Nuclear Engineering Plant Design

AWS Builds Custom Liquid Cooling System for Data Centers Amazon Web Services (AWS) is sharing details of a new liquid cooling system to support high-density AI infrastructure in its data centers, including custom designs for a coolant distribution unit and an engineered fluid. “We've crossed a threshold where it becomes more economical to use liquid cooling to extract the heat,” said Dave Klusas, AWS’s senior manager of data center cooling systems, in a blog post. The AWS team considered multiple vendor liquid cooling solutions, but found none met its needs and began designing a completely custom system, which was delivered in 11 months, the company said. The direct-to-chip solution uses a cold plate placed directly on top of the chip. The coolant, a fluid specifically engineered by AWS, runs in tubes through the sealed cold plate, absorbing the heat and carrying it out of the server rack to a heat rejection system, and then back to the cold plates. It’s a closed loop system, meaning the liquid continuously recirculates without increasing the data center’s water consumption. AWS also developed a custom coolant distribution unit, which it said is more powerful and more efficient than its off-the-shelf competitors. “We invented that specifically for our needs,” Klusas says. “By focusing specifically on our problem, we were able to optimize for lower cost, greater efficiency, and higher capacity.” Klusas said the liquid is typically at “hot tub” temperatures for improved efficiency. AWS has shared details of its process, including photos: https://lnkd.in/e-D4HvcK

𝐇𝐨𝐭 𝐚𝐧𝐝 𝐂𝐨𝐥𝐝 𝐀𝐢𝐬𝐥𝐞 𝐂𝐨𝐧𝐭𝐚𝐢𝐧𝐦𝐞𝐧𝐭: 𝐎𝐩𝐭𝐢𝐦𝐢𝐳𝐢𝐧𝐠 𝐃𝐚𝐭𝐚 𝐂𝐞𝐧𝐭𝐞𝐫 𝐂𝐨𝐨𝐥𝐢𝐧𝐠 Temperature and humidity control are critical to protecting IT infrastructure, reducing failures, and extending equipment lifespan. One of the most effective strategies? Aisle Containment. 🔹 𝐖𝐡𝐚𝐭 𝐢𝐬 𝐀𝐢𝐬𝐥𝐞 𝐂𝐨𝐧𝐭𝐚𝐢𝐧𝐦𝐞𝐧𝐭? It’s a smart arrangement of server racks that physically separates hot (exhaust) and cold (supply) air. This simple yet powerful setup minimizes air mixing, allowing operators to raise supply temperatures or reduce fan speeds improving energy efficiency and reducing operational costs. 🔹 𝐇𝐨𝐭 𝐯𝐬. 𝐂𝐨𝐥𝐝 𝐀𝐢𝐬𝐥𝐞 𝐂𝐨𝐧𝐭𝐚𝐢𝐧𝐦𝐞𝐧𝐭 Both strategies aim to prevent the mixing of hot and cold air, but they do it differently: 🔸 𝐇𝐨𝐭 𝐀𝐢𝐬𝐥𝐞 𝐂𝐨𝐧𝐭𝐚𝐢𝐧𝐦𝐞𝐧𝐭 (𝐇𝐀𝐂): In this system, server racks are arranged so that the fronts face the cold aisle, and hot air is exhausted into a sealed hot aisle or plenum. The hot air is then removed by cooling units. This setup allows for higher server density without increasing cooling costs, as heat is prevented from migrating into other areas. Hot aisle containment also improves airflow uniformity and reduces the risk of hotspots 🔸 𝗖𝗼𝗹𝗱 𝗔𝗶𝘀𝗹𝗲 𝗖𝗼𝗻𝘁𝗮𝗶𝗻𝗺𝗲𝗻𝘁 (𝗖𝗔𝗖): Here, racks are arranged with their fronts facing a contained cold aisle, where cold air is supplied through perforated tiles or doors. The cold air passes through the equipment and is exhausted into a return plenum. Cold aisle containment can improve cooling efficiency by up to 30% by minimizing the mixing of cold and warm air, ensuring a steady supply of cold air to equipment. 🔹 𝗪𝗵𝗶𝗰𝗵 𝗖𝗼𝗻𝘁𝗮𝗶𝗻𝗺𝗲𝗻𝘁 𝗦𝘁𝗿𝗮𝘁𝗲𝗴𝘆 𝗶𝘀 𝗥𝗶𝗴𝗵𝘁 𝗳𝗼𝗿 𝗬𝗼𝘂? Choosing between hot and cold aisle containment depends on: 👉 Space & Operations: Hot aisle containment suits larger, custom built data centers with dedicated on-site staff. Cold aisle containment is often more predictable and suitable for colocation environments. 👉 Budget: Cold aisle systems are more cost-effective and easier to retrofit, while hot aisle systems may have higher upfront costs but offer better long-term savings. 👉Layout Flexibility: Existing facility layouts may restrict containment choices, while new builds provide greater design freedom. 👉Equipment & Climate: High-density or heat-intensive equipment often benefits from HAC, especially in warmer climates. CAC may be more practical in moderate or cooler regions. ✅ Conclusion Both hot and cold aisle containment are proven methods for managing airflow and maintaining optimal environmental conditions in data centers. Your choice should reflect the size, layout, workload, climate, and budget of your facility. Sources are listed in the first comment.

How Full Liquid Cooling Is Powering the Next Generation of AI Data Centers.... . . As AI workloads grow, traditional cooling methods are no longer enough. Modern high-performance data centers are now built around full liquid cooling architectures designed to manage the extreme heat generated by advanced AI processors. At the facility level, water from the building cooling system flows into in-row Coolant Distribution Units (CDUs). Inside, a liquid-to-liquid heat exchanger transfers cooling capacity to a secondary fluid that circulates directly to each rack, creating an efficient bridge between facility cooling and IT equipment. Inside every server, a dedicated liquid loop is engineered to match the processor layout and power density of AI hardware. Instead of relying on air, this loop absorbs heat directly from CPUs, GPUs, and memory modules, removing thermal energy at the source. The heated liquid then returns to the CDU, where high-performance heat exchangers move the heat away from the IT space toward the facility cooling system. From there, rooftop chillers or dry coolers reject the heat into the ambient environment. Even in fully liquid-cooled data centers, air still plays a supporting role. Air handlers remove residual heat from components not connected to the liquid loop, creating a balanced ecosystem where liquid handles high-density loads and air maintains room stability. Full liquid cooling is becoming a foundation for AI-ready infrastructure, enabling higher rack densities, better efficiency, and stable performance under extreme compute demand. As a Data Center Operations & Maintenance Engineer, I closely follow how these cooling architectures are transforming operations and facility design. Always happy to connect with professionals working on next-generation, AI-ready data centers. Video copyright: BOYD © Abdullah Mahrous – CC BY 4.0

Mastering Data Center HVAC: An MEP Deep Dive ❄️💡 Data centers are the digital world’s engine rooms — and their lifeline depends on robust MEP systems, especially advanced HVAC design. Cooling isn’t just about comfort — it’s about reliability, efficiency, and scalability in an always-on world. Let’s explore the essentials every modern MEP engineer must master: ⸻ 🔹 1️⃣ Nailing the Fundamentals: 🔢 Precision Load Calculation: Accurately quantify heat loads from IT equipment (kW or BTU/hr). Rule of thumb: Electrical input ≈ Thermal output. Even small miscalculations can cascade into costly over/under-sizing errors. 🛡️ Design for Redundancy: High uptime demands high resilience. Deploy N+1, 2N, or even 2(N+1) cooling architectures to safeguard operations. Example: 2N cooling ensures full capacity even during maintenance or failures. 🔹 2️⃣ Mastering Airflow Management: ↔️ Hot/Cold Aisle Best Practices: Separating intake and exhaust airflow is foundational — improving efficiency by up to 30%. 🔒 Containment Systems: Elevate performance with Hot or Cold Aisle Containment. Energy tip: Well-sealed containment can reduce fan energy by 10–20% and cut cooling costs significantly. 🌬️ Optimized Distribution Methods: Raised floor, overhead ducting, or hybrid systems — the choice depends on space, density, and maintenance access. ✨ AI-Enhanced Airflow: Leading-edge data centers now leverage AI-based airflow control — dynamically adjusting cooling in real time based on server load profiles. ⸻ 🔹 3️⃣ Choosing the Right Cooling Technologies: ❄️ CRAC vs. CRAH: Understand the core difference: CRAC (direct expansion) vs. CRAH (chilled water). CRAHs generally offer higher efficiency when paired with centralized chiller plants. 🎯 Precision Cooling Innovations: In-Row Cooling, Rear-Door Heat Exchangers, and Close-Coupled Cooling are ideal for high-density zones (>20-30 kW per rack). 💧 Liquid Cooling Revolution: Adoption of direct-to-chip and immersion cooling is rising rapidly, especially in AI/HPC (High-Performance Computing) clusters. By 2025, up to 20% of new data centers are projected to deploy some form of liquid cooling. ⸻ 🔹 4️⃣ Driving Efficiency and Standards Compliance: ⚙️ PUE (Power Usage Effectiveness): Strive for a PUE close to 1.2 or better. Every 0.1 reduction slashes operational costs and carbon footprint dramatically. 🌍 Harness Free Cooling: Utilize air-side or water-side economizers when ambient conditions allow. Example: In cool climates, economizers can reduce chiller runtime by over 50% annually. 💧 Precise Humidity Control: Maintain RH between 40%–60% (per ASHRAE 2021 Thermal Guidelines) to mitigate risks of static buildup and condensation. 📊 Integrated Monitoring: Deploy advanced DCIM (Data Center Infrastructure Management) and BMS (Building Management Systems) platforms. Real-time visibility + AI analytics = proactive optimization.

Evolutionary Structural Optimisation (ESO) is an optimisation process where small pieces of a structural zone are nibbled away iteratively, reanalysing between each nibble, to reveal an optimum stress path structure. It forms complex shapes which would be ideal for manufacture using 3d printing/ additive manufacture. It is also possible to enforce certain rules on how the material can be nibbled away. For example, if you want to cast the shape and pull it out of a mould then you want to enforce mould pull out rules such that only elements at the mould surface can be removed, without removing elements within the material leaving voids. It is also possible to carry out natural frequency, linear buckling, and deflection checks with each ESO iteration for multiple loading cases concurrently. The two videos attached show an example ESO solution for a corbel zone around a column of a floor plate. The light blue vertical downward loads at the corners are from the infill floor panel (not shown). The column and top of floor slab are in green and not erodible. The blue volume is able to be nibbled away. One video shows the ESO solution where there are no rules other than stress criteria. The other shows the ESO solution when mould pull out rules are applied with pull out in the vertical upward direction. Possible usages for this could be for bespoke connections, precast pad foundations, or even as an organic scheming tool for bridge structures. #EvolutionaryStructuralOptimisation #StructuralEngineering #StructuralAnalysis #StructuralDesign #FiniteElementAnalysis #Optimisation #StructuralOptimisation #AdditiveManufacture

𝗙𝗿𝗼𝗺 𝗦𝘁𝗿𝗮𝗶𝗴𝗵𝘁 𝗙𝗶𝗻𝘀 𝘁𝗼 𝗔𝗜-𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗲𝗱 𝗚𝗲𝗼𝗺𝗲𝘁𝗿𝘆: The illustration below highlights three generations of thermal optimization: (𝟭) 𝟭𝗗 𝗢𝗽𝘁𝗶𝗺𝗶𝘇𝗮𝘁𝗶𝗼𝗻 – 𝗗𝗶𝗺𝗲𝗻𝘀𝗶𝗼𝗻 𝗦𝗰𝗮𝗹𝗶𝗻𝗴 Engineers optimize basic parameters such as: • Fin height • Fin thickness • Fin spacing • Fin length 𝗔𝗱𝘃𝗮𝗻𝘁𝗮𝗴𝗲𝘀 • Simple design and manufacturing • Low cost • Suitable for conventional air-cooled heat sinks 𝗟𝗶𝗺𝗶𝘁𝗮𝘁𝗶𝗼𝗻𝘀 • Limited design freedom • Lower heat transfer efficiency • Higher thermal resistance at extreme heat fluxes    (𝟮) 𝟮𝗗 𝗢𝗽𝘁𝗶𝗺𝗶𝘇𝗮𝘁𝗶𝗼𝗻 – 𝗙𝗶𝗻 𝗚𝗲𝗼𝗺𝗲𝘁𝗿𝘆 Instead of only changing dimensions, engineers optimize fin shapes using: • Pin fins • Offset fins • Louver fins • Wavy fins This increases: • Surface area • Turbulence • Coolant mixing • Heat transfer coefficient The result is significantly improved cooling performance while still using conventional manufacturing methods. (𝟯) 𝗧𝗼𝗽𝗼𝗹𝗼𝗴𝘆 𝗢𝗽𝘁𝗶𝗺𝗶𝘇𝗮𝘁𝗶𝗼𝗻 – 𝗠𝗮𝘁𝗲𝗿𝗶𝗮𝗹 𝗗𝗶𝘀𝘁𝗿𝗶𝗯𝘂𝘁𝗶𝗼𝗻 This represents the next generation of thermal engineering. Rather than designing individual fins, optimization algorithms determine 𝘄𝗵𝗲𝗿𝗲 𝗺𝗮𝘁𝗲𝗿𝗶𝗮𝗹 𝘀𝗵𝗼𝘂𝗹𝗱 𝗲𝘅𝗶𝘀𝘁 𝗮𝗻𝗱 𝘄𝗵𝗲𝗿𝗲 𝗶𝘁 𝘀𝗵𝗼𝘂𝗹𝗱 𝗯𝗲 𝗿𝗲𝗺𝗼𝘃𝗲𝗱 to maximize thermal performance while minimizing pressure drop and weight. Using 𝗖𝗙𝗗, 𝗙𝗗𝗔, 𝗙𝗘𝗔, 𝗮𝗻𝗱 𝗴𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝘃𝗲 𝗱𝗲𝘀𝗶𝗴𝗻, engineers create complex organic structures that would be nearly impossible to design manually. 𝗞𝗲𝘆 𝗯𝗲𝗻𝗲𝗳𝗶𝘁𝘀 • Maximum heat transfer per unit volume • Lower junction temperatures • Uniform coolant distribution • Reduced pumping power • Lightweight, high-strength structures • Optimized pressure drop vs. thermal performance These geometries are typically manufactured using 𝗺𝗲𝘁𝗮𝗹 𝗮𝗱𝗱𝗶𝘁𝗶𝘃𝗲 𝗺𝗮𝗻𝘂𝗳𝗮𝗰𝘁𝘂𝗿𝗶𝗻𝗴 (𝗦𝗟𝗠/𝗗𝗠𝗟𝗦), enabling cooling solutions beyond the capabilities of traditional machining. 𝗪𝗵𝘆 𝗶𝘁 𝗺𝗮𝘁𝘁𝗲𝗿𝘀 Thermal management has become one of the primary constraints on AI infrastructure performance. As chip power densities continue to rise, the industry is moving beyond incremental fin optimization toward 𝗮𝗹𝗴𝗼𝗿𝗶𝘁𝗵𝗺-𝗱𝗿𝗶𝘃𝗲𝗻 𝘁𝗵𝗲𝗿𝗺𝗮𝗹 𝗮𝗿𝗰𝗵𝗶𝘁𝗲𝗰𝘁𝘂𝗿𝗲𝘀. The future of AI cooling will be designed not only by engineers—but also by optimization algorithms that simultaneously solve for 𝗵𝗲𝗮𝘁 𝘁𝗿𝗮𝗻𝘀𝗳𝗲𝗿, 𝗳𝗹𝘂𝗶𝗱 𝗱𝘆𝗻𝗮𝗺𝗶𝗰𝘀, 𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗮𝗹 𝗶𝗻𝘁𝗲𝗴𝗿𝗶𝘁𝘆, 𝗺𝗮𝗻𝘂𝗳𝗮𝗰𝘁𝘂𝗿𝗮𝗯𝗶𝗹𝗶𝘁𝘆, 𝗮𝗻𝗱 𝗰𝗼𝘀𝘁. #𝗧𝗵𝗲𝗿𝗺𝗮𝗹𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 #𝗗𝗮𝘁𝗮𝗖𝗲𝗻𝘁𝗲𝗿 #𝗔𝗜𝗖𝗼𝗼𝗹𝗶𝗻𝗴 #𝗟𝗶𝗾𝘂𝗶𝗱𝗖𝗼𝗼𝗹𝗶𝗻𝗴 #𝗖𝗼𝗹𝗱𝗣𝗹𝗮𝘁𝗲𝘀 #𝗛𝗲𝗮𝘁𝗧𝗿𝗮𝗻𝘀𝗳𝗲𝗿 #𝗖𝗙𝗗 #𝗙𝗘𝗔 #𝗧𝗼𝗽𝗼𝗹𝗼𝗴𝘆𝗢𝗽𝘁𝗶𝗺𝗶𝘇𝗮𝘁𝗶𝗼𝗻 #𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝘃𝗲𝗗𝗲𝘀𝗶𝗴𝗻 #𝗔𝗱𝗱𝗶𝘁𝗶𝘃𝗲𝗠𝗮𝗻𝘂𝗳𝗮𝗰𝘁𝘂𝗿𝗶𝗻𝗴 #𝗛𝘆𝗽𝗲𝗿𝘀𝗰𝗮𝗹𝗲 #𝗛𝗣𝗖 #EngineeringInnovation

Hyperscale Data Centre Airflow & Heat Transfer Engineering This engineering drawing presents an optimized airflow and thermal management concept for a hyperscale data centre. The design illustrates hot aisle and cold aisle containment, raised-floor air distribution, ceiling return air, CFD airflow patterns, server heat transfer, temperature gradients, pressure management, and psychrometric principles. Developed through multiple engineering review and optimization cycles, the final design enhances reliability, energy efficiency, maintainability, scalability, and compliance with industry best practices, including ASHRAE thermal guidelines.

⚙️ Pressure Differential in Data Centers: The Invisible Force Behind Airflow Stability 🌬️ Why It Matters In data centers, air doesn’t just move — it’s driven by pressure differentials. Without proper balance between hot and cold aisles, you risk: Air recirculation 🔁 Hot spots near racks 🌡️ Higher fan energy and poor cooling efficiency ⚡ Loss of containment performance 🧩 Ideal Pressure Ranges Server Room to Corridor: +2 to +5 Pa (positive pressure keeps dust out). Cold Aisle Containment: Slightly positive vs hot aisle (for directed flow). Raised Floor Plenum: Maintained by CRAH supply to ensure uniform airflow. Too low? Air short-circuits. Too high? Tiles lift or doors slam — both are warning signs ⚠️ 🧠 How It’s Controlled VFD Fans: Adjust automatically via static pressure sensors. BMS Feedback Loops: Maintain target ΔP using real-time monitoring. Containment Design: Defines the direction and strength of airflow. 🔍 Common Mistakes ❌ Ignoring sensor calibration. ❌ Poor CRAH balancing leading to uneven airflow. ❌ No pressure zoning in large halls. 💡 Pro Tip Always validate pressure differentials during commissioning with calibrated manometers — not just the BMS readings. Even a 2 Pa difference can make or break your airflow containment strategy. #DataCenter #MEP #HVAC #AirflowManagement #FacilityManagement #Commissioning #PressureDifferential #MechanicalEngineering #CoolingOptimization #BMS #EnergyEfficiency

Cooling the Kilowatt Chip: The Next Architecture Constraint The next generation of AI chips will not be limited only by transistor density, memory bandwidth or interconnect speed. Increasingly, it will be limited by how effectively we extract heat from dense, heterogeneous packages with highly localised hotspots. We are moving from “cooling the server” to thermal co-design across chip, package, rack and facility. Beyond TDP A 1.4 kW chip is challenging, but total power is only part of the story. The harder questions are where the heat is generated, how concentrated it is, how quickly it fluctuates, and how far the coolant is from the junction. W/chip is not W/cm². A lab-scale heat-flux result is not a bankable rack-level solution. Cooling technologies must be compared under common boundary conditions: junction temperature, coolant inlet temperature, pressure drop, pump power, reliability, serviceability and facility impact. No Single Winner The future will be a layered thermal stack: reduce heat at source, spread heat near the junction, minimise TIM and package resistance, extract heat using advanced single-phase or two-phase liquid cooling, buffer transients, and orchestrate flow, chemistry and heat rejection intelligently. Enhanced direct-to-chip liquid cooling will remain the near-term workhorse. Two-phase DTC will grow as socket powers move beyond the hydraulic comfort zone of single-phase loops. Immersion will be valuable where whole-server heat capture and fan removal matter. Embedded microfluidics and direct-to-silicon cooling may become the long-term frontier for 3D-native compute. The Tropical Dimension For tropical data centres, high ambient temperature, humidity, condensation risk, water stress and serviceability constraints change the optimum. The best solution is not necessarily the one that gives the lowest chip temperature. It is the one that safely maintains junction temperature while maximising coolant temperatures, reducing compressor lift, minimising water use and preserving maintainability. That means warm-water liquid cooling, robust CDU isolation, fluid-chemistry control, leak blast-radius design, and metrics beyond PUE: WUE, carbon intensity, heat-reuse potential and compute-energy productivity. Cooling as Compute Strategy Bring the coolant closer to the heat. Reduce every avoidable thermal resistance. Use phase change only where instability can be controlled. Treat coolant chemistry as a reliability discipline. Optimise chip, package, rack and facility as one coupled thermal system. Cooling is no longer infrastructure support. It is becoming a determinant of compute density, silicon reliability, data-centre bankability and sustainability performance. The organisations that understand this first will shape the architecture of AI infrastructure itself. #AI #DataCentres #LiquidCooling #ThermalManagement #Semiconductors #AdvancedPackaging #Sustainability #GreenDataCentres #MechanicalEngineering #EnergyEfficiency

Breaking the thermal wall with material innovation Performance is now limited by heat as much as logic. Beating the thermal wall demands a materials‑first approach paired with tight electro‑thermo‑mechanical co‑design. What moves the needle - Next‑gen TIMs: liquid‑metal gallium alloys for ultra‑low interface resistance; sintered silver for near‑bulk conductivity and high‑temp stability; phase‑change and graphene/graphite‑enhanced TIMs for thin, reliable bond lines. - Heat spreading ultrathin vapor chambers, pyrolytic graphite sheets, and composite lids (e.g., Cu‑diamond) to flatten hot spots before the sink. - Microchannel cooling: single‑phase cold plates for hundreds of W/cm² with modest ΔP; two‑phase and jet impingement for the highest flux; additive‑manufactured manifolds and fins to unlock flow and surface area. - Package co‑design: direct‑to‑die cooling, embedded spreaders, and low‑CTE, high‑k substrates to manage both heat and warpage. From concept to production - Engineer the interface: flatness, roughness, bondline control, and clamp load dominate real‑world Rθ. - Prove reliability: resist pump‑out, dry‑out, creep, and galvanic effects; ensure coolant/material compatibility. - Model and measure: disciplined compact models and standardized test methods keep simulations honest. How we can help We combine materials science with system co‑design to turn thermal limits into headroom. We have all the Credence design tools and can help with thermal management using the best TIMs and microchannel solutions for your challenging application. Share your power map, allowable pressure drop, and constraints—we’ll deliver a material stack and cooling architecture with modeled junction temps, flow/pressure requirements, and a clear reliability plan.