As the AI data centre boom drives record levels of investment and construction, the industry is being urged to ensure commissioning standards and environmental accountability keep pace with the breakneck speed of delivery. Louis Charlton, Group CEO of the Global Certification & Verification Group, offers his view on the commissioning and verification standards that should be considered, while Structure Research’s report outlines the balance between unprecedented infrastructure growth and long-term sustainability.
The AI data centre boom is putting commissioning and verification standards under pressure and people are starting to pay attention.
The data centre industry is racing to build the necessary infrastructure to support the next phase of the AI buildout. This year, the five biggest AI infrastructure companies will spend somewhere between $660 billion and $690 billion on building AI data centres, almost twice what they spent in 2025. Worldwide data centre power consumption is forecast to grow by 50% between 2025 and 2027, rising by as much as 165% by the end of the decade.
The hurdles that have arisen as a result of meeting this generational surge in infrastructure demand are well documented. From securing the necessary access to power from ageing grids to securing skilled engineers and technicians, many challenges inherent to executing the AI boom are being widely discussed. But the accelerated pace and sheer scale of data centre demand poses other problems that, as yet, aren’t getting the attention they deserve from the industry.
As AI demand pushes for faster data centre construction and compresses delivery timelines, the race to deliver the next generation of digital infrastructure is putting unprecedented pressure on the commissioning process. This is happening at a time when new data centre designs, bigger facilities and shorter project windows are making independent verification and certification more critical than ever. The industry’s focus on speed-to-market is ramping up pressure on testing and validation processes and the resultant incentive to cut corners is raising the potential operational and commercial risk if the need for speed compromises verification.
Global Commissioning is sounding the alarm that, as AI demand accelerates data centre construction and compresses delivery timelines, commissioning and independent verification are becoming more critical than ever.
Commissioning: The invisible, invaluable last line of defence
Commissioning rarely makes headlines. But when it goes wrong, the consequences certainly do.
The commissioning process is widely understood as a set of tests that take place close to the completion of a project. Its technical and regulatory necessities are broadly agreed upon, but ask why commissioning matters at a business level and many people in the industry will struggle to articulate their answers.
In practice, commissioning is so much more than an exercise in compliance. A rigorous commissioning programme begins at design review, long before a cable is pulled and runs through every layer of a building’s systems, from the component level up to integrated performance under full operational load. It is the process that determines whether a data centre actually does what its designers intended.
The industry shorthand for this is L0 to L6: a structured testing methodology that progressively validates each system layer, culminating in integrated systems testing and operational readiness. When it’s done properly, it produces a test record that is a genuine risk management instrument. That record protects developers, operators and investors alike.
When that process is compressed, deferred, or treated as a box-ticking exercise, that protection disappears and the consequences tend to surface at the worst possible moment.
This is the moment to centre commissioning, not set it aside
Commissioning is not just a technical exercise or final-stage checklist. It is a critical risk management and accountability process that protects long-term data centre performance, resilience and investor confidence.
The data centre sector is building faster than at any point in its history. Hyperscale demand, AI infrastructure investment and energy transition pressures are all compressing timescales and raising the stakes on every delivery decision. In that environment, the tolerance for substandard commissioning, for incomplete test records, deferred defect resolution and integrated testing that never quite happened, is shrinking fast.
Operators who have been through difficult handovers are restructuring how they procure commissioning authority. Developers are being asked harder questions about what their documentation actually reflects. And the wider market is converging around a new understanding of what credible, independent verification looks like: one that integrates commissioning, certification and operational validation into a single, accountable chain.
The data centre industry is engaged in an infrastructure buildout of generational scale and significance to the fabric of the modern world. The industry responsible for certifying and verifying that infrastructure is fit for purpose should not be relegated to a box-ticking exercise.
The environmental cost of building at speed
The commissioning challenge is just one pressure point. The same AI-driven buildout is also generating growing scrutiny of the industry’s environmental footprint, as a new report from Structure Research makes clear. The independent research firm’s 2026 State of Environmental Impact Report sheds light on the impact of accelerating AI infrastructure growth, examining environmental data from 38 data centre providers and nine hyperscale cloud platforms, tracking trends across carbon emissions, energy consumption, renewable energy usage, water consumption and operational efficiency between 2020 and 2025.
As hyperscale AI deployments and high-density compute workloads continue to reshape the infrastructure landscape, the report finds that data centres accounted for an estimated 1.23% of global energy consumption in 2025, up from 0.81% in 2020. Total data centre energy consumption increased from 198.7 TWh in 2020 to 361.6 TWh in 2025.
Despite this rapid growth, the industry continues to make measurable progress in efficiency and renewable energy adoption.
“AI infrastructure growth is fundamentally changing the scale and resource profile of the global data centre market,” said Philbert Shih, Managing Director of Structure Research. “The industry is facing increasing scrutiny around energy availability, emissions and water consumption, but at the same time we are seeing meaningful improvements in efficiency, transparency and carbon-free energy adoption. The challenge moving forward will be balancing unprecedented infrastructure growth with long-term sustainability goals.”
One of the key findings from the report highlighted how AI workloads are reshaping infrastructure demand. The data shows that the total operational IT capacity in the data centre industry reached an estimated 80,242 MW in 2025, up from 44,046 MW in 2020.
Hyperscale self-build capacity grew at a five-year CAGR of 17.6%, reflecting the rapid expansion of AI and cloud infrastructure globally. Meanwhile higher-density AI workloads are driving increased adoption of liquid cooling technologies and larger-scale campus developments.
The report also found that renewable and carbon-free energy adoption is accelerating among industry leaders. Renewable energy usage among ESG leaders grew at a five-year CAGR of 26.2%, significantly outpacing overall energy consumption growth. The results showed that hyperscalers sourced approximately 92% of their energy usage from carbon-free energy in 2025, while data centre providers reached 69%.
In addition, nuclear energy, natural gas partnerships and direct power procurement strategies are becoming increasingly important as providers confront grid constraints in major markets.
The report also found that efficiency gains are continuing despite growing consumption. Average Power Usage Effectiveness (PUE) for data centre providers improved from 1.44 in 2020 to 1.38 in 2025. Hyperscalers maintained industry-leading average PUEs of approximately 1.21.
The average emissions per GWh of energy consumption decreased from 328.3 mtCO2e/GWh in 2020 to 229.3 mtCO2e/GWh in 2025, demonstrating improved operational efficiency and cleaner energy sourcing.
The research found water usage emerged as a critical industry focus. The total water consumption by ESG leaders increased from 55.8 million m3 in 2020 to 114.9 million m3 in 2025 as AI-related liquid cooling deployments accelerated.
The report highlighted increasing adoption of closed-loop liquid cooling systems, hybrid cooling designs and non-potable water strategies to improve water efficiency and reduce environmental impact.
The AI data centre boom shows no signs of slowing and neither does the pressure this is placing on the industry. As investment scales and timelines compress, the message from both stories is clear: building fast and building responsibly are not mutually exclusive, but achieving both will require the industry to treat commissioning standards and environmental accountability as priorities, not afterthoughts.
