Data centers have undergone many changes over the past year or so. The main reason is the rise of Generative AI (GenAI). The massive rise in use and demand for AI services is fueling data center developments around the world which is bringing us a lot of innovations. This is what we see thanks to the State of the Data Center Report by AFCOM.
This is the eighth annual edition of the report, DataCenterKnowledge details. This report gives us a good idea of what’s going on in the data center industry and it’s quite the picture. According to the report, currently the industry is in the midst of a “massive boom in infrastructure design and construction,” while also experiencing several problems in key areas. Among them are energy, cooling and sustainability.
“This year, we delve into the intricacies of GPUs, the explosive growth of the data center sector, and, particularly, the adoption of AI solutions. A significant majority of respondents are gearing up to integrate AI-capable solutions into their data centers. The primary driver? An overwhelming interest in supporting the newest wave of generative AI applications,” said Bill Kleyman, program chair for AFCOM’s Data Center World event and co-author of the report.
The Big boom
The 2024 State of the Data Center Report says that hyperscalers like Google, Meta and Amazon are currently making big changes to their facilities. The main goal – they are adapting to the changes that AI demands. One main one is high rack load and more powerful hardware to accommodate the AI workloads.
For example, 53% of respondents say that new AI workloads will “definitely increase capacity requirements,” including for colocation. 56% of respondents say they are planning to deploy AI solutions within their own data centers. In most cases, 43% of these AI solutions will be GenAI.
AFCOM also adds that currently, there are almost 10 000 colocation and wholesale data centers in operation around the world. There’s still a lot more room to grow as AI will drive a massive jump in demand for capacity. And while some of that can be covered by expanding current facilities and increasing densities, there will be a big need for entirely new buildings, too.
“We expect new data center space will need to be added for years to come. Already, the hyperscale cloud and colocation service providers are in a race to build newer and more facilities globally. New data centers will need to be built from the ground up (rather than retrofitted) to be AI-ready, as the infrastructure has unique cooling and power requirements,” Dell’Oro Group analyst Baron Fung, who told Data Center Knowledge.
This is further supported by a study by Gartner. They predict that the overall spending on IT will reach $5 trillion in 2024. This would mean an increase of 6.8% compared to 2023. From that money, about $261 million will be spent on data centers. This would result in a 7.5% growth.
“AI will continue to fuel data center growth in the European data center industry. We also see it being a catalyst to overcoming the challenges (e.g. economy, supply chain, availability of land and power, local regulations) it presents by enabling innovation,” Oliver Menzel, CEO of Maincubes, a European data center operator, told Data Center Knowledge.
Generate this
GenAI specifically has had quite an effect on data centers. DataCentreMagazine has checked with industry experts and operators to see exactly what GenAI has already done and what they expect to happen.
First, the positives. According to Sascha Giese, Tech Evangelist at SolarWinds, GenAI will be very beneficial when it comes to simplifying the process of interacting with infrastructure. For example, helping engineers handle more tasks even when they might include programming languages they don’t know. “That’s especially true when it’s tailored for voice commands. GenAI models can be trained to ‘translate’ spoken statements and enter commands into the command-line interface of devices or orchestrators. This is particularly effective in vendor-agnostic scenarios,” he says.
According to Karsten Winther, EMEA president at Vertiv, the transformative power of AI will be of benefit for predictive maintenance of data centers, optimizing performance and reducing downtime. It will also help data center operators to reduce the environmental impact of their facilities and find new sustainability solutions. This way AI can help solve the problem it will create – the jump in hardware demand will result in increased energy consumption, thus increased carbon footprint. AI could help optimize and improve the entire process of the data center operation and thus help lower the ecological impact.
“The integration of AI in smart energy systems is particularly interesting. AI algorithms are able to analyse real-time data, enhancing grid management, demand response and efficient resource use. By continuously monitoring energy demand and supply, AI can enhance the distribution of renewable energy, maximising its effective use. Moreover, AI algorithms can predict changes in energy patterns, enabling proactive measures to maintain grid stability and reliability,” says Winther.
Yes, but…
Of course, not everything is roses. “Given its critical role, training GenAI in a data centre environment is a huge challenge. The margin for error is very slim, with any hiccup having huge knock-on effects for the organisation — including potential blackouts. As such, training AI models involves thorough planning and careful deployment over an extended period of time,” Giese says.
Also, data centers can become victims of the very technology they are supposed to carry. Cyber criminals could use AI to infiltrate data centers and extract data. GenAI would without a doubt be used by hackers in this regard. Some more experienced and better funded hackers could develop or even modify their own GenAI modules for the specific use of cyberattacks. There are already some projects discovered in this area on the Web.
So, how to respond? With more AI, of course. GenAI could help and become a tool for improving digital defenses. AI systems will also bring ethical risks. Among them are malicious usage of GenAI by users, bad actors or simply privacy and economical concerns. Data center operators should be mindful of all of this and be prepared to react, with technical measures or even PR statements. AI will demand changes in all areas and will force data center operators to be ready for everything as will all of us.
Innovations are here
The AI boom is already showcasing multiple areas data centers need to improve. Including power and cooling. And while AI can help, most of the innovations in this area are already happening without it.
“Our sector’s expansion and critical role, along with its substantial energy needs, have necessitated a heightened emphasis on sustainable practices and the exploration of renewable energy sources. Three in four respondents (73%) plan to utilize renewable energy, including 27% who are currently doing so,” states the AFCOM State of the Data Center report.
Additionally, 63% are planning to decrease their water usage. 18% of them are already working with active solutions, while 27% are looking for solutions. For 8% this is in their 12-month plan. 9% expect to do this within three years.
Meanwhile, data center construction is also ramping up. AFCOM notes that 42% of respondents are interested in a hybrid construction approach. This means using prefabricated modules along with traditional building techniques. “Diving deeper into prefabricated data center designs, we were curious to learn which part would be prefabricated and delivered. Respondents planning to leverage prefabrication in new data center builds are most likely to report exploring prefabricated power modules (78%) and prefabricated cooling modules (72%),” the report said.
A separate assessment by Omdia from November 2023 states that prefabricated data centers can shorten build time by 40% to 80%. Another benefit is that it will reduce the carbon footprint for the construction by two-thirds. Of course, this solution is not applicable everywhere as some data centers demand specific solutions based on their location and goals. But for the most part, prefabricated modules can be used to help shorten construction times and slash costs.
Then there’s the topic of energy. “The development of AI algorithms specifically designed for cognitive infrastructure will further enhance the capabilities of smart energy systems. These algorithms will enable data centres and other industries to make intelligent decisions in real-time, optimising energy usage, predicting equipment failures and proactively managing power consumption. By leveraging AI’s ability to analyse vast amounts of data and learn from patterns and trends, data centres can continuously improve their energy efficiency and reduce their environmental impact,” says Winther.
AI is here to stay
“With forecasts indicating that applications based on Large Language Models will be as ubiquitous as web applications within the next half-decade, the industry is at a pivotal juncture. Yet, the narrative doesn’t end with AI. Our comprehensive report goes beyond unraveling the complexities underpinning these emerging trends. We are at the cusp of crafting a digital era like no other. In this journey, challenges are inevitable, and barriers are meant to be broken. In this rapidly evolving landscape, there is very little tolerance for complacency. Embrace experimentation with new technologies; they could revolutionize your business and user experience,” Kleyman says.
GenAI is just getting started. In 2023 it showed us how quickly it can evolve and it’s still in its infancy. The advancement of AI in general will accelerate more and more as it gets better and better. So, we can expect it to be significantly more capable by the end of the year in all areas.
As such, data centers have no other choice but to continue to improve and accelerate the rate of their advancements. This is the only way they can answer the ever-increasing demands of AI and the expectation of data center clients, mainly enterprises, who will want to be able to access all the latest and greatest AI technologies.