Data Center Architecture

AI will drive 2x growth in data centre power capacity in 5 years. This means data centres must evolve. This week, I visited Barcelona to see how Schneider Electric assembles its prefabricated and containerised data centres at its factory. Also had the chance to attend briefings by senior members of its data centre division. Here are my thoughts. 𝟭/ 𝗔𝗜 𝘄𝗼𝗿𝗸𝗹𝗼𝗮𝗱𝘀 AI is here to stay. There are no pathways that don't include AI in some shape or form. But compute systems with lower power demands won't disappear either. This means our current way of designing data centres - often carving out a separate section for AI while non-AI workloads reside elsewhere, won't work. We need a new approach to cooling for maximum flexibility and sustainability for supporting both AI and non-AI workloads. - A new end-to-end design approach. - Partnerships for an AI-inclusive ecosystem. - New systems suited for this new paradigm. 𝟮/ 𝗗𝗶𝗱 𝘀𝗼𝗺𝗲𝗼𝗻𝗲 𝘀𝗮𝘆 𝗹𝗶𝗾𝘂𝗶𝗱 𝗰𝗼𝗼𝗹𝗶𝗻𝗴 Unsurprisingly, the new AI-centric data centres of the future must support liquid cooling. < 𝟰𝟬/𝗸𝗪 𝗿𝗮𝗰𝗸𝘀 - Liquid cooling offers better efficiency. ~ 𝟱𝟬/𝗸𝗪 𝗿𝗮𝗰𝗸𝘀 - Air cooling still possible, but barely. > 𝟱𝟬/𝗸𝗪 𝗿𝗮𝗰𝗸𝘀 - Liquid cooling is a must-have. I just wrote a post about why liquid cooling is the future of data centres yesterday. (Read: https://lnkd.in/g5jhCNcX) 𝟯/ 𝗡𝗼 𝘄𝗮𝗹𝗸 𝗶𝗻 𝘁𝗵𝗲 𝗽𝗮𝗿𝗸 But liquid cooling isn't trivial. - Local sustainability standards differ. - Not all liquid cooling solutions scale well. - Efficiency might come at the expense of other areas. Throw in the need for continued need for air-cooling, and it just gets... complicated. 𝟰/ 𝗠𝗮𝗸𝗶𝗻𝗴 𝗶𝘁 𝘄𝗼𝗿𝗸 What will the data centres of the future look like? The industry must come for a new generation of more sustainable, flexible, data centres. We need: - New data centre design and modeling software. - New high-capacity power trains, systems for AI. - Easy way to determine data centre efficiency. - Greater innovation across the ecosystem. And yes, Schneider Electric says it has developed a reference design for an AI-centric data centre with Nvidia - I'll share more about it in another post. 👉 Would love to hear your thoughts about how data centres must evolve. 𝗣𝗵𝗼𝘁𝗼: Schneider Electric's Sant Boi factory in Barcelona. --- My name is Paul Mah and I write about tech that matters in #EverydayTechStories 📆 Get weekly updates: www.techstories.co/updates 👀 See my other posts: www.techstories.co 🙋 Follow me on LinkedIn: https://lnkd.in/gu5EMKQg #datacentre

Over the past year I have had a growing number of conversations with ports, energy operators, universities and government teams about AI infrastructure. The discussion often ends up framed around hyperscale cloud: massive facilities, centralised compute and the assumption that AI capability must sit inside a handful of global platforms. That model makes sense for large AI training clusters. It is far less obvious that it is the right architecture for the systems that actually operate national infrastructure. Ports, hospitals, transport systems and energy networks increasingly rely on AI inference close to where decisions are made. It is in these critical infrastructure use cases that latency, resilience and governance begin to matter more than raw scale. This is where local, modular edge data centres are becoming strategically interesting. They allow compute capacity to be deployed close to industry, scaled incrementally and integrated with local energy infrastructure, including renewables. They can also be designed so that the hardware, networking and governance all remain within UK or European jurisdiction, removing hidden dependencies on US or Chinese infrastructure where that is a concern. Alongside these modular deployments, there is a growing role for high performance "edge" compute using on-premises workstations such as the Dell Precision T2, Lenovo ThinkStation P7 and AMD's Threadripper Pro based systems - which are increasingly capable of supporting inference workloads directly within organisations operational environments. This all shifts the architecture away from centralised dependency and towards locally controlled capability. In the short video below I walk through the practical differences between hyperscale and edge architectures and why inference workloads are changing how organisations think about AI infrastructure. The video runs just under six minutes and explains the architecture in straightforward terms. If you are involved in ports, energy systems, transport networks or other forms of critical infrastructure, this is a discussion that is appearing more frequently in strategy conversations. #AI #EdgeComputing #Blackwell #DellProPrecision #Nvidia #DellTech #DellPromaxGB10 #TechnicalWorkflows

🚀 AI's Impact on Data Centers: A Call for Modular Design The AI revolution, led by applications like ChatGPT, is reshaping the demands on data centers. These powerful tools require unprecedented levels of power, data, and bandwidth, challenging even modern facilities. 📈 Changing Power Dynamics: Just a year ago, 10-kilowatt racks were the norm. Now, we're looking at 25, 50, or even 100-kilowatt racks. This shift can strain traditional designs, affecting everything from performance to maintenance. 🌐 Bandwidth & Connectivity: High-density AI racks need robust network support. Without it, we risk inefficiencies and bottlenecks. ❄️ Cooling Concerns: As power distribution becomes uneven, our cooling systems face new challenges, leading to potential hot spots. ⚙️ The Modular Solution: The future of data centers is modular. This design offers the adaptability needed to meet changing demands, from network topology to airflow. It's the key to supporting AI's growing needs efficiently. In the AI era, adaptability is crucial. Modular data centers are our way forward, ensuring we're ready for the next wave of AI innovations. https://lnkd.in/gtHt8Mcn

Germany’s new national data centre strategy is out and it (finally) sends a clear signal to investors. The ambition is substantial: ➡️ Double data centre capacity by 2030 ➡️ Quadruple AI compute capacity From an international investor perspective, this is more than just policy, it is a statement of intent to remain a relevant location in an increasingly competitive global market. It has been recognized that additional data centres are needed to support Germany’s goals for data protection and digital sovereignty. But what does it actually mean for expansion and market entry into Germany? 1. Germany is open for data centre investment, but more structured than before. The strategy explicitly welcomes investment while aiming to strengthen local and European value creation. Expect a more guided approach to site selection, energy planning, and connectivity. 2. Power access becomes more “managed” New mechanisms such as staged connections and flexible grid agreements are likely to become standard. For developers, this means: 👉 Earlier engagement with TSOs/DSOs 👉 More realistic ramp-up scenarios 👉 Stronger linkage between project maturity and power reservation In short: speculative land banking without credible delivery pathways will become harder. 3. Site selection shifts beyond traditional hubs While Frankfurt remains constrained, the strategy clearly supports decentralised growth and the use of brownfield sites. 👉 New regions with available grid capacity will gain relevance 👉 Early-stage spatial planning becomes a competitive advantage 4. Permitting remains a bottleneck...for now. There is recognition that planning and permitting must accelerate, but concrete, binding timelines or standardised frameworks are still missing. For investors, this means continued reliance on: 👉 Local relationships 👉 Early stakeholder engagement 👉 Deep understanding of municipal dynamics 5. Community acceptance becomes a decisive factor One of the biggest practical risks today is not regulation, but local opposition. Projects increasingly succeed or fail at municipal level, regardless of technical suitability. For developers entering Germany, this changes the playbook: 👉 Transparent communication is no longer optional 👉 Local value creation (heat reuse, tax contribution, jobs) must be clearly articulated 👉 Municipal alignment needs to happen early, not during permitting 6. Energy costs and talent remain structural challenges The strategy acknowledges both, but does not yet provide fully actionable solutions. Given that energy can represent ~50% of OPEX, this remains a key factor in global capital allocation decisions. ---- Bottom line: Germany is positioning itself as a leading European data centre market, but success will depend less on ambition and more on execution. For investors, the opportunity is real. But so is the need for a more disciplined, locally embedded, and infrastructure-aware development approach.

Data centers were never designed for this. Enterprise facilities built for 5–10kW racks are hitting a hard limit. AI workloads now demand 50kW+ rack densities and that is forcing a total rethink of how we design, build, and operate data centers. What’s changing? Cooling: Air is no longer enough. Direct-to-chip and immersion liquid cooling are moving mainstream. Silicon: Hyperscalers are building their own chips (AWS Trainium, Google TPU, etc.) to optimize for AI. Networking: Old three-tier models are giving way to spine-leaf, 800G+, and SDN automation to handle east–west traffic. Power: AI demand could require the equivalent of 35 nuclear plants by 2030. Utilities and hyperscalers are joining forces to build new renewable capacity. My point of view: this represents a new era of infrastructure. Hyperscale campuses are becoming as critical as ports, airports, and power plants. Whoever masters this architecture will not only win in tech but also shape the backbone of the global economy. The real question: are we ready to balance this growth with sustainability, community impact, and energy realities? Google | DLVS consultancy ltd #DataCenters #Hyperscale #AIInfrastructure #CloudComputing #Google #LiquidCooling #SustainableTech #LabDesign

💡 How China Builds Its Cloud. By Design, Not by Market In most countries, cloud infrastructure grows where markets find cheap land, fast internet, and stable power. In China, it grows where policy decides. The Eastern Data, Western Computing (EDWC) initiative is a state-orchestrated plan to move data centers inland, shifting processing power from energy-hungry coastal cities to renewable-rich western provinces. 🔹 8 national computing hubs have been established, such as in Guizhou, Inner Mongolia, Gansu, Ningxia, and Qinghai, forming a vast “digital backbone” that supports Beijing, Shanghai, and Guangdong’s data demand. 🔹 US$6.1 billion has already been invested, with total project commitments topping US$28 billion. 🔹 1.95 million server racks have been installed, 63% already operational. They are policy-built ecosystems aligning computing geography with energy policy, data sovereignty, and regional development. Guizhou, once one of China’s poorest provinces, now hosts Apple’s iCloud operations through a state-partnered local firm, and houses Huawei, Tencent, Alibaba Group, and Baidu, Inc. data campuses. Some centers are even built into natural mountain caves to optimize cooling efficiency. This is “the cloud by design.” Infrastructure is a strategic instrument of the state. And it raises a bigger question for us in Malaysia. As AI and compute infrastructure become the new national assets, should we continue letting market forces alone decide where data, power, and compute grow or should we design them by intent? I share semiconductor insights everyday. Follow me 👉 Andrew Chan Yik Hong for actionable perspectives on policy, strategy & industry shifts and ring the bell 🔔 to get notified whenever I post. 💬 If this post resonates with you, re-post, drop a comment or leave a like. I’d love to hear your thoughts.

For years, data center conversations were anchored around uptime, connectivity, and expansion.   What’s becoming increasingly clear now is this: the conversation has moved upstream.   As AI workloads scale, infrastructure decisions are no longer being shaped at the deployment stage; they are being defined at the power and design layer itself.   Capacity planning is becoming fundamentally power-led, where grid readiness, energy strategy, and cooling architecture are starting to dictate what is actually possible to build and scale.   In my recent interaction with Voice&Data, I’ve shared a perspective on how this shift is turning data centers into engineered systems where long-term success will depend on how well we align power, cooling, and sustainability from the outset.   Read here → https://lnkd.in/dXFNs-DT   #AIInfrastructure #DataCenters #Energy #Sustainability #DigitalInfrastructure #TechnoDigital

🚀 AI is eating the world… and data centers are paying the price. 💡 Did you know that AI-driven data centers now consume more electricity than some entire countries? With demand skyrocketing, the infrastructure supporting AI is under massive strain—pushing energy grids to their limits and raising serious sustainability concerns. 🔌 The problem? AI models require enormous computational power—and that power has to come from somewhere. Data centers now use over 4% of U.S. electricity, with most of it still coming from fossil fuels. The race for bigger, faster AI is driving exponential demand, forcing grid expansions and supply chain bottlenecks. But here’s the twist… AI can actually help solve this problem (surprise surprise)! 🔍 AI can transform the way we DESIGN data centers: ✅ Automating site selection – AI can identify the best locations based on environmental, geotechnical, and grid availability data. ⚡ Grid infrastructure analysis – AI can predict where power demand will surge and help plan the necessary upgrades. 📐 Generative design for optimal layouts – AI can maximize space efficiency, optimize building layouts, cooling strategies, and energy usage. 🔥 Surrogate models for faster simulations – Instead of waiting hours for thermal, structural, and seismic analysis of data center buildings, AI can generate results in minutes. 🚚 Supply chain optimization – AI can streamline material sourcing and procurement, reducing construction delays and emissions. 📢 The takeaway? AI is both the challenge and the solution. To sustain the next wave of AI innovation, we must rethink how we design, build, and operate data centers. What’s your take? Should AI companies be more accountable for their energy footprint? Or is the race for intelligence too fast to slow down? Drop your thoughts in the comments! 👇 Sources: https://lnkd.in/dPZ46-Tu #ArtificialIntelligence #DataCenters #Sustainability #AEC #GenerativeDesign

India’s #datacentre sector is clearly entering a more serious phase of policy and infrastructure support. In my earlier post, I highlighted how the discussion around nuclear power privatization for data center operators was an important signal of India’s long-term intent to enable energy-secure digital infrastructure. Now, Andhra Pradesh’s reported #DISCOM licence model for large data centres takes that conversation a step further. This is significant because it shows that India is beginning to address one of the sector’s biggest constraints more directly: reliable, scalable, and flexible power delivery. For years, data centre growth discussions have focused on land, connectivity, and incentives. Those remain important. But as the market scales, especially with the rise of AI workloads, higher rack densities, and larger hyperscale campuses, power architecture is becoming the defining variable. That is why this development matters. It suggests a broader shift in thinking: ⚡️from viewing data centres as conventional real estate, ⚡️to recognizing them as strategic digital infrastructure; ⚡️from generic industrial policy, ⚡️to more targeted enablement around energy access and distribution. When seen alongside the earlier debate on private participation in future energy supply models, including nuclear, the direction is becoming clearer: India wants to compete seriously for long-term data centre and AI infrastructure investment. The next wave of growth in this sector will not be determined by demand alone. It will be determined by which regions can offer: - dependable power, - scalable distribution frameworks, - cleaner energy pathways, - and policy confidence for large capital deployment. This is why moves like this deserve attention. They are not isolated policy decisions. They are early markers of a more mature infrastructure strategy for India’s digital economy. #DataCenters #DataCentre #DigitalInfrastructure #India #AndhraPradesh #PowerInfrastructure #PowerDistribution #EnergyTransition #RenewableEnergy #CleanEnergy #NuclearEnergy #Infrastructure #Investment #Hyperscale #Colocation #CloudInfrastructure #AI #ArtificialIntelligence #GenerativeAI #AIInfrastructure #DigitalEconomy #Sustainability #EnergySecurity #DataCenterIndia #TechInfrastructure

When leaders think about reorganising IT, the debate often collapses into a false binary: centralise or decentralise. In reality, modern states and complex organisations cannot be run as either a fully centralised command economy or a loose federation of independent units. Both extremes fail, just in different ways. What matters is clear role allocation, shared foundations, and disciplined governance, while preserving enough autonomy where it actually creates value. Yesterday, we presented Estonia’s ICT consolidation plan. A necessary next step in maturing our digital state and ensuring that we build less duplication, more capability, and have a stronger strategic focus. The core pillars are straightforward, but ambitious: 1) Consolidation of state IT houses and in-house IT units. Including infrastructure consolidation, to reduce fragmentation, improve resilience, and free up capacity for higher-value work. 2) Clearer public-private roles. With the gradual transfer of product development to the private sector, while the state focuses on governance, standards, platforms, and long-term direction. 3) More core tools and services. Public-sector digital solutions developed once and reused across sectors and ministries - similar to eesti.ee, the state app, abd consent service. 4) Step-by-step implementation of a nationwide ICT governance framework, aimed at: - strategic management of IT costs and investments - guiding the creation and rollout of central solutions - setting clear, cross-government priorities As a concrete next step, we are going to present the detailed IT house consolidation plan by June 2026. This is not about centralisation for its own sake. It is about building a state that can govern complexity, scale digital services, and remain sustainable in the AI era. More about the news: https://lnkd.in/d9hC-rDM