The 2-Z of 2025 predictions: D is also for data centres

Concept artwork on sustainable data
centres generated by Dream
by WOMBO.

D is also for data centres 

"The year 2025 will be a pivotal one for the data centre industry. The rapid rise of AI-driven workloads, coupled with stricter sustainability mandates, is driving a fundamental shift in how data centres are designed and operated. AI-supported innovations like ChatGPT now require four to five times more processing power than traditional Internet searches, while ASEAN data centre demand is projected to grow at 20% annually over the next five to seven years, according to Maybank," said Jason Low, APAC Director at Iceotope.

The rise in AI agents will ultimately lead to a significant increase in the amount of data that is created and utilised, with implications for data centres, said Terry Maiolo, VP & GM, Asia Pacific, OVHcloud.

"With more data generated than ever before and new data sources
emerging, increasing amounts of data will become scattered across disparate systems, limiting its accessibility and utility – unless governed by robust infrastructure and a trusted cloud. Moreover, the sheer volume of data will place increasing demands on data centres, necessitating scalable storage and processing capabilities," he said.

Tay Bee Kheng, President, ASEAN, Cisco, said the reality check on generative AI has been harsh, partly because of shortfalls with data centres. "It’s been more than a year since AI became the dominant theme of the business world. The pressure to deploy it is relentless, with nearly all companies in Cisco’s 2024 AI Readiness Index reporting a heightened urgency to implement AI solutions over the past year," she said.

"As companies embark on their AI adoption journeys, they are realising that leveraging AI is not as straightforward as anticipated. Only 19% of companies in ASEAN are fully ready to capture AI’s potential, as reality sets in on what is needed to succeed. Despite AI being a priority investment, many companies are saying that returns on these investments are falling short of expectations."

The primary challenge remains infrastructure readiness, with gaps in compute, data centre network performance, and cybersecurity, amongst other areas, Tay said. "Only 30% of companies have the necessary GPUs to meet current and future AI demands and 39% have the capabilities to protect data in AI models with end-to-end encryption, security audits, continuous monitoring, and instant threat response. As companies weigh the decision to build or buy AI solutions, modernising data centres and leveraging plug-and-play AI infrastructure that evolves with their needs without adding complexity is crucial," she elaborated. 

Compute fabrics

Source: NVIDIA. Shainer.

Gilad Shainer, Senior VP of Networking, NVIDIA, said that data centre architecture would transform into an integrated compute fabric that enables thousands of accelerators to efficiently communicate with one another via scale-up and scale-out communications, spanning miles of cabling and multiple data centre facilities.

"Scale-out communication across networks will be crucial to large-scale AI data centre deployments — and key to getting them up and running in weeks versus months or years," Shainer said. 

"As agentic AI workloads grow — requiring communication across multiple interconnected AI models working together rather than monolithic and localised AI models — compute fabrics will be essential to delivering real-time generative AI."

From data centre to AI factory

Source: NVIDIA. Buck.

"With GPUs becoming more widely adopted, industries will look to accelerate everything, from planning to production. New architectures will add to that virtuous cycle, delivering cost efficiencies and an order of magnitude higher compute performance with each generation," said Ian Buck, VP of Hyperscale and HPC, NVIDIA.

"As nations and businesses race to build AI factories to accelerate even more workloads, expect many to look for platform solutions and reference data centre architectures or blueprints that can get a data centre up and running in weeks versus months. This will help them solve some of the world’s toughest challenges, including quantum computing and drug discovery."

Charlie Boyle, VP of DGX Platforms, NVIDIA, expects AI factories to manage more data in 2025. 

Source: NVIDIA. Boyle.

"(In 2025) enterprises will expand these factories to leverage massive amounts of historical and synthetic data, then generate forecasts and simulations for everything from consumer behaviour and supply chain optimisation to financial market movements and digital twins of factories and warehouses. AI factories will become a key competitive advantage that helps early adopters anticipate and shape future scenarios, rather than just react to them," he said.

Shainer went on to predict that all data centres will become accelerated as new approaches to Ethernet design emerge. Hundreds of thousands of GPUs might support a single workload in future, he said, helping to democratise AI factory rollouts for multitenant generative AI clouds and enterprise AI data centres.

"Companies will build data centre resources that are more geographically dispersed — located hundreds or even thousands of miles apart — because of power limitations and the need to build closer to renewable energy sources. Scale-out communications will ensure reliable data movement over these long distances," Shainer said.

Average rack densities have been increasing steadily over the past few years, with predictions of AI factory racks of 500 to 1,000 kW or higher presenting an "unprecedented disruption", said Vertiv, a global provider of critical digital infrastructure and continuity solutions. 

"As a result of the rapid changes, chip developers, customers, power and cooling infrastructure manufacturers, utilities and other industry stakeholders will increasingly partner to develop and support transparent roadmaps to enable AI adoption. This collaboration extends to development tools powered by AI to speed engineering and manufacturing for standardised and customised designs. In (2025) chip makers, infrastructure designers and customers will increasingly collaborate and move toward manufacturing partnerships that enable true integration of IT and infrastructure," added Vertiv in a list of 2025 predictions.

Power, cooling and sustainability

"In 2025, we will see increased scrutiny on businesses to act in an environmentally-sustainable manner, with some markets introducing legislation imposing tighter restrictions on the disclosure of climate risks. For data centres, the challenge will be balancing the need for innovation and scalability with its carbon footprint," stated Lenovo in its list of predictions.

Scalability will tip the scales towards sustainability, Vertiv said, as the volume of compute-intense workloads intensifies. "Advanced computing will continue to shift from CPU to GPU to leverage the latter’s parallel computing power and the higher thermal design point of modern chips. This will further stress existing power and cooling systems and push data centre operators toward cold-plate and immersion cooling solutions that remove heat at the rack level. Enterprise data centres will be impacted by this trend, as AI use expands beyond early cloud and colocation providers," the company predicted.

"AI adoption is accelerating across the Asia Pacific region, with enterprises in markets like Singapore, Malaysia, and Australia leading the way in harnessing AI for transformation—unlocking new levels of efficiency, enhancing customer experiences, and solving complex challenges with agility. Our predicted trends for 2025 highlight the critical need for investment in energy-efficient and innovative digital infrastructure to unlock AI’s full potential," said Paul Churchill, VP and GM, Vertiv Asia. 

Source: Hitachi Vantara.
Hardman.

Matthew Hardman, CTO, APAC, Hitachi Vantara, said that energy demands for data centres are skyrocketing. "To tackle this, companies are turning to AI-powered digital twins—virtual models of physical infrastructure—to optimise energy usage and simulate efficiency improvements before implementation," he said. 

"This approach isn’t just theoretical; retrofitting existing data centres with these technologies is already reducing power consumption to be more energy efficient. In Southeast Asia alone, the data centre market projected to grow by over 5% annually through 2029, reaching US$14.41 B -- this shift will have a massive impact on both sustainability and cost savings."

Liquid cooling

"The rise of AI workloads and next-generation chip architectures has brought the limitations of traditional cooling methods into sharp focus. Air cooling systems, which have long been the backbone of data centre operations, are increasingly unable to handle the immense heat generated by high-performance computing environments," said Low.

Recent advancements in liquid cooling technology are paving the way for more sustainable and efficient solutions, Low added. "By leveraging forced convection for heat transfer, these systems deliver superior cooling performance, achieving thermal resistances comparable to advanced water-cooled systems without the dependency on water," he said. 

Boyle said: "As AI workloads continue to drive growth, pioneering organisations will transition to liquid cooling to maximise performance and energy efficiency. Hyperscale cloud providers and large enterprises will lead the way, using liquid cooling in new AI data centres that house hundreds of thousands of AI accelerators, networking and software." 

Boyle noted that the demand for liquid cooling will be driven by enterprises which increasingly choose to deploy AI infrastructure in colocation facilities rather than build their own, "in part to ease the financial burden of designing, deploying and operating intelligence manufacturing at scale. Or, they will rent capacity as needed". 

"These deployments will help enterprises harness the latest infrastructure without needing to install and operate it themselves. This shift will accelerate broader industry adoption of liquid cooling as a mainstream solution for AI data centres," he said.

Source: OVHcloud. Maiolo.

Transforming energy consumption has become a priority for data centres. "These facilities, already energy-intensive, consume 1-1.5% of global electricity with potentially-significant implications on energy use from AI and machine learning in upcoming years," said Maiolo.

"In fact, as the immense computational demands of AI and quantum workloads require substantial processing power and storage capacity, the resulting heat output necessitates increased energy consumption to maintain optimal operating temperatures. This will strain cooling systems, escalating energy consumption, and exacerbating overall data centre energy usage. 

"Singapore, a key player in SEA’s data centre landscape, continues to lead the charge towards sustainable data centre development. The nation has not only established standards to facilitate the construction and operation of energy-efficient data centres, but also has plans to introduce standards for IT equipment and energy-efficient water cooling by 2025. 

"Concurrently, governments worldwide are increasingly recognising the need for cleaner energy sources, with hydrogen fuel cells emerging as a promising alternative to fossil fuels."

Maiolo said businesses must prioritise green data centres – which are designed to optimise energy efficiency, scalability and environmental sustainability. "By implementing advanced cooling technologies such as water cooling, data centres can significantly reduce their water and electricity consumption, thereby minimising their environmental impact," he said. 

"The good news? We are already seeing markets like Thailand take steps, with others like Vietnam announcing plans to follow in the coming years. By investing in energy-efficient infrastructure and adopting ecofriendly practices, businesses will be able to achieve long-term success in the cloud and ensure that its digital ambitions are not curtailed by avoidable setbacks."

Vertiv highlighted emerging cooling solutions as well. "In 2024, we predicted a trend toward energy alternatives and microgrid deployments, and in 2025 we are seeing an acceleration of this trend, with real movement toward prioritising and seeking out energy-efficient solutions and energy alternatives that are new to this arena. 

"Fuel cells and alternative battery chemistries are increasingly available for microgrid energy options. Longer-term, multiple companies are developing small modular reactors for data centres and other large power consumers, with availability expected around the end of the decade. Progress on this front bears watching in 2025," Vertiv said. 

Small modular reactors were also highlighted by Dr Werner Vogels, CTO, Amazon in his 2025 predictions."The advent of generative AI, coupled with a broad push for electrification across various sectors—from transportation to manufacturing—have significantly increased energy demand. These developments have intensified the urgency for more robust and scalable energy solutions," he said.

While renewables are an up-and-coming solution, Dr Vogels noted that they "cannot meet our accelerating energy demands fast enough", and that "a constant, modular, round-the-clock complement" like nuclear power is ideal. "Nuclear has re-emerged as a promising solution. It’s not about returning to old models but embracing cutting-edge technologies such as small modular reactors (SMRs). These reactors are smaller, flexible, and easier to maintain and operate than traditional nuclear plants," he said, noting that Amazon has invested in X-Energy to develop SMRs.

"Moreover, advances in fabrication technologies, such as local-electron beam welding, which reduces the time it takes for nuclear grade welds from a year to roughly a day, have the potential to drastically reduce build times. Innovations in seismic safety, like the work being done by Japan’s Atomic Energy Agency, will make nuclear power a safer option in earthquake-prone areas. What’s even more interesting is the opportunity that SMRs present to repower existing sites already connected to the grid, such as decommissioned coal plants, providing an accelerated path to fill critical energy needs and alleviate stress on the grid."

"While liquid cooling has yet to achieve widespread adoption, the need for sustainable solutions that balance operational efficiency with environmental goals is accelerating its appeal," agreed Iceotope's Low.

Low listed some drivers for more widespread adoption of liquid cooling:

- Governments must set clear benchmarks and incentivise data centres to adopt sustainable technologies. "For instance, Singapore’s updated Green Mark certification, which will include liquid cooling standards by 2025, highlights how policy can encourage innovation," he said.

- Collaboration between technology providers, operators, and regulators will be essential to scale the technology across diverse environments.

- Industry-wide recognition of liquid cooling’s strategic value is equally important. "Beyond compliance, liquid cooling enables higher compute densities, reduces operational costs, and aligns with environmental goals—key factors as clients and investors increasingly prioritise sustainability in vendor selection," Low explained.

Vertiv's take, on the other hand, is that:

- AI racks will require UPS systems, batteries, power distribution equipment and switch gear with higher power densities to handle AI loads that can fluctuate from a 10% idle to a 150% overload in a flash.

- Hybrid cooling systems, with liquid-to-liquid, liquid-to-air and liquid-to-refrigerant configurations, will evolve in rack-mount, perimeter and row-based cabinet models that can be deployed in brown/greenfield applications.

- Liquid cooling systems will increasingly be paired with their own dedicated, high-density UPS systems to provide continuous operation.

- Servers will increasingly be integrated with the infrastructure needed to support them, including factory-integrated liquid cooling, ultimately making manufacturing and assembly more efficient, deployment faster, equipment footprint smaller, and increasing system energy efficiency.

Dr Vogels emphasised that the flip side of the energy equation has to be considered as well when energy consumption at data centres is only projected to go up over time. "The best thing that we can do is move away from legacy infrastructure and shift to hyperscale data centres, which will decrease energy consumption by nearly 25% through improved efficiency and economies of scale. 

"Beyond that, the next frontier will involve designing data centres as flexible load centres, capable of adjusting their consumption based on real-time grid needs. This will not only optimise energy usage, but also enhance the stability and resilience of the grid we all rely on," he said.

"This transformative shift will require more than technological innovation; it demands a skilled workforce, from nuclear engineers and machinists to grid management experts and material scientists. Investing in education, training, and reskilling programmes will be crucial. And it will reshape economies by creating high-paying, highly skilled jobs and driving innovation across sectors."

Regulations

Another trend in play, Vertiv said, involves government scrutiny. Expected increases in energy consumption due to AI may place demands on the grid that many utilities can’t handle, attracting regulatory attention from governments around the globe and leading to potential restrictions on data centre builds and energy use, as well as spiking costs and carbon emissions. "These pressures are forcing organisations to prioritise energy efficiency and sustainability even more than they have in the past," Vertiv stated.

The company also noted that AI regulations will initially focus on applications of the technology, "but as the focus on energy and water consumption and greenhouse gas emissions intensifies, regulations could extend to types of AI application and data centre resource consumption." 

Explore

Read the full set of 2025 predictions from 2-Z at https://www.techtradeasia.com/2025/01/the-techtrade-asia-2024-roundup-2025.html

Hashtag: #2025Predictions