Artificial intelligence (AI) has been rapidly integrated into diverse industries such as finance, cybersecurity, healthcare, transportation, and manufacturing. AI and analytics are boosting productivity, delivering innovative products and services, accentuating corporate values, addressing supply chain challenges, and fueling innovation.
However, a challenge is arising: The growing reliance on AI computation demands greater use of data center resources. As a CIO, I am excited by the possibilities generative AI brings, but I’m also concerned about the impact this will have on my company and others who rely on data centers and cloud infrastructure, such as cost increases, sustainability issues, and compliance concerns.
Impact on Costs
The surge in generative AI development is placing significant strain on data center capacity, resulting in higher costs for customers. Tirias Research forecasts that generative AI data center server infrastructure plus operating costs will exceed $76 billion by 2028.
Cloud service providers are not immune to the increased demand for data center resources. They are also having to expand their infrastructure to meet the growing demand for AI computation. The basic principles of supply and demand, combined with rising interest rates, have inevitably caused prices to increase. Those prices will then be passed along to customers.
Sustainability Concerns
Generative AI models, especially advanced ones, place a significant energy demand on data centers. These models are intricate and require a lot of computational power, particularly during their training phase with vast amounts of data. Specialized computer components, like graphics processing units (GPUs), are designed to handle these heavy tasks, but they still consume a lot of energy when they’re constantly in use.
Additionally, the intense work these models undertake generates a lot of heat. To prevent overheating, data centers use cooling systems, such as air conditioners and fans. These systems, while necessary, also consume energy. So, the combination of the AI’s computational needs and the cooling requirements leads to substantial energy usage.
Microsoft is experimenting with a novel approach: powering its data centers with small nuclear reactors. These reactors are safe and reliable, and they do not produce greenhouse gases. They are a green energy innovation for data centers. A further obstacle involves managing the heat generated by AI workloads in data centers. Approaches such as air-assisted and direct liquid cooling offer more efficient cooling compared to conventional methods. Microsoft is exploring the concept of submerging data centers underwater to utilize natural cooling from the water.
Navigating Policy and Compliance
The rapid growth of generative AI has led to an increase in the amount and sensitivity of data being used, posing challenges for CIOs. These AI models need large datasets, often containing sensitive details. Any mistakes in handling this data can result in privacy breaches and legal issues. As AI becomes more common, governments around the world are setting stricter data protection rules. Some even have specific limits on AI research, especially if they see it as a potential risk to their country’s safety or well-being.
This complex set of rules means CIOs have to be very careful about compliance. They might also need to think about where to place their data centers based on different country’s rules about AI. For instance, some countries mandate that customer data on their country’s citizens may not be stored in data centers outside their country. While navigating these rules, CIOs also have to ensure that they’re getting the most out of their AI projects and not letting strict data rules hold back innovation.
Preparing for the Future
Generative AI models, which require a lot of computing power, are changing how data centers operate. CIOs need to check if their current setup, from the main processing units to storage, can handle these powerful AI demands. But getting ready for the AI age isn’t just about adding more equipment; it’s about making sure everything works well together. This means thinking about power use, cooling systems, and having backup plans for any system issues. As AI keeps advancing, data centers need to be flexible and ready to adapt, which could entail designs that are easy to update or combining local and cloud-based resources, and CIOs need to be prepared for what’s coming next, ensuring their setups are both strong and adaptable.
