Powering AI Doesn’t Have to Mean Abandoning Sustainability Goals

This year’s Earth Day arrives at a pivotal moment for the digital economy. AI acceleration is driving a surge in demand for compute, and in turn, a surge in demand for energy.

This is a sustainability issue, but it’s more than just that: It’s a scaling challenge. The availability, cost and carbon impact of energy are quickly becoming defining constraints on how and where AI can be deployed.

Some organizations are beginning to question whether their sustainability commitments can keep pace with AI-driven growth. A recent custom survey from 451 Research, S&P Global found that 39% of respondents either “agreed” or “strongly agreed” that growing electricity demands from AI systems are directly limiting their ability to invest in carbon-free energy initiatives.[1]

However, this is a false trade-off. Leading companies in the AI era won’t choose between innovation and sustainability: They’ll build infrastructure strategies that deliver both.

Earth Day is a reminder that long-term progress requires long-term thinking. For enterprises embracing AI, that means rethinking how they power, optimize and sustain digital infrastructure, so that growth doesn’t harm the communities and ecosystems that support them.

On this Earth Day, we at Equinix reaffirm our commitments: We won’t pursue growth at the expense of our values, and we won’t betray the trust that customers and stakeholders have placed in us. Our values will drive us to pursue growth in a responsible and sustainable manner, and we’ll help our customers do the same.

To scale AI sustainably, enterprises must look at digital infrastructure through a new lens.

Three Levers of Sustainable AI Infrastructure:

• Energy supply: Where power comes from
• Operational efficiency: How power is used
• Life-cycle impact: The hidden carbon footprint of data centers

Increasing clean and renewable energy coverage

Data center operators are major consumers of energy. They should therefore support the energy grids of the communities in which they operate.

When colocation providers invest in local energy grids, it doesn’t just unlock the power of enterprise AI; it helps avoid passing higher energy costs on to the community. Power purchase agreements (PPAs) are one mechanism to achieve this.

A PPA is a long-term contract between an energy buyer and an energy developer. The buyer agrees to purchase electricity from a new renewable energy development at predetermined rates over a certain time. This makes renewable energy projects viable where they might not be otherwise:

• Mitigating financial uncertainty: Large-scale renewable energy projects can be expensive and risky. Banks are more likely to provide funding if a major energy buyer is arranged in advance. Also, predetermined pricing is a hedge against future volatility in the energy market.
• Adding to the grid: PPAs are highly impactful because they directly contribute new renewable energy capacity to local grids. Thus, they benefit entire communities, not just the signatories.

In addition to renewables PPAs, leading data center operators are pursuing on-site clean energy generation that can directly reduce demand for grid energy:

• Fuel cells can produce reliable electricity via an electrochemical reaction instead of combustion, with fewer emissions than the grid average and zero particulate matter. When they use a clean fuel such as green hydrogen, on-site fuel cells create even lower emissions.
• On-site solar panels can be an economically viable way to augment grid energy in certain locations.
• Nuclear small modular reactors (SMRs) aren’t widely used yet, but we expect they’ll play a pivotal role in providing clean energy for the next generation of data centers.

Investing in efficiency improvements

Using cleaner energy is one way to reduce the carbon emissions of data center operations, but it’s equally important to use energy more efficiently.

Data center operators pursue greater efficiency via several different initiatives:

• Advanced cooling: As servers run, they create heat. How efficiently cooling systems remove that heat determines how efficient the facility is overall. Liquid cooling is inherently more efficient than air cooling, and it’s become the gold standard for high-density workloads like AI.
• Airflow management: Ensuring separation between cold supply aisles and hot exhaust aisles helps prevent unnecessary cooling that wastes energy. This is a widely accepted best practice in the data center industry, yet it hasn’t been adopted consistently.
• Optimizing operations with AI: Training machine learning models with operational data from real data centers can help proactively identify efficiency improvements.

Power usage effectiveness (PUE) is the metric that data center operators use to measure the impact of efficiency initiatives. When a colocation provider demonstrates consistent PUE improvements over time, customers can feel confident that their equipment will run as efficiently as possible.

In addition to everything colocation providers do to improve efficiency, customers can help by building an environment that’s sized appropriately for their needs. This helps avoid underutilization of servers, which is the ultimate efficiency killer.

Optimizing for lower emissions throughout the data center life cycle

There’s more to data center sustainability than just operational energy consumption. Everything that goes into building and running a data center has an impact on its total carbon footprint, from the first raw materials harvested all the way through to the eventual decommissioning.

Data center operators must address carbon emissions throughout the entire life cycle of their facilities:

• Embodied carbon: This refers to the emissions created by harvesting data center materials and manufacturing components. Working with the right suppliers can help avoid unnecessary emissions, reduce unavoidable emissions, and pursue innovative low-carbon alternatives.
• Circularity and responsible disposal: Since all data center components have emissions attached to them, it’s important to minimize waste by reusing and recycling components once they reach end of life. When e-waste cannot be reused securely, it must be disposed of properly.
• Whole building life-cycle assessment: A WBLCA shows carbon impact across every stage in the data center life cycle. It can help data center operators identify areas for improvement.

What sustainability means for business leaders

All the different sustainability and efficiency initiatives summarized above clearly show that enterprises don’t have to choose between executing their AI strategies and pursuing their sustainability goals. The right approach can address both priorities simultaneously:

• Energy strategy is AI strategy: Increasing demand for energy doesn’t have to mean eating into existing energy supplies; it can be the driver behind new clean and renewable energy projects that directly increase capacity.
• Sustainability is a global priority: AI is happening all over the world, so it’s not enough to pursue isolated sustainability initiatives in just a few places. When customers use a colocation provider that understands this, they can drive better results wherever their AI strategy takes them.
• The right partners make all the difference: Colocation providers work with renewable energy developers and sustainability-minded contractors and suppliers in pursuit of common goals. They can pass on the benefits of these partnerships to customers.

Marking our sustainability progress

Equinix’s continued commitment to our sustainability priorities, even during this time of AI-driven demand, allowed us to achieve the following results during 2025:

• Increased renewable energy coverage: We signed four new PPAs during 2025, and 14 previously contracted projects came online. As of year-end, we’d installed 73 MW of fuel cells, with an additional 35 MW in progress. We achieved 96% renewable coverage globally, and we’re on track to meet our goal of 100% coverage by 2030.
• Improved efficiency: Our global annualized average PUE was 1.37. This is an efficiency improvement of about 5% from 2024, and is down about 31% from our 2019 baseline. We achieved this via a $36 million investment to replace existing infrastructure across 79 sites with more efficient technology.
• Reduced waste: In 2025, we recycled more than 3,900 metric tons of operational waste. We’re also collecting waste data across 140 different sites. We’ll use this data to better understand our operational waste and identify areas for further improvement.

AI is pushing infrastructure to its limits, and energy is one of the defining factors in how far it can scale. With the right strategy, enterprises don’t have to choose between growth and sustainability. They can achieve both. At Equinix, we’re helping organizations turn energy and efficiency into a strategic advantage, so they can scale AI with confidence.

Learn more about our perspective on managing sustainability goals in the AI era, and the progress we made toward those goals in 2025: Access our sustainability report.

[1] 451 Research, S&P Global Energy Horizons report commissioned by Equinix, December 2025.