At the European Data Center Associations Summit hosted by Vertiv, industry leaders argued that AI growth is forcing data centres to rethink their relationship with power grids, communities and policymakers.

Data centres are increasingly being asked to do more than simply consume electricity and provide computing capacity. As AI drives unprecedented demand for digital infrastructure, industry leaders across Europe are beginning to position data centres as active participants in the wider energy ecosystem.

That was one of the central themes emerging from the European Data Centre Associations Summit, hosted by Vertiv in Zagreb, where representatives from data-centre associations across Sweden, the Netherlands, France and Ireland discussed the infrastructure supporting Europe’s digital economy.

The conversation began with a reminder of how invisible data centres remain to most users. Consumers may interact with data centre infrastructure dozens of times every day, through banking, digital payments, video calls, cloud applications and government services, without ever thinking about the physical systems behind those services.

But AI is making that infrastructure increasingly difficult to ignore.

From grid burden to grid participant

As data centre demand expands, access to electricity has become one of the sector’s biggest constraints. Yet the panel argued that data centres should not necessarily be viewed purely as an additional burden on already stretched electricity networks.

Stijn Grove, Managing Director of the Dutch Data Center Association, pointed to the potential for data centres to locate closer to renewable-energy generation and absorb power that might otherwise require significant additional transmission infrastructure.

He also highlighted a broader opportunity: data centres could potentially help stabilise grids as renewable generation becomes more variable.

That idea was echoed by Ronan Kelly, CEO of Digital Infrastructure Ireland, who discussed the role of battery energy-storage systems and on-site backup capacity in supporting electricity networks during periods of peak demand.

The direction of travel is significant. Data centres are beginning to move from simply asking “How much power can the grid give us?” towards asking “How can our infrastructure interact with the grid?”

Waste heat becomes an asset

The panel also highlighted heat reuse as one of the clearest examples of how data centres can integrate more deeply into local communities.

Isabelle Kemlin, Vice Chair of the Board at Swedish Datacenter Industry, cited examples where waste heat from data centers is being reused through district-heating systems and even agricultural applications.

In the Netherlands, Grove pointed to projects where data-centre heat is being used to replace natural gas in buildings and potentially support greenhouse operations.

Such projects challenge the perception of data centres as isolated industrial buildings that simply consume electricity and generate heat. Increasingly, the heat itself can become part of another energy system.

AI changes the efficiency conversation

AI is also forcing the industry to reconsider how data center efficiency should be measured.

Traditional measures such as Power Usage Effectiveness (PUE) remain important, but several speakers argued that they do not always capture the complete picture.

For example, equipment installed to recover and redistribute waste heat may consume additional electricity and therefore worsen a facility’s PUE, even though the overall energy system becomes more efficient.

The discussion therefore moved towards a newer metric increasingly associated with AI infrastructure: tokens per watt.

Instead of measuring only how efficiently a facility delivers electricity to IT equipment, tokens per watt attempts to connect energy consumption with the amount of useful AI computation produced.

As AI factories become larger and more power-intensive, the ability to convert electricity into useful compute efficiently may become as important as simply minimising facility overhead.

Europe’s sovereignty challenge

Energy is not the only reason Europe will continue to require significant local data center capacity.

The panel also highlighted digital sovereignty.

Europe’s fragmented national landscape means governments, public-sector organizations, and regulated industries frequently need to consider where data is stored and processed. Moving workloads across borders may be technically possible, but sovereignty, security, and latency requirements can make local infrastructure essential.

That creates a very different environment from markets where computing resources can be concentrated across a smaller number of enormous geographic regions.

The public-perception problem

Perhaps the industry’s biggest challenge, however, is not technical.

Several panellists acknowledged that public perceptions of data centers remain dominated by concerns around electricity consumption, water usage, land requirements and limited employment creation.

Michaël Reffay, Managing Director of France Data center, argued that many of these criticisms overlook the wider economic and digital services supported by data center infrastructure.

Kelly made a similar point, arguing that discussions about data centre carbon emissions frequently focus only on the electricity consumed by facilities while ignoring emissions potentially avoided through digital services such as remote working, digital banking and electronic distribution.

The industry therefore faces a communication challenge alongside its engineering one.

Data centers will consume significant amounts of energy as AI expands. But the sector increasingly wants policymakers and the public to judge that consumption alongside the digital services, economic activity, renewable-energy investment, and wider infrastructure benefits it enables.

AI makes infrastructure strategic

Perhaps the clearest conclusion from the discussion was that digital infrastructure is no longer simply a back-end utility.

AI is making access to power, cooling, grids and computing capacity strategic economic issues.

For Europe, the next phase of the data centre debate may therefore be less about whether more facilities should be built and more about where they are built, how they interact with energy systems, and how effectively the industry can demonstrate their value to society.

Kayvan Karim, Programme Director of MSc Software Engineering, School of Mathematical and Computer Sciences, Heriot-Watt University Dubai 

For the past few years, the AI infrastructure race has largely been measured in scale: more GPUs, larger data centres and greater power capacity. That expansion is continuing, but the economics are beginning to change. Competitive advantage may increasingly depend not only on securing additional power, but on extracting more useful AI work from the power already available. Part of the reason is a change like AI demand. We are moving from relatively simple prompt-and-response systems towards agents that can reason across multiple steps, call tools, inspect results, revise plans and continue working autonomously. Anthropic’s latest Economic Index notes that Claude usage is increasingly shifting towards long-running agentic tasks and that more computationally intensive conversations tend to be associated with higher-value outputs.

This transition could have significant implications for infrastructure demand. A chatbot might generate one answer to one prompt. An agent performing a software-development, research or business task may invoke a model dozens of times, maintain a large context, call external tools and generate many intermediate reasoning steps before producing an outcome. Demand may therefore grow in two ways: more people using AI and more inference, model calls, and tokens processed for each task. Major AI laboratories are already working on this efficiency problem. OpenAI says one of its primary inference objectives is to serve more tokens from the same hardware, using techniques including scheduling, caching, kernel optimisation and improved model implementation. It also describes GPT-5.6 as being trained to accomplish more work per token. Google DeepMind is pursuing a similar direction: its Gemini 3.6 Flash was designed for scaled agentic workloads and uses fewer output tokens than its predecessor on several evaluations. In contrast, its recent agentic video system reduced token consumption by up to 88% for that workload.

Other approaches address efficiency at the model architecture level. DeepSeek-V3, for example, uses a Mixture-of-Experts design with 671 billion total parameters but activates 37 billion per token, so only part of the network is used for each computation. Meta has similarly worked on inference efficiency through grouped-query attention and more efficient tokenisation; the Llama 3 tokeniser was reported to require up to 15% fewer tokens than Llama 2 for equivalent text. Taken together, these approaches show that model capability is increasingly being developed alongside the cost of delivering it.

Model-level efficiency, however, is unlikely to remove the infrastructure constraint on its own. Global data-centre electricity consumption was approximately 415 TWh in 2024, according to the International Energy Agency, and its base case projects this to reach around 945 TWh by 2030. AI is expected to drive most of that growth. Efficiency is therefore improving while aggregate demand continues to rise. One reason is the Jevons, or rebound, effect: efficiency improvements reduce the resources required for each unit of work, but lower costs can also encourage greater overall use. If agents become much cheaper to operate, organisations may respond by deploying more of them, running them for longer, or applying them to tasks that were previously uneconomic. Efficiency can reduce the compute required for an individual task while still increasing total demand.

That increased demand meets infrastructure that cannot expand as quickly. Models and software can improve quickly, but grids, substations, transformers and power-generation infrastructure usually have much longer development cycles. The IEA notes that while a data centre can sometimes be developed within two or three years, the broader energy infrastructure required to support it often involves longer planning and construction periods. Where grid capacity is constrained, each available megawatt becomes a more valuable production resource. The amount of power available remains important, but so does the amount of useful computation that can be produced within that power envelope.

That changes how we should understand capacity. Improvements in accelerator performance, model architecture, workload scheduling, caching, utilisation and inference software can increase computational output without increasing a site’s electrical connection. OpenAI’s recently reported Jalapeño inference hardware illustrates the direction of travel: the company says the chip can deliver more AI work per unit of power while increasing throughput and reducing latency. Efficiency can therefore act as a form of virtual capacity. If two operators each control 100 MW, but one can consistently deliver substantially more useful AI work within that power envelope, their nominal capacity may be identical while their productive capacity is not.

The same constraint applies to physical space and cooling. AI systems are concentrating more computational power into individual racks, increasing both power density and heat output. Packing more accelerators into the same building only creates useful capacity if the electrical and thermal infrastructure can support them. This is one reason liquid cooling is moving from a specialist technology towards a more central part of AI data-centre design. Microsoft, for example, has introduced a closed-loop chip-level cooling architecture that it says eliminates evaporative water consumption for cooling and could avoid more than 125 million litres of water annually per data centre. The example also shows why power, cooling, water use and rack density cannot be treated independently.

The same shift creates a measurement problem. Power Usage Effectiveness, or PUE, has been valuable for showing how much facility energy is required beyond the electricity IT equipment consumes. It does not, however, measure whether that IT equipment is producing useful work efficiently. Uptime Institute’s 2025 survey placed average PUE at around 1.54 and noted that the headline industry figure had changed little for six years. Uptime has consequently argued for productivity measures that relate computational work to energy consumption. For AI inference, tokens per kilowatt-hour might offer one operational measure. Still, even that is incomplete: an efficient model that solves a task in 1,000 tokens may be more valuable than one generating 10,000. A more useful long-term measure may therefore be useful AI work per unit of energy, water and infrastructure.

Capacity will remain essential. The AI industry will continue to build larger data centres, secure new power supplies and deploy large quantities of computing hardware. As agentic systems create more persistent inference demand and physical resources become harder to expand, however, efficiency may increasingly determine the productive value of that capacity. Operators that can support more useful computation within the same power, cooling, water, and space constraints can accommodate more workloads without waiting for equivalent growth in physical infrastructure.

For AI infrastructure, installed megawatts will remain a headline measure. The more consequential measure may increasingly be how much useful AI work those megawatts can support.

Alteryx, the agentic analytics and automation company, today announced new AI capabilities across Alteryx One that connect enterprise-grade business logic directly to the AI agents’ teams already use. By extending governed workflows and datasets to external AI tools, organizations can “build once and govern once,” eliminating the need to recreate complex business logic from scratch. This approach allows enterprises to scale AI action with confidence, helping to reduce both security risks and runaway token costs.

Key Findings

• 71 percent of IT leaders report that AI initiatives are most successful when IT and business teams collaborate closely to bridge the gap between AI agents and enterprise business logic.
• NextWave achieved a 20x reduction in LLM token consumption using an Alteryx workflow during a complex Office of the CFO reconciliation between front-office and back-office data.
• Up to 93 percent reduction in token consumption and up to 85 percent increase in speed on tasks involving raw, ungrounded data, when combining an LLM with an existing, trusted Alteryx workflow.
• Up to 83 percent reduction in token costs and up to 65 percent increase in speed on tasks involving clean, grounded data.
• 65 percent of analysts confirm that AI delivers the most value when business logic is managed at the business level.

“Generative AI is brilliant at brainstorming, but it often struggles with the precision required for enterprise execution. Organizations don’t need agents that guess at business rules and burn through tokens; they need AI that operates on the same trusted business logic and governance that underpin the rest of the business,” said Ben Canning, Chief Product Officer at Alteryx. “By connecting existing tools to a governed business logic layer, we are allowing enterprises to stop the ‘re-work’ tax of rebuilding business rules for every new agent, ensuring that every AI-driven action is as reliable as the calculations they already trust.”

New Capabilities

The latest capabilities include:

Ask Alteryx: With this release, Ask Alteryx evolves from an embedded assistant into the primary way users interact with Alteryx One, guiding new users step-by-step through their first workflow in Designer and giving everyone a natural-language front door to their data through Ask Alteryx for Live Query, with connections to Snowflake, Big Query, and Databricks for reading and writing data directly. Ask Alteryx checks existing workflows and data first, delivering a governed answer when one exists or building a new workflow when it doesn’t, with every output remaining inspectable, editable, reusable, and schedulable within Alteryx One.

Agent Studio: Enables business users to turn existing, already-governed datasets into conversational agents without rebuilding anything. Analytics teams maintain full control over which datasets and KPIs power agent responses, while finance and operations departments can instantly scope an agent to their reconciliation dataset or KPI dashboard data, so stakeholders can ask trend, root-cause, and variance questions in plain language and get governed, explainable answers back.

Alteryx Insights for OpenAI: Available through the ChatGPT Plugin Directory, this capability allows business users to generate answers based on analyst-approved data, calculations, and workflows. Employees can investigate revenue variances or resolve reconciliation issues directly, accessing trusted business logic without opening the platform or requiring an Alteryx seat. Alteryx will be expanding this surface integration strategy to bring governed logic to where teams already collaborate, including upcoming support for Claude, Gemini, Slack, and Microsoft Teams.

Alteryx MCP Server: The governed way to use Alteryx from whatever AI platform or agent you already work in. It lets AI agents interact with Alteryx as easily as a human would, finding the right data, building multi-step solutions, and turning them into governed, repeatable workflows. Agents can build, run, schedule, and discover Alteryx assets directly, with external AI requests inheriting Alteryx’s authentication, workspace context, role-based access controls, and permissions automatically, so every interaction carries the same security model and audit trail as if a person had done it. More capabilities, including connection creation and expanded scheduling, are expected to roll out later this year on the same governed connection.

Alteryx Skills: A GitHub install that teaches third-party agentic interfaces, OpenAI Codex, Microsoft Copilot, Claude Code, Gemini CLI, and more, how to build Alteryx assets the right way, closer to how Ask Alteryx already builds them. Rather than each tool guessing at Alteryx’s patterns on its own, Skills gives them Ask Alteryx’s own workflow-building know-how, so a financial calculation or reconciliation workflow gets built correctly the first time, without rebuilding logic or permissions separately for every tool your team uses.