At this year’s Data Center World in Washington, DC, a wide array of critical issues took center stage, from high-performance, cost-effective data center cooling to key sustainability challenges.
On the Data Center Knowledge news desk, analysts and experts examined hardware limitations, emerging opportunities, and major trends. A significant focus was on AI’s dual role: driving high energy demand while enabling data centers to optimize performance and efficiency.
The discussions also covered pressing trends such as the persistent lack of diversity, the rise of new markets in Latin America and Asia-Pacific, and the potential applications of large language models (LLMs) on smartphones. Experts emphasized AI’s explosive potential as a driving force behind these developments, with one researcher declaring: “There’s no question that this is the next industrial revolution.”
Watch each interview below to catch up on the latest data center industry trends:
Liquid Cooling: Opportunities and Challenges
The biggest challenge facing data centers is the heavy demands of generative AI, according to Steve McDowell, chief analyst at NAND Research. Citing Nvidia’s high-performance Blackwell GPUs as an example, McDowell laid out how new technology requires more power and runs hotter than traditional processors. “I can only put a few of those in a rack before power and heat dissipation is a big problem,” he said.
The answer, the analyst said, is liquid cooling. But the lack of standardization in performance, quality, and costs of liquid cooling injects anxiety and uncertainty into the industry, McDowell said, comparing the liquid cooling market to “VHS versus Betamax.”
Still, hardware is only half the AI equation, he added, detailing how the applications of AI can optimize efficiency. “We talk a lot about generative AI right now, but really there’s more traditional AI techniques around causal AI, around predictive AI and predictive analytics. Those are the technologies that are being used to build intelligence into data centers that then helps to control the sustainability piece.”
Will Generative AI Be as Transformative as the Printing Press?
“There’s no question that this is the next industrial revolution,” said an optimistic Vlad Galabov, research director at Omdia, about AI’s ability to significantly increase productivity. One fascinating use case is Intel, TSMC, and Nvidia’s plan to run a semiconductor within a digital twin. Calling the semiconductor industry a trial-and-error business, Galabov outlined how generative AI can allow these companies to test potential disruptions and challenges when creating a factory, and then use predictive AI as they analyze their findings. “I think it’s so interesting… using multiple AI techniques at the same time.”
Speaking to the challenge of power capacity, Galabov outlined the evolution of data center servers away from general-purpose design and towards being application-specific. Using YouTube as an example, Galabov described how building processors for video encoding into their servers increased efficiency. “The dramatic change that will happen is… you end up [replacing] several existing servers with one.” A similar consolidation is happening in IT, Galabov said, pointing to the scaling of CPUs in Intel cores.
Diversity is a Central Challenge in the ‘Human Supply Chain’
Have data centers become more sophisticated in recruiting diverse candidates? “Quite frankly, no,” said principal and CTO of StrategITcom Carrie Goetz, describing a failure to reach new talent as a “very large self-inflicted wound.” Goetz, who published a best-selling book on diversity in data centers, called for new strategies to attract candidates, as well as more apprenticeships and programs to help people who are new to the industry. “We have 300,000 open jobs that need to be filled by next year, and you ask these companies, and most of them will tell you that they’re having problems hiring talent… HR is doing exactly the same thing they’ve always done.”
Making the case for a more inclusive industry, Goetz called data centers welcoming to every trade, as well as a place for collaborative learning. “Curriculum can’t keep up with the industry. We have to learn from each other.” The ideal candidates for data centers, Goetz said, are lifelong learners from all kinds of backgrounds and perspectives. “There’s a place for everybody – we just need to make sure everybody knows they have a place.”
Emerging Markets Try to Keep Pace with AI’s Surge
As AI drives increased power demand globally, new markets are emerging to meet those needs, said Perkins Liu, senior research analyst on the Data Center Services and Infrastructure team of 451 Research, pointing in particular to markets in Vietnam, the Philippines, and Europe. Liu’s research team has tracked the “hockey stick” of demand over the last 12-18 months.
The conversation around how data centers can keep up with this growth has changed, said Liu, describing the emphasis on liquid cooling. “Last year people were still comparing the pros and cons between air cooling and liquid cooling, but this year it’s more about how and when [with] liquid cooling.” Liu also noted a more prominent effort to look at alternative energy resources to help solve the energy demands. “Sustainability is not a siloed effort by itself, it will be coupled with other improvements.”
Sustainability Is a Global Data Center Concern
“Sustainability is a topic that almost bubbles underneath without being said,” noted Omdia chief analyst Roy Illsley, describing the most pressing questions on the minds of data center professionals. “They know we’ve got to change… and it’s [a question of] how they can do it differently and understanding the impact of that change.” What is the nature of that change? In step with several other analysts, Illsley pointed to the overwhelming interest in liquid cooling, as well as the ways data center operators are making use of AI systems to improve efficiency and predict when a failure may be likely.
Another trend on Illsley’s radar is the importance of sovereign cloud, particularly in Europe and countries in the Middle East and Africa. Illsley said this was a response to the dominance of US and Chinese cloud providers. “The European market is very focused on the sovereignty aspects of how they can comply with the privacy laws and the different country laws.”
In Latin America, a Trend Toward Larger Data Centers
Asked about the emerging market in Latin America, researcher at Omdia Aaron Lewis described a large capital push to capture expected growth. “The current ecosystem mostly comprises small to midsize data centers. But if you look at the projects that are currently being built… many of them are very large data centers,” Lewis said, referencing a new construction by Cloud HQ in Sao Paolo, which sits around 1.6 million square feet. The timeframe for these projects vary, Lewis said – a large data center planned by Cloud HQ in Mexico hasn’t broken ground and it’s difficult to precisely predict when new large-scale data centers will break ground. “But what we can confirm is that this progress is happening at a very rapid pace, and it is likely to continue this way as long as demand is high,” said Lewis.
Looking Ahead: The Capability of LLM’s on Your Smartphone
“How do we get these large language models small enough to run on our smartphones?” This is a question all the major smartphone manufacturers are asking, said Michael Azoff, senior principal analyst at Omdia. That’s desirable in part because at that scale, LLMs are more cost-effective to train, and because of this technology’s many potential use cases. One example Azoff cited was the impact LLM’s might have on automobiles. “Some years ago, there was a lot of talk about autonomous driving. That’s gone a little bit quiet because they’re really struggling. But I think there’ll be a case for large language models… because they’ve shown they can do logical reasoning. And that’s something I think would be really valuable in automobiles.”
Don’t Forget About Air Cooling
Principal Analyst at Omdia Shen Wang said that without cooling, data centers won’t be able to have the AI capacity they need, and won’t be able to grow their capacity in the future. “We have to take a complete relook at AI technologies and consider what is the best solution at this stage in terms of cost [and] in terms of performance.” While liquid cooling is the fastest-growing cooling technology, Wang made the case for air cooling as well. “Air cooling is not old tech. There are still many… innovative approaches, and you can still use that for existing data centers, even the very advanced GPU-based AI data centers.” Different approaches are necessary, Wang said, for different markets: one for retrofitting existing designs – and maximizing the efficiency of air cooling devices – and one for new designs.
Cloud and Data Center Advancements in Asia-Pacific
While there have been major expansions in the data center market in North America and Western Europe, South Asia and India are rapidly growing markets, according to Omdia analyst Siraj Aziz. Beyond the top-tier market of Singapore, Aziz highlighted Malaysia, Indonesia, and the Philippines as areas with new hyperscalers and colocation providers. “In terms of scale, Princeton Digital Group, a Singaporean digital infrastructure provider, is going all out in Johor.” What makes the Malaysian market desirable, Aziz said, is the relatively low cost, reliability, and government support for new data centers. “We are seeing these three factors – power, reliability, land, and government support to be the dominant factors that actually drive data center development.”