Data Center Cooling Pivot: Magnetic Levitation & Modularity Overtake AI Growth as Power Scarcity Deepens
Introduction
The data center industry is facing a fundamental shift in its cooling strategy as power constraints and edge deployment demands force operators to abandon traditional air cooling in favor of advanced thermal management solutions. Recent weeks have witnessed a dramatic pivot toward liquid cooling and emerging magnetic levitation technologies, with established players like ABB and Clivet unveiling innovations that promise radical efficiency gains while modular designs enable rapid, localized deployment. 
According to the latest Data Centre Magazine report, ABB and NOVENCO have paired their IE5 motor and drive with the ZerAx fan, delivering up to 85% system efficiency and 15-40% energy savings in data center cooling applications. This represents a significant step beyond incremental improvements, targeting the fundamental thermodynamics of heat dissipation that are becoming the bottleneck for AI and edge workloads. 
Meanwhile, Clivet's debut of next-generation magnetic levitation compressors at the AI Infra Summit 2026 signals a technological leap that could render traditional vapor-compression systems obsolete in high-density environments. These developments coincide with Microsoft's recent #2 ranking in Data Centre Magazine's Top 250 Data Centre Companies, highlighting how liquid cooling and emerging thermal technologies are becoming the differentiator in competitive data center rankings.
The urgency driving this pivot is clear: power constraints are forcing operators to rethink how they cool AI accelerators and edge nodes. Traditional air cooling is no longer viable for densities that exceed 50kW per rack, while liquid cooling and magnetic levitation represent the twin pillars of the next-generation thermal management strategy.
In Indonesia's rapidly expanding edge computing market, these cooling innovations are particularly relevant. As local telecom operators deploy modular micro facilities to serve autonomous logistics and telemedicine applications, the ability to rapidly scale thermal management without extensive power infrastructure becomes critical.
Power Constraints Force Cooling Overhaul
Recent power grid constraints in key data center markets are driving operators to prioritize efficiency above all other considerations. The energy savings promised by advanced cooling systems are no longer optional features but essential survival tools for data center economics.
According to Ankush Gulati, Energy Efficiency Program Lead at ABB's Motion Services Division, data center energy demand is going to double by 2030, and in AI-only infrastructure, it's on track to triple. "We believe that the fastest available capacity gain is already inside the facility, hiding in plain sight." Despite cooling making up to 40% of a data center's total electricity consumption, fan systems remain fairly low on the efficiency agenda — the very gap that drove the ABB-NOVENCO collaboration, moving from better components to system-level efficiency where gains are real and accountability is shared.
GXO's new whitepaper reveals how global supply chain orchestration and predictive intelligence are accelerating activation and sustaining hyperscale performance. The report argues that the data center sector has entered a new phase of complexity driven by AI and machine learning, with deployment timelines compressed over the past two to three years. Leading hyperscalers are expected to invest approximately $650 billion in capital expenditure in 2026, and securing a return on that investment depends on how quickly infrastructure can be brought into service and sustained once live. The supply chain is now part of the critical infrastructure that determines speed to market and long-term performance: a delayed server rack, a customs issue, or a mismatch between construction readiness and hardware arrival can translate into lost revenue. GXO's report makes the case for transforming logistics into a strategic layer of data center governance, examining what separates providers that simply move equipment from genuine logistics partners capable of managing risk, sequencing critical hardware and supporting long-term operational resilience.
Vertiv's acquisition of King Environmental Services expands thermal and fluid management across EMEA data centers, underscoring how modular thermal solutions are becoming essential for operators who need flexibility to respond to changing demand patterns. Sensata's sensing and protection components are helping engineers manage cooling, power, and on-site resilience as these new cooling technologies become standard.
Magnetic Levitation: The Cooling Technology That Could Replace Vapor Compression
Clivet's magnetic levitation compressor technology represents a quantum leap in thermal management. Under the "DUO POWER BOOST" theme, Clivet made the global debut of its next-generation magnetic levitation compressor at the AI Infra Summit 2026 in Santa Clara, California. The company's portfolio also comprised magnetic cooling sources, liquid cooling distribution, and supporting products for data centers, alongside a joint power-and-cooling coordination solution with CLOU.
The portfolio's core is Clivet's next-generation magnetic levitation compressor, built on the DUO POWER BOOST dual-rotor variable-speed architecture. In a two-stage compression structure, two rotors and their impellers can adjust speed independently based on operating conditions, adapting to different flow rates and pressure ratios. According to Clivet, this design brings data center load variations, water temperature conditions, and system operating strategies into cooling source design, supporting optimization under complex conditions.
The portfolio's capabilities span facility-side cooling through air-cooled magnetic levitation chillers, and high-capacity CDUs (cooling distribution units) handling secondary-side heat exchange and distribution with capacity ranging up to 3 MW (at 3K approach temperature difference, ATD) or 5 MW (at 5K ATD). Predictive temperature control keeps fluctuations within ±0.5°C, while key components support online replacement, helping ensure continuous operation in high-availability environments.
From a portfolio covering facility-side, liquid-cooling, and outdoor heat rejection stages, through the power-and-cooling coordination showcase, to the launch of the next-generation magnetic levitation compressor, Clivet's appearance at the event presented its technology and product layout for data center thermal management. The magnetic levitation design eliminates mechanical contact between moving components, dramatically reducing wear, maintenance, and energy losses compared to traditional vapor-compression systems. The system operates with efficiency gains that make it particularly attractive for high-density AI workloads where every percentage point of efficiency translates to millions in operational savings.
In a two-stage compression structure, two rotors and their impellers can adjust speed independently based on operating conditions, adapting to different flow rates and pressure ratios. This design brings data center load variations, water temperature conditions, and system operating strategies into cooling source design, supporting optimization under complex conditions.
Modularity: The Edge Advantage
While hyperscale providers like Microsoft and Google continue to build massive centralized facilities, the real story is in modular deployment. Modular designs enable operators to deploy thermal management capacity exactly where it's needed, without the commitment to massive power infrastructure.
Vertiv CEO Gio Albertazzi discussed the acquisition of King Environmental Services, expanding thermal and fluid management across EMEA data centers. This strategic move underscores how modular thermal solutions are becoming essential for operators who need flexibility to respond to changing demand patterns.
In Southeast Asia, where power constraints are particularly acute, local telecom operators are deploying modular edge facilities that combine liquid cooling with magnetic levitation technologies. These deployments enable operators to provide AI inference services without requiring extensive power grid upgrades, addressing one of the key barriers to edge computing adoption. The ability to rapidly scale thermal management capacity without extensive power infrastructure becomes critical for edge computing in Indonesia's archipelago geography — thousands of islands, dispersed population centers make centralized processing not just slow but impractical for local services.
Edge deployments at submarine cable landing points in Batam and Palembang are already reducing round-trip time for cross-border traffic by 40-60 milliseconds, a measurable improvement for financial services and content delivery. Modular construction also matters — deploying edge facilities in weeks rather than quarters shortens the window where power procurement and grid interconnection make or break a project. Each site is becoming a node of a larger distributed system, and power resilience is the constraint that decides which nodes stay online when the grid hiccups.
Liquid Cooling: The Transition in Progress
Liquid cooling has moved from experimental to mainstream, with operators increasingly recognizing that air cooling cannot meet the demands of modern workloads. As rack densities climb because new AI chips draw 700 watts or more, air can't carry heat away fast enough. Direct-to-chip loops and immersion systems are being specified in new edge builds because air cooling cannot handle the power density of modern AI accelerators at distributed sites.
In Batam and Bintan, where new industrial parks are rising, operators are designing for 50kW per rack from day one. That requires liquid loops, not raised floors and CRAC units. The power constraint is structural. PLN's grid expansion in secondary cities lags demand, and on-site generation adds cost and emissions. Some operators are evaluating small modular reactors for baseload, but those are years out. In the meantime, battery storage and grid-interactive UPS systems buy flexibility at the edge. The sites that secure firm power contracts now will operate when competitors are still waiting for transformers.
Liquid cooling is no longer experimental. Direct-to-chip and immersion systems are being specified in new edge builds because air cooling cannot handle the power density of modern AI accelerators at distributed sites. The economics are clear: while traditional air cooling systems require substantial power for fans and cooling towers, liquid cooling systems can achieve the same heat removal with 50-70% less energy consumption. For operators facing power constraints, this efficiency translates directly to improved margins and competitive advantage.
Investment Patterns Reflect the Cooling Pivot
KKR's Global Head of Real Assets, Raj Agrawal, revealed Samsung's US$1 billion Helix investment targets AI data centers, power and connectivity. This significant investment in thermal management infrastructure signals how major technology companies are prioritizing cooling solutions as a critical component of their AI infrastructure strategy.
Alibaba's Eddie Wu unveiled the Zhenwu V900 at its annual Apsara conference, with shares surging 5% to a monthly high in Hong Kong. The company's focus on advanced cooling solutions highlights how Chinese tech giants are investing heavily in thermal management to support their AI ambitions. The Zhenwu V900 represents Alibaba's next-generation approach to data center infrastructure, optimized for the specific thermal and power demands of AI workloads.
As AI drives rack densities higher, Sensata's sensing and protection components are helping engineers manage cooling, power, and on-site resilience. These components become essential enablers for the new cooling architectures, providing the monitoring and protection capabilities that ensure reliable operation under demanding conditions.
The New Cooling Reality
The data center industry's pivot toward liquid cooling and magnetic levitation technologies represents more than a technical shift; it's a fundamental rethinking of how thermal management works. The old paradigm of air cooling massive centralized facilities is giving way to a new reality of modular, liquid-cooled, and magnetic levitation-based thermal management systems.
For operators, the implications are clear. Those who can adopt these technologies early will gain significant competitive advantages in efficiency, flexibility, and operational costs. The companies leading this transition include established players like ABB, Clivet, and Vertiv, as well as technology giants like Microsoft, Samsung, and Alibaba.
More investment is coming, and the bottleneck won't be capital. It will be power, cooling, and the security frameworks that keep distributed nodes from becoming weak points. Operators that solve all three first will set the standard for the region. For context on data center expansion, see Reuters on data center growth. The race is open, and the finish line is closer than most boards realize. The question is not whether edge computing will expand; it is which operators secure the contracts first, and on what timeline. The gap between pilot projects and production fleets remains the next real test of execution.
Looking Forward
As power constraints continue to tighten and edge computing demands increase, the cooling technology race will only accelerate. Magnetic levitation, liquid cooling, and modular designs will become the standard rather than the exception. Operators who can navigate this transition successfully will position themselves as leaders in the next wave of computing infrastructure.
The question is not whether to adopt these technologies, but how quickly operators can implement them. As Sensata's components become essential for managing these new cooling systems, the entire data center industry is being reshaped by thermal innovation. The enterprises that will thrive in this environment are the ones treating edge as infrastructure, not a novelty. They are building repeatable deployment playbooks, securing firm power and cooling contracts upfront, and designing security into the node, not bolted on later. The finish line isn't a milestone date. It's the point where every latency-sensitive workload has a logical place to run, and the bill for moving data long distances becomes a line item, not a given. Operators that ignore this reality will lose customers to competitors who have already moved. For more on the infrastructure side, see Reuters on data center growth.
References
-
Data Centre Magazine: Top 250 Data Centre Companies 2026
-
ABB and NOVENCO: IE5 motor and drive with ZerAx fan
-
Clivet: Magnetic levitation compressor at AI Infra Summit 2026
-
GXO: Orchestrating Global Data Centre Supply Chains at Hyperscale
-
Vertiv: Acquisition of King Environmental Services
-
Sensata: Components for cooling, power, and on-site resilience management
-
Abu Dhabi: Zhenwu V900 at Apsara conference
You can read more about data center trends in the Cloud & Edge Computing category for our ongoing series on distributed infrastructure.