Evolving a data centre into an Artificial Intelligence factory
The industrial landscape is undergoing a transformation. Traditionally, foundries have converted raw materials into basic components while factories assemble products. However, emerging AI factories and foundries now represent a novel approach to product creation and innovation.
These new creations are poised to fundamentally alter software development, resource consumption, and management. This shift is impacting how businesses operate and the value they deliver.
Unlocking productivity with Generative AI (GenAI)
Companies are increasingly leveraging GenAI to enhance productivity. Hitachi Vantara, for instance, utilises GenAI copilots and large language models (LLMs) to support customer service, marketing, and sales teams.
Studies by MIT and Stanford indicate that GenAI can empower customer service agents to resolve 14% more issues per hour. Additionally, GenAI has driven productivity improvements of 34% for new and less-experienced employees. In Australia, every sector and occupation sees potential productivity gains from increased automation, with education, professional services, and financial services expected to benefit the most by 2030.1
GenAI and AI are also reshaping software development and functionality. Software engineers can achieve significantly higher productivity with GenAI assistance.
Research by McKinsey suggests that with GenAI, developers can document code functionality for maintainability in half the time, write new code in nearly half the time, and optimise existing code (code refactoring) in nearly two-thirds the time.
Businesses are also integrating AI and machine learning (ML) into their software, enabling data-driven decisions based on real-time customer insights and use cases, rather than static rules.
Preparing IT infrastructure for the AI era
As organisations embrace GenAI and software with embedded AI and ML, they need to ensure their IT infrastructure possesses the necessary power and flexibility. This presents an opportunity for collaboration with trusted partners to upgrade and modernise data centres.
It should be recognised that no single entity possesses all the hardware and software required for successful AI and GenAI endeavours. This necessitates establishing highly integrated processes throughout the product lifecycle and in business operations.
Companies and key partners must also ensure compliance with all relevant regulations and enforce best practices across the entire supply chain. This includes adhering to proper processes and implementing checks and balances on materials and manufacturing processes.
It also requires adherence to software design best practices and ensuring efficient transportation and delivery of solutions. Alignment and tight integration are crucial, particularly with GenAI, which requires substantial compute and storage resources and can lead to uncontrolled compute costs, energy consumption, and carbon emissions if left unchecked.
Optimising for efficiency and sustainability
A recent report highlights that a single version of Nvidia's Blackwell chip for data centres consumes a staggering 1,200 watts of electricity. This is significantly higher than just a few years ago.
GenAI, along with the AI foundries and factories that support GenAI applications, rely heavily on compute power, interconnect networks, and storage for massive datasets.
This necessitates optimisation similar to that of FedEx's delivery approach. The company is constantly optimising routes and implementing measures to ensure timely delivery while minimising fuel consumption, costs, and carbon footprints.
However, optimisation goes beyond just efficiency. Sustainability is a growing concern, and businesses must find ways to leverage AI and GenAI while minimising their environmental impact.
This could involve exploring alternative cooling solutions for data centres, utilising renewable energy sources, or employing energy-efficient hardware and software configurations.
Building the right infrastructure for GenAI workloads
While GenAI is a new workload (or suite of workloads), it's crucial to understand that it presents unique challenges compared to traditional workloads. Organisations are still learning how to best tune their infrastructure around these new workloads.
Building the right infrastructure - a combination of cloud and on-premises systems - requires extensive analysis and expertise. Collaboration with innovative partners with a proven track record of deploying and managing mission-critical infrastructure will enable organisations to maximise the value of GenAI.
Embrace the iterative process
Optimising GenAI is an iterative process, and there are no quick fixes. Embrace solutions that streamline infrastructure and automation and seek partners with capabilities extending from data preparation (including data cleansing and obfuscation) to scalable, flexible, and cost-effective data storage, AI model training, and inference. Partnering with specialists with experience in your specific industry vertical and data-centric workflows is crucial for success.
AI and GenAI are fundamentally data-driven. It's paramount to have the most relevant and complete data, along with the appropriate data infrastructure to locate, secure, and safeguard it. This includes robust data governance practices to ensure data quality, security, and compliance with relevant regulations.
The future of AI factories
The concept of AI factories is still in its early stages, but it holds immense potential to transform how businesses operate and create value. By embracing GenAI, optimising their infrastructure, and prioritising collaboration and sustainability, organisations can position themselves to become leaders in this emerging landscape.
As AI factories and foundries evolve, they will likely play an increasingly crucial role in driving innovation across various industries, shaping the future of business and technology.
