AI data centre surge tests labour, energy & grids

AI-driven data centre expansion is expected to intensify pressure on power grids, labour markets and environmental targets in 2026, according to industry executives, as operators and governments worldwide work to align infrastructure, regulation and workforce capacity with rapidly rising compute demand.

Predictions from companies including Onnec, Informatica, Adactin and Neara point to a year in which energy use, data quality, skills supply and local community consent move to the centre of global AI infrastructure planning.

Flexible labour

UK-based connectivity and data centre specialist Onnec expects a structural shift in how data centre projects are staffed as AI workloads grow.

Claire Keelan, Managing Director UK at Onnec, said that flexible labour models will become central to delivery across the design, build and operations lifecycle.

"Flexible labour models will underpin almost every new data centre project. Traditional staffing can't scale at the speed AI demands. By 2026, flexible, crowdsourced, project-based teams will fill critical gaps across design, building, and operations. This shift isn't about replacing expertise, it's about redeploying it. Clear standards, accreditation, and safety frameworks will make flexibility viable at scale, turning part-time professionals and returning workers into a reliable, high-quality talent engine."

Diversity push

Keelan also highlighted workforce composition as a constraint on expansion, pointing to the low representation of women in data centre roles. "With women making up less than 8% of the current workforce, the imbalance is holding the sector back. In 2026, diversity will shift from a talking point to an operational priority. This means targeted recruitment, retraining programmes, and mentorship networks designed to bring more women into engineering, safety, and leadership roles. Diversity will be treated as a business resilience issue, not just a social goal. This is because the industry can't meet AI's demands while sidelining a sizeable portion of its potential workforce."

Regional hubs

Onnec expects AI demand to accelerate the shift of UK data centre development away from London and the M4 corridor and towards regional locations.

"Regional 'AI growth zones' will emerge as the new engines of capacity. In 2026, Manchester, South Wales, and Scotland will continue to gain momentum thanks to lower land costs, access to renewable energy, and close ties to academic institutions. This regional diversification will help balance power use and strengthen resilience against local constraints. The days of London and the M4 corridor as the single dominant hub are fading; the future of data centres is distributed, collaborative, and regionally connected," said Keelan.

Retrofit pressure

The UK's large base of legacy facilities is expected to face mounting pressure to adapt to AI workloads. Onnec sees a sharp rise in retrofitting activity as operators seek to increase density and performance.

"With the UK home to one of the world's largest portfolios of legacy data centres, next year operators must prove how fast they can innovate to stay ahead in the new AI landscape. In 2026, we'll see a surge in retrofitted data centres as operators rush to upgrade legacy sites to meet soaring AI demand. Power and cooling will be complex, but cabling and network capacity will be the real bottlenecks. Poor-quality or overcrowded cabling limits density, throttles performance, and makes future upgrades almost impossible. Smart operators will invest early in high-grade structured systems that support modular expansion and long-term flexibility. "Retrofit-ready" will become the new benchmark for responsible, future-proof design," said Keelan.

Australia's data focus

In Australia, the national AI capability plan has shifted policy debate towards the role of data as an asset underpinning sovereign AI capabilities, according to Informatica.

Alex Newman, Country Manager ANZ, Informatica, said, "Australia's announcement of a national AI capability plan and its emphasis on data is a welcome and important step. Policymakers have moved the conversation beyond raw compute and rightly signalled that data is central to any sovereign AI ambition. Recognising data as strategic infrastructure is the foundation for trustworthy, useful and competitive AI."

"However, recognising data is only the first step. The early public debate has understandably zoomed in on hyperscale facilities, GPU counts and latency - the visible elements of an AI ecosystem. But compute and chips are only half the story. The other half is the quality, governance and operational readiness of the data that feeds those machines."

"AI only performs as well as the data that fuels it. Without high-quality, well-integrated, and trusted data, even the most powerful models will underdeliver, producing biased, brittle or unusable outcomes. To turn Australia's growing compute footprint into a genuine sovereign capability, we must build the invisible, but essential, data layer alongside it," said Newman.

Newman outlined four principles for implementation: integration at scale, quality and trust, governance and transparency, and the continuous operationalisation of data pipelines.

"Four practical principles should guide the next phase of implementation: Integration at scale: AI models require data from hundreds of sources - enterprise systems, public datasets, IoT streams and more. That data must be unified and accessible in forms that models can ingest rapidly and reliably. Quality and trust: "Garbage in, garbage out" is amplified at AI scale. Data cleansing, enrichment, and validation aren't optional; they are mission-critical for accuracy and fairness. Governance and transparency: Regulators, citizens and customers will demand lineage, explainability and controls. Built-in metadata, provenance tracking, and auditability are essential for demonstrating how models were trained and why they make the decisions they do. Continuous operationalisation: AI is not a one-off project. Pipelines must feed fresh data into training, retraining and deployment cycles so systems remain accurate and relevant."

"The good news is that these are solvable engineering and governance problems but they require targeted investment and a clear national posture. Australia's plans should couple hardware commitments with funding, standards and incentives for the modern data management infrastructure that ensures data is discoverable, clean, secure and compliant."

"The Australian government is to be commended for elevating data in the national AI conversation. Now let's match that leadership with the practical, long-term data investments that will turn potential into performance and ensure Australia's AI future is robust, responsible and nationally valuable," said Newman.

Energy and emissions

Alongside data strategy, Australian commentators expect AI-related demand to intensify scrutiny of electricity use and emissions from data centres.

Navneesh Garg, CEO, Adactin, said, "With AI usage soaring, we're seeing an extraordinary demand for computing power, with some research suggesting Australia will need up to 175 new data centres by 2030. The demand for electricity to power those centres will ensure the issue of environmental and social responsibility stays squarely on state and federal government agendas next year and every one after that."

Garg said operators and enterprises will be expected to adopt a range of measures to manage environmental impacts. "Businesses will increasingly be expected to do some heavy lifting: finding ways to reduce the carbon footprint of IT workloads; utilising carbon-aware scheduling; deploying digital twin technology to simulate and optimise energy consumption across physical assets, supply chains and operational environments; and reducing e-waste via more rigorous re-use and recycling campaigns."

Grid constraints

Neara expects electricity networks themselves to become a potential brake on AI build-out if grid upgrades struggle to keep pace with the connection needs of new facilities.

Taco Engelaar, SVP and Managing Director, Neara, said, "The AI bubble might burst in 2026, but not for the reason people think. Investment has poured into new data centre developments to power global AI ambitions, but the energy systems required to support them are on their knees. To increase capacity for new data centres, policymakers are proposing extensive grid expansion; but if history is anything to go by, public opposition could stop it in its tracks."

Engelaar said public resistance to new network infrastructure could echo previous disputes over transmission projects. "We've seen it before with pylons: not-in-my-backyard (NIMBY) attitudes can significantly slow - or stop - new infrastructure plans. Now, AI data centres are primed to become the next NIMBY battleground."

He said digital modelling could allow operators and planners to use existing assets more intensively and reduce the need for new builds. "Ironically, AI could be the solution. When used to deliver digital modelling, for example, it can help detect untapped capacity in the grid. This could unlock more room for data centre connections while reducing the need for new infrastructure. Balancing AI's energy demands with those of the public will be a critical challenge in 2026. Success will start and end with investing in the grid."