What is an “AI-powered smart data center”?
Artificial intelligence is the most common topic in the tech industry in 2023. After years of development, OpenAI tools have become popular, sparking many questions about the long-term implications of this technology, as well as the immediate changes it may bring. In the data center sector, the focus is on the infrastructure required to support AI. There is a general consensus that AI will increase demand for data center space, especially from large clients or those with massive capabilities. Additionally, industry media point to the emergence of a new category of facilities—”AI smart data centers”—expected to completely reshape the market.
On a broader scale, AI appears to have fallen victim to the usual pattern that emerges when a new technological breakthrough begins: media hype. Although AI is poised to bring about major changes in the global technology landscape, we believe there is a lot of hype surrounding the specific impacts of AI on the data center industry. To understand the long-term impacts, we must first take a step back and define what an AI smart data center is—or more simply, what future data centers will look like.
The wave of hype and its impact on understanding future data centers
The interesting feature of the big hype wave, as with AI, is that it pushes us to explore what is possible, what is plausible, and what is pure fiction. When it comes to the impact of AI on data centers, the real question that arises is: “What will data centers look like in the future?” So far, AI finally seems to have achieved the high density that the data center industry has been talking about for over a decade.
Projections indicate that a new category of facilities may emerge—data centers dedicated to AI—but in the short term, most data center providers appear capable of adapting to high-density requirements using their existing designs. Over the long term, graphics processing unit (GPU) deployments are likely to continue pushing densities to higher levels, and some providers are already beginning to consider redesigning facilities to meet these requirements. However, AI smart data centers will likely have impacts beyond density, including facility cooling, networking, and data center management.
High density in smart data centers
Over the past few years, the data center sector has seen increasing demand for space, driven by cloud service providers. Since 2020, the need for cloud services has increased significantly, prompting cloud providers to expand their infrastructure to serve end users. By late 2021 and early 2022, a new wave of demand emerged, but with a notable difference: companies now require facilities that offer greater amounts of power per square foot.
Providers agree that AI workloads require high-density infrastructure at the data center level. This indicates that the new demand can be attributed to growing interest in AI. However, in the short term, most data center providers have not been forced to completely redesign their facilities. For example, some providers claim they can already accommodate deployments of up to 50 kW per rack.
Nonetheless, most providers agree that redesigns will likely be necessary over the next five years as rack densities are expected to continue rising with the emergence of more power-hungry systems.
Cooling challenges: A pressing issue
One of the main issues associated with high-density data centers is the excess heat generated by new generations of high-performance servers. As rack densities rise, adopting new cooling methods becomes essential. Currently, air cooling systems remain the most common, but liquid cooling is starting to turn heads.
According to a recent 451 Research survey titled Voice of the Enterprise: Datacenters, Datacenter Infrastructure 2023, 43% of IT decision-makers stated they use air cooling and plan to continue doing so over the next five years. Meanwhile, 51% said they use air cooling systems but plan to transition to liquid cooling during the same period. Among them, 31% plan to do so within 12 months, 17% within two to four years, and 4% within five years.
As for those who have already started using liquid cooling, the most popular options include immersion cooling (34%), direct-to-chip cooling (19%), and a combination of both (47%).
Although immersion cooling requires radical changes to existing facility designs, providers appear to lean toward other options that can be implemented more quickly.
Networking: The foundation for running AI
AI relies on high traffic between components. The work requires parallel processing, where GPUs, CPUs, and memory work independently, and work must wait for the slowest element. For this reason, networking is a major bottleneck for AI performance.
According to the Voice of the Enterprise: Artificial Intelligence and Machine Learning, Infrastructure 2023 survey, 45% of respondents indicated they need high-performance networking to optimize AI and machine learning workload performance. This outranked any other infrastructure resource, including cloud accelerators (37%) and faster servers (36%).
Data center management using artificial intelligence
As demand for higher-density racks increases and cooling methods evolve, managing data center space effectively becomes increasingly important. AI can be used as part of the solution to the “problem” it causes, as it is already being used to identify infrastructure issues within data centers.
Data Center Infrastructure Management (DCIM) systems are used by companies like Google and Huawei to control cooling and identify hot and cold spots within facilities. This helps maintain a uniform temperature and ensures cooling units operate efficiently.
Future benefits and challenges
Future benefits of AI in data centers
- Improving energy efficiency:
Energy consumption is one of the major challenges facing data centers. By using AI, energy usage can be monitored and analyzed more accurately. Smart systems can automatically adjust power consumption based on current loads, reducing waste and achieving higher efficiency in daily operations. - Achieving environmental sustainability goals:
In line with the global push toward sustainability, AI can contribute to reducing data centers’ carbon footprints. Smart systems are capable of identifying areas where efficiency can be improved, reducing reliance on traditional energy sources while supporting the use of renewable energy sources such as solar and wind power. - Process automation:
AI can independently perform many routine tasks, such as equipment monitoring, failure prediction, and adjusting the working environment within data centers. This reduces the need for human intervention and ensures uninterrupted operations. - Reducing operating costs:
Thanks to AI’s capabilities to optimize resource utilization and reduce the need for emergency maintenance, providers can significantly cut operating costs. For example, a smart management system can detect issues early and prevent them from developing into major failures, saving time and money. - Enhancing operational efficiency:
AI helps manage data flows effectively and improves network performance, leading to better response times and increased productivity. It also enables the prediction of behavioral patterns in hardware and infrastructure, allowing for better resource management.
Future challenges of AI in data centers
- Skilled labor shortage:
While AI offers innovative solutions, operating these systems requires advanced technical skills. The sector faces a shortage of qualified personnel to develop and manage smart technologies, making it essential to invest in training and skill development. - High initial cost:
Adopting AI systems requires substantial upfront investments to develop infrastructure and purchase advanced equipment. This may present a barrier for small and medium-sized enterprises that might struggle to afford these costs. - Cybersecurity challenges:
With increased reliance on AI, the need to secure these systems against cyberattacks grows. Smart systems, despite their effectiveness, can be targets for hackers seeking access to sensitive data. - Compatibility with legacy systems:
Many data centers still use traditional systems, making integration between old and new technologies a major challenge. Strategies must be designed to ensure compatibility and maximize returns on previous investments. - Risks associated with over-automation:
Despite the benefits of automation, complete reliance on smart systems can leave gaps in the event of unexpected failures. Backup plans must be in place to handle emergencies and ensure business continuity. - Ethical and regulatory challenges:
As AI evolves, questions arise regarding privacy, data management, and ensuring the technology is used responsibly. This requires establishing clear regulations and policies to prevent the misuse of these technologies.
The future: Combining benefits and addressing challenges
To address these challenges and maximize benefits, data center providers must:
- Invest in workforce training: Develop the skills of technical teams and prepare them to handle smart technologies.
- Adopt flexible strategies: Enabling adaptation to rapid technological changes.
- Enhance cooperation with government entities: To establish a regulatory framework that encourages innovation while protecting data and ensuring privacy.
- Focus on sustainable innovation: Striking a balance between technological advancement and environmental preservation.
Despite the challenges, adopting AI in data centers presents enormous opportunities to improve operations, reduce costs, and support sustainability. The key lies in adopting a strategic and balanced approach that enhances data centers’ ability to meet future demands.
