambers. “There is a backlog for the latest GPU units and long lead times; it’s an arms race everyone wants the latest and greatest. As a large IT organisation in New Zealand and Australia, Datacom has the commercial relationships in place for large orders of laptops, servers, storage arrays and software. We can leverage our supply chain so our customers can get access to this technology. Fortunately, we have Dell as a partner and they say, ‘Yeah, of course we can help’. So it’s Dell asking Nvidia for chips, not just a local business making the request.”
Dell provides storage, data protection and compute solutions for Datacom’s private cloud and they've created the AI Factory “a curated framework for an end-to-end AI infrastructure and partner ecosystem”. Dell provides specific configurations, in conjunction with Nvidia, that are ready made, fully supported, giving faster time to value.
Datacom is already using their GPUaaS to explore how locally hosted AI platforms help augment Datacom’s wider investment in AI, including a secure, self-hosted AI model for natural language processing tasks. That is just the first step in the journey.
“Trying to predict AI out much past a year is tough, but, we’ll continue to scale our offering,” says Chambers.
"We’ll also offer more choice for customers, with a ‘good, better, best’ model as we lifecycle GPU through our platform. With Dell’s help it’s likely we’ll introduce more GPU options and greater value for our customers with specific tooling for AI orchestrations and automation.”
From there, it’s about providing a superior service that lets organisations pay for what they need, giving them scalable, cost effective and flexible AI tools within their businesses.
“These GPUs are the core building blocks of the AI infrastructure journey. We’ll certainly be adding more functionality to our services, so that GPUaaS becomes more of a platform, with more toolsets and more value on top of the infrastructure. That’s the second phase of the journey. From the infrastructure to the code, we’ve got long-term aspirations to do much more in this space.”
Australia's critical infrastructure industries run on digital, interconnected infrastructure, from power grids to cloud-enabled services. A complex web of networks powers growth, but also exposes critical systems to cyber, physical and operational risks. It's imperative organisations in critical infrastructure industries are resilient. Digital resilience is the capacity to anticipate, withstand, adapt to and recover from disruption, including cyber attacks and supply chain shocks. It weaves technology, process, compliance and culture to keep operations secure, scalable and future-ready.
