From Air to Intelligence – How Technology is Transforming Data Center Sustainability

The challenge facing data centers is paradoxical: double the capacity (415 TWh to 945 TWh by 2030) while halving the environmental footprint. Meeting this demand requires more than incremental improvements. It demands innovation across the entire stack – from how we cool servers to how we manage power, from the intelligence that optimizes operations to the creative reuse of what was once waste.

The good news? A wave of technologies is emerging to tackle different pieces of this puzzle. Together, they're reshaping what's possible in data center sustainability.

This is Part 2 of our Green Data Centers series. Read Part 1 here.

The Brain: AI-Driven Data Optimization

If hardware is the body of a data center, artificial intelligence is becoming its brain. Data centers generate massive streams of real-time data – temperatures, loads, power draw, humidity levels. Optimizing these complex, interconnected systems is fundamentally a big-data problem, and AI excels here.

AI-driven cooling optimization represents the most immediate opportunity. Cooling systems in traditional data centers operate on static setpoints or manual adjustments, often leaving significant efficiency on the table. AI can continuously fine-tune settings based on weather patterns, workload predictions, thermal modeling, and historical performance data.

Google's DeepMind cooling system is the poster child for this approach. It reportedly cut cooling energy use by 40% in Google's facilities by learning optimal control strategies and adjusting settings in real-time. Startups like CoolGradient in the Netherlands are bringing similar capabilities to the broader market, offering AI-powered software that connects "roof-to-floor" – from chillers to servers – to analyze and optimize resource use. Early deployments demonstrate double-digit percentage reductions in cooling energy.

Phaidra, founded by ex-Google DeepMind engineers, offers AI control systems that automatically adjust cooling equipment. In tests, their algorithms maintained server inlet temperatures while significantly cutting chiller power usage. TycheTools focuses on real-time environmental monitoring and airflow management optimization, alerting operators to inefficiencies like stranded cooling or over-provisioned airflow.

The value proposition is compelling: even a 5-10% improvement in energy efficiency at a large data center can save hundreds of thousands of dollars and many tons of CO₂ annually. At scale across thousands of facilities, the impact multiplies dramatically.

Next-generation Data Center Infrastructure Management (DCIM) platforms are evolving beyond traditional monitoring. These systems now incorporate real-time PUE tracking, carbon accounting that integrates with energy market data to calculate facility carbon intensity, and predictive maintenance. They give operators unified visibility across multi-site operations, enabling benchmarking and knowledge transfer from efficient sites to underperforming ones.

The Uptime Institute notes that optimization of data center operations can extend across the entire life cycle of infrastructure, from design through operation to decommissioning. This digital layer – software and intelligence – represents an asset-light, highly scalable approach to sustainability that can be deployed across existing facilities without massive capital expenditure.

Breaking the Cooling Bottleneck

While AI optimizes existing systems, advanced cooling technologies are fundamentally reimagining how we remove heat from IT equipment.
Traditional air-based cooling has reached its limits. As computing density increases – particularly with AI accelerators generating 500+ watts per chip – air simply can't transfer heat efficiently enough. Moreover, air cooling is energy-intensive, consuming roughly 40% of total data center energy in typical facilities.

Immersion cooling represents a quantum leap. By submerging servers in specially designed dielectric fluids, heat transfer efficiency increases dramatically. Submer, a Barcelona-based startup that raised $55 million in Series C funding, reports that immersion cooling can cut cooling power consumption by up to 95% while enabling much higher server densities. The warm fluid leaving the tanks is also easier to capture for heat reuse than hot air from traditional systems.

Asperitas in Amsterdam offers modular immersion tanks that can integrate into existing facilities. Their systems have been deployed in French data centers where the captured waste heat feeds into local heating networks. Iceotope in the UK focuses on sealed chassis-level liquid cooling, claiming up to 40% energy savings on cooling while maintaining compatibility with standard racks.

Free cooling leverages environmental conditions to reduce or eliminate mechanical refrigeration. In moderate and cold climates – abundant in Northern Europe – outside air can cool data centers for significant portions of the year. Facilities in Stockholm and other Nordic locations achieve PUE ratings around 1.1-1.2 by maximizing free cooling. Smart systems that automatically adjust between cooling modes based on weather forecasts and real-time conditions can dramatically reduce compressor use.

Geothermal cooling taps stable underground temperatures. While requiring specific geology and upfront investment, it provides consistent, low-energy cooling year-round. Some European data centers are exploring shallow geothermal loops as part of their cooling strategy.

The cooling revolution isn't just about efficiency – it's about enabling the next generation of computing. High-performance AI workloads simply can't be deployed at scale with traditional air cooling in many existing facilities. Advanced cooling unlocks capacity that would otherwise remain stranded.

Turning Waste into Assets

Perhaps the most elegant innovations involve transforming data center byproducts into valuable resources.

Heat reuse converts the industry's biggest "waste" product into a commodity. Data centers generate enormous amounts of heat – traditionally expelled to the atmosphere through cooling towers. But that heat represents energy that cost money to create. Why not use it?

District heating integration is gaining traction across Northern Europe. Facilities in Sweden, Denmark, and the Netherlands channel waste heat into municipal heating networks. These systems can offset natural gas consumption for thousands of homes, reducing both emissions and heating costs for communities.

More creative approaches are emerging. Heata, a UK startup, installs distributed computing units in residential homes. Each unit's waste heat provides free hot water to the household while performing cloud computing tasks for paying clients. The model saves each household roughly £300 and 750 kg of CO₂ per year. At scale, it addresses both computing demand and fuel poverty while recycling energy that would otherwise be lost.

Deep Green took this concept to swimming pools, installing a mini data center at a public pool to heat the water for free. Qarnot Computing in France markets "digital boilers" – smart radiators with embedded processors that provide heating while doing cloud computation.

These distributed models face challenges around workload reliability and hardware maintenance, but they represent fundamentally new thinking about where computing happens and how energy flows through systems. They blur the line between data center and energy infrastructure in ways that create value on both sides.

Thermal storage systems can store excess cooling capacity for later use. Phase-change materials or ice storage allow facilities to create cooling during off-peak hours (when grid electricity is cleaner and cheaper) and use it during peak demand periods. This reduces grid stress while improving economics and sustainability.

Powering the Green Transition

Beyond efficiency, the source of power matters. Innovations in renewable integration and energy management are making data centers more sustainable and resilient.

On-site generation is becoming more common as solar costs fall. While data centers' huge power appetites mean rooftop solar typically covers only a fraction of demand, every bit helps – and more importantly, pairing generation with storage creates flexibility. Some facilities co-locate with renewable plants to directly use power that might otherwise be curtailed due to grid constraints.

Smart grid interaction turns data centers from passive loads into active grid participants. By modulating demand in response to grid conditions – drawing power when renewable generation is high, reducing load when the grid is stressed – facilities can earn revenue through demand response programs while improving grid stability. Battery systems sized for backup power can serve double duty, participating in frequency regulation or energy arbitrage when not needed for emergency backup.

Microsoft's successful test of hydrogen fuel cells running backup generators for 48 hours straight demonstrates that clean backup alternatives to diesel are becoming viable. As green hydrogen production scales and costs fall, fuel cells could replace generators entirely, eliminating one of the last major sources of on-site emissions.

These energy innovations transform data centers from mere consumers to prosumers – both using and potentially supplying power to the grid, integrating with renewable resources, and providing valuable flexibility services to energy systems under strain from the transition to variable renewable generation.

Closing the Loop on Resources

Circular economy principles are extending beyond energy to all resources data centers consume.

Water conservation is critical as facilities face increasing scrutiny over consumption. Epic Cleantec's water recycling systems can recycle up to 95% of wastewater for reuse in cooling towers and other non-potable applications. This virtually eliminates draw on municipal water supplies – particularly important in water-stressed regions or where data center growth is competing with other water needs.

Some facilities are exploring closed-loop cooling systems that use minimal water, or advanced dry-cooling technologies that eliminate water use entirely, though often at the cost of higher energy consumption. The Climate Neutral Data Centre Pact includes water usage effectiveness metrics, pushing the industry toward solutions that balance water and energy efficiency.

IT asset lifecycle management addresses embodied carbon and e-waste. Rather than disposing of servers after three years of hyperscale use, companies are refurbishing and reselling them for less demanding workloads. This extends hardware lifespan while making capable equipment available to markets that couldn't afford new systems. Proper recycling of components, batteries, and electronics reduces the environmental footprint of inevitable end-of-life disposal.

Sustainable construction involves using low-carbon concrete, recycled materials, and modular designs that can be reconfigured as needs evolve rather than demolished and rebuilt.

No Silver Bullets, But a Complete Arsenal

The transformation of data center sustainability isn't driven by a single breakthrough technology. It's emerging from an ecosystem of complementary innovations – each addressing specific pain points, each enabling others.

AI optimization makes existing facilities dramatically more efficient. Advanced cooling enables the density needed for modern AI workloads while slashing energy use. Heat reuse turns waste into revenue. Renewable integration and smart grid participation reduce carbon intensity while creating new business models. Circular approaches minimize resource consumption across the entire lifecycle.

Together, these technologies are proving that the choice between digital growth and environmental responsibility is a false dichotomy. The data centers emerging from this innovation wave are simultaneously more powerful, more efficient, more sustainable, and increasingly, more economical to operate than their predecessors.

The question is no longer whether technology can deliver green data centers. It's how quickly we can deploy these solutions at scale.

This blog post is Part 2 of our Green Data Centers series. Part 1 is here. Stay tuned for Part 3 where we explore the investment landscape

We are here to invest in and support startups making a difference. Get in touch: hello@greencode.vc