The Difficulties with Data Centres Adopting Gen AI Workloads
The rapid ascent of AI, particularly the widespread implementation of Gen AI, is already significantly altering business operations across nearly every sector. In the retail industry, for instance, companies are utilising Gen AI to maintain optimal inventory levels, cut costs through enhanced logistics, and personalise customer interactions. In healthcare, Gen AI is facilitating easier access to medical information for patients and aiding physicians in diagnosing patients more quickly and accurately. Within the financial services sector, firms are benefiting from Gen AI through improved client care, personalised financial advice, and the transformation of branches into experience centres.
Given this swift pace of change, IT organisations are racing against time to modernise and expand their data centres to accommodate complex new compute clusters, stringent data compliance requirements, energy efficiency needs, and various other Gen AI-specific deployment considerations. According to the latest Cisco AI Readiness Index, 61% of companies report having a maximum of one year to implement their Gen AI strategy before it negatively impacts their business.
Despite the recognition by many organisations of Gen AI's vast potential to enhance business outcomes, determining the way forward can be challenging due to numerous barriers to progress. The Cisco AI Readiness Index indicates that 86% of companies feel they are not fully prepared to harness Gen AI to its fullest potential.
Data centre sustainability and the bottom line
Rising energy consumption and costs, along with compliance with stringent new environmental regulations, are some of the biggest challenges organisations face as they look to integrate Gen AI-capable solutions into their data centres. These factors can also be motivating drivers that align with sustainability goals. In addition, the EU’s Corporate Sustainability Reporting Directive (CSRD) and Energy Efficiency Directive (EED) have both come into effect in 2024, introducing an obligation to monitor and report the energy performance of data centres and increasing the urgency for organisations to take strong steps towards zero-emissions data centres.
According to the International Energy Agency, global electricity demand, driven by Gen AI growth, is expected to double by 2026, starting a vicious cycle of requiring more power, which generates more heat, which requires more efficient cooling methodologies, which also comes at a cost. With air cooling consuming up to half of the total energy in the data centre, liquid cooling is more efficient but requires organisations to implement additional technologies at additional cost.
Increased energy consumption and more sustainable cooling solutions are significant challenges for organisations as they deploy their Gen AI strategies. These include higher operational costs, difficulty in securing energy supply, and the potential for legal and financial penalties for not meeting various sustainability regulations.
The solutions to these challenges are multifaceted. Choosing energy-efficient hardware purpose-built for the requirements of Gen AI is essential, as is adopting a platform approach to energy management that provides real-time visibility and actionable insights into energy usage. Beyond energy usage and cooling, organisations also need to implement processes designed to manage resources efficiently and reduce waste. These include recycling programmes for equipment, choosing hardware with a longer lifecycle, and designing data centres to minimise the use of materials and resources during construction.
As Gen AI workloads continue to become more complex, achieving sustainability through these and other best practices will be vital to the bottom line, customers, and regulatory agencies.
Overcoming uncertainty and complexity
Uncertainty about how to scope, define, design, and deploy Gen AI in existing data centre environments and data centre operations is a critical concern within IT teams. Waiting to have all the answers in hand before moving forward means falling further behind the competition for the business. It’s imperative for IT organisations to start putting the required infrastructure in place now to remain competitive with early adopters. Already on the edge of innovation, industries like financial services and manufacturing are adopting Gen AI to, for example, enable personalised financial guidance for clients or improve manufacturing yields.
Dealing with the complexity inherent in Gen AI infrastructure is another common deployment barrier for many organisations. In a survey of over 8,000 global companies, Cisco found that 95% of businesses are aware that Gen AI will increase infrastructure workloads, but just 17% of organisations have network and compute platforms that can handle this complexity. A lack of manageability and a shortage of Gen AI-specific IT skills in many organisations are additional factors that are compounding this problem, making data centre operations that much more challenging.
A platform approach to managing Gen AI deployments and data centre operations offers greater manageability by simplifying the inherent complexity of many AI-specific tasks. By automating activities that would otherwise need to be done manually by highly skilled and scarce IT staffers, a platform approach reduces manual error and improves operational efficiency by giving IT teams a set of advanced tools that simplify and streamline data centre operations and monitoring. Equipped with these tools, organisations can refocus their efforts on providing high-quality digital experiences and achieving their strategic business objectives sooner.
Cisco actively works to lower the barriers to Gen AI adoption in the data centre using a platform approach across the entire AI infrastructure stack to address complexity and skills challenges while helping monitor and optimise energy usage. Discover how the Cisco AI-Native Infrastructure for Data Center solution can help your organisation build your AI data centre of the future.
******
Make sure you check out the latest edition of AI Magazine and also sign up to our global conference series - Tech & AI LIVE 2024
******
AI Magazine is a BizClik brand