AI pushes dense, decentralised & sustainable data centres

This article draws on insights shared during Ingram Micro's B2B Tech Talk podcast episode, "Powering the Future with High-Density Innovation." As digital transformation accelerates and data volumes explode, the conversation around infrastructure has fundamentally shifted. The question facing IT leaders today is no longer if infrastructure needs to evolve, but how rapidly it can adapt to a new reality. We are witnessing a seismic shift driven by AI-driven workloads, advanced analytics, and a massive proliferation of data at the edge. These are not merely incremental changes; they represent a distinct era of "hungry" workloads that place unprecedented pressure on power, cooling, and environmental management.

Traditional data centres, designed for the workloads of yesterday, are increasingly ill-equipped to handle the thermal and power demands of this new wave. The challenge, stripped to its core, is a matter of physics: delivering power effectively to the chip and extracting heat efficiently back to the chiller. Modern chips require significantly more power, transforming the data centre equation into a complex logistical challenge of moving energy from the grid - often starting at 22 kilovolts - down through transformers, racks, and power distribution units (PDUs), directly to the silicon. Simultaneously, the heat generated must be captured and dissipated through innovative means, such as liquid cooling or rear-door heat exchangers.

To solve this, the industry is moving toward validated architectures - pre-tested designs where server portfolios and cooling solutions are integrated to ensure scalability and speed of deployment. However, a pervasive myth remains that high-density computing is the exclusive domain of hyperscalers and massive centralised facilities. This is demonstrably false. We are seeing "green shoots" of high-density innovation emerging at the edge. Research indicates that by 2028, nearly 50% of AI workloads will be processed at the edge, decentralising compute power and bringing high-performance infrastructure closer to where data is created.

This decentralisation creates a massive opportunity - and a significant responsibility - for Managed Service Providers (MSPs) and IT partners. The incumbent data centre often lacks the capacity to run these intensive AI workloads, opening the door for partners to provide specialised hosting facilities and value-added services. By managing these specialised workloads outside of the customer's legacy data centre, partners allow clients to focus on their core business outcomes rather than infrastructure maintenance.

However, high-density compute at the edge inevitably introduces Operational Technology (OT) into the mix. Unlike hyperscale data centres, which benefit from massive redundancy, edge deployments often lack that safety net. Therefore, monitoring the reliability of OT becomes critical. Partners who can bridge the gap between IT and OT, ensuring convergence and reliability, will unlock significant value for their clients, particularly in areas like cybersecurity and sustainability.

There is, however, a readiness gap that must be addressed. While 77% of end clients indicate a preference for deploying AI solutions with their existing trusted IT partners, the question remains: are those partners ready? The market is calling for partners to educate themselves on the nuances of AI workflows and the specific architectural requirements of high-density setups. It is not enough to simply sell hardware; partners must understand the entire ecosystem, from Data Centre Infrastructure Management (DCIM) to service management.

Finally, this infrastructure revolution cannot happen without a rigorous commitment to sustainability. Powering the future of AI requires a strategy that goes beyond efficiency and targets decarbonisation and the circular economy. A proven framework for this journey involves three steps: Strategise, Digitise, and Decarbonise. Organisations must first establish a baseline, then leverage AI and digital tools to monitor energy use, and finally implement technologies to strip carbon from their operations.

This commitment extends to the lifecycle of the hardware itself. Goals such as recycling one metric tonne of product for every ten sold, and achieving 100% recycled packaging, are becoming industry benchmarks. Sustainability is not just about the product; it is about process efficiency and resource utilisation across the board.

The first step for any organisation is understanding its current state. Through initiatives like "Get Efficient" assessments, partners can help customers visualise and monetise the value of modernisation - calculating specific savings in power and cooling. The future of the data centre is dense, decentralised, and demanding. Whether through liquid cooling innovations, validated server architectures, or the convergence of IT and OT, the path forward requires a deep understanding of both the power that drives the chip and the ecosystem that supports it. The tools are available, and the demand is real - now is the time to build the infrastructure that powers the next generation of innovation.

Key takeaways for IT leaders:

• High-density is everywhere: With 50% of AI workloads moving to the edge by 2028, high-density infrastructure is no longer just for hyperscalers; it is a necessity for decentralised compute.
• The physics have changed: Success depends on efficient "grid-to-chip" power delivery and innovative heat extraction, requiring validated architectures that marry compute with cooling.
• The partner opportunity: Clients prefer incumbent partners, but partners must upskill in IT/OT convergence and offer specialised hosting for intensive workloads to bridge the readiness gap.
• Monetise efficiency: Use assessments to establish a baseline and quantify the monetary value of modernisation through power and cooling savings.
• Sustainability is holistic: Combine the "Strategise, Digitise, Decarbonise" framework with circular economy goals (recycling and packaging) to ensure long-term resilience.