Data Center

Cadence and Microsoft Present New Insights on Data Center CFD Modeling at ITherm

Corporate — Tue, 19 May 2026 16:00:00 GMT

As AI workloads continue to drive unprecedented rack power densities, the limits of traditional air cooling are becoming increasingly visible. At IEEE ITherm Conference (Orlando, FL – May 26-29), Cadence and Microsoft will jointly present new ...(read more)

Physics Underpinning Decisions: Simulation‑Trained AI Optimizes Tokens per Watt

Corporate — Wed, 15 Apr 2026 17:30:00 GMT

The rapid rise of AI factories is pushing data center infrastructure beyond the limits of traditional planning. Rack densities have increased by an order of magnitude, liquid cooling is becoming standard, and AI workloads no longer behave predictabl...(read more)

Industry’s First UALink-200G Controller and PHY Running in 3nm!

HW202512191014 — Mon, 13 Apr 2026 16:00:00 GMT

AI systems are running into a familiar problem. Compute keeps scaling, but the infrastructure connecting that compute is starting to dominate system behavior. Interconnect latency, bandwidth efficiency, and power now have as much impact on performanc...(read more)

Scale Up vs. Scale Out in Modern AI Factories

HW202512191014 — Sat, 11 Apr 2026 00:30:00 GMT

Choosing pod fabrics, planning bisection bandwidth, and managing ordering semantics Two Worlds Inside the AI Factory As AI factories scale into tens of thousands of accelerators, architects must navigate two very different networking worlds. Inside a...(read more)

Accelerating the AI Factory: Switch and Cadence Redefine High-Density Design

Corporate — Tue, 17 Mar 2026 16:00:00 GMT

"We are redefining what is possible for next-gen AI factories with our patent-pending EVO Chamber solution—delivering up to 2MW per cabinet through advanced hybrid cooling in a modular, future-proof design. Using the Cadence Reality Digital Twi...(read more)

Digital Twins Enable the Next Era of AI Infrastructure

Corporate — Mon, 16 Mar 2026 20:30:00 GMT

Artificial intelligence (AI) is reshaping the data center. As AI workloads scale in size and complexity, traditional hyperscale designs are giving way to AI factories—purpose-built environments engineered to manufacture intelligence efficiently, reliably, and at scale.

In an AI factory, infrastructure performance is no longer measured solely by availability or power efficiency. Instead, success is defined by tokens generated per watt, workload throughput, and the ability to rapidly deploy and operate next-generation accelerated computing platforms. Meeting these demands requires a new approach to infrastructure design and operations—one grounded in system-level understanding, workload awareness, and continuous optimization.

This is where digital twins play a foundational role, and where Cadence is working closely with NVIDIA and industry leaders to enable the next generation of AI infrastructure.

From Hyperscale Data Centers to AI Factories

AI factories represent an evolution of the hyperscale model. While hyperscale data centers are designed to support a broad mix of workloads, AI factories are optimized specifically for AI training with an increasing emphasis on inference environments. These workloads introduce high rack-power densities of 100KW+ per rack, higher thermal loads, extreme AI workload variations, and critical sensitivity to network topology and latency in order to deliver the best performance per watt.

As a result, infrastructure decisions can no longer be made in isolation. Power delivery, cooling architecture (such as liquid-cooling and hybrid-cooling), workload placement, and network design are deeply interconnected. Campus-level choices—such as building layout, cooling plant strategy, and fiber routing—directly affect AI performance, efficiency, and scalability.

Designing AI factories, therefore, requires holistic, workload-specific optimization, validated before deployment, and continuously refined during operation.

Cadence and NVIDIA: Enabling AI Factory Design at Scale

As part of the effort to address these challenges, Cadence has collaborated with NVIDIA to deliver a Cadence Reality DC Elements model of NVIDIA's latest high-performance accelerated computing platform, the NVIDIA GB300 NVL72 system, for the Cadence Reality Digital Twin Platform.

This Cadence Reality DC Elements model is integrated with the NVIDIA Omniverse DSX Blueprint for AI factory digital twins and is available as a SimReady model for use via NVIDIA Omniverse libraries through the Cadence Reality DT Experience. SimReady assets provide use-case-specific technical payloads, from high-fidelity visualization and BOM data to lightweight behavioral models for rapid simulation. It enables data center designers and operators to accurately model, simulate, and optimize AI factory designs and changes during operations as part of performance-aware lifecycle management—reducing uncertainty, accelerating design cycles, and enabling confident decision-making.

By providing validated, high-fidelity infrastructure models, Cadence helps organizations design AI factories that are ready for today's workloads while remaining adaptable to future generations of accelerated computing.

Digital Twins Across the Infrastructure Lifecycle

Leading AI factory designers and operators rely on the Cadence Reality Digital Twin Platform to support both design and operational optimization. Unlike static planning tools, digital twins create a continuous feedback loop between design intent and real-world operation.

Key capabilities include:

Operational digital twins delivered by the Cadence Reality Digital Twin Platform. Visualization and cross-team collaboration of operational digital twins is enhanced by Cadence Reality DT Experience, powered by NVIDIA Omniverse.
Cadence Reality DC Elements models now integrated in the NVIDIA Omniverse DSX Blueprint.
AI surrogate models that accelerate simulation and optimization, allowing teams to explore more scenarios and tradeoffs much faster, typically in minutes.
High‑fidelity AI server and CDU models that speed end‑to‑end system design, shortening the time to deployment while optimizing AI factory architectures.

Together, these capabilities enable engineers to move beyond conservative margins and instead operate infrastructure at validated optimal points, harmonizing performance, energy efficiency, and reliability to discover new revenue opportunities.

Optimizing Infrastructure for AI Performance

In AI factories, performance is measured by the number of tokens generated per second and efficiency by tokens per watt. The Cadence Reality Digital Twin Platform enables designers and operators to optimize infrastructure directly against these metrics by simulating real AI workload behavior across power, cooling, and networking domains.

Using AI-accelerated simulations, teams can:

Maximize token throughput by operating hardware at the most efficient tokens-per-watt point. Digital twin analysis validates running more GPUs at lower power (MaxQ), increasing token generation by up to 30% while improving overall energy efficiency (Q) and improving tokens per watt by 17%.
Minimize cooling energy to unlock more power for compute. Optimized liquid cooling strategies, airflow, 45C inlet cooling temperatures, and thermal setpoints reduce cooling overhead and free additional power capacity for NVIDIA AI infrastructure.

By enabling high-fidelity simulation of AI workload scenarios, these capabilities accelerate gigawatt-scale AI factory buildouts and help unlock billions of dollars in potential revenue by maximizing performance while controlling energy and infrastructure costs. These results highlight the importance of treating infrastructure as a single, integrated system, validated in a high-fidelity digital twin before deployment.

Momentum Across the AI Ecosystem

Digital twins are rapidly becoming foundational to AI factory design and operations across the ecosystem:

NVIDIA is collaborating with Cadence on Reality DC Element models for next-generation Vera Rubin technology.
NV5 is relying on the Cadence Reality DC Design software to create digital twins that future-proof AI data centers, optimize infrastructure efficiency, and reliability. These workflows have been applied at scale across environments powered by thousands of NVIDIA Grace Blackwell GPUs, such as NVIDIA GB200, where engineering-grade, CFD-driven simulation is critical to identifying and minimizing operational risks before deployment.

Together, these examples reflect a growing industry consensus: Digital twins are essential infrastructure for AI factories.

Engineering the Future of AI Infrastructure

As AI continues to scale, the challenge is no longer simply delivering more compute—it is to efficiently convert power into intelligence. AI factories demand a new engineering mindset in which infrastructure is designed, optimized, and operated as an integrated system. With the Cadence Reality Digital Twin Platform, now integrated into the NVIDIA Omniverse DSX Blueprint, including DSX SimReady assets and AI-accelerated simulation capabilities, Cadence is helping customers design and operate AI factories with greater efficiency, speed, and confidence. The future of AI will be shaped not only by algorithms and silicon, but by infrastructure engineered to manufacture intelligence at scale.

Learn more about Cadence's partnership with NVIDIA and the Cadence Reality Digital Twin Platform.

Designing AI Factories with Digital Twins

Reela Samuel — Tue, 03 Mar 2026 04:32:00 GMT

Engineering AI infrastructure as a performance system The role of the data center is changing. What was once built to run applications is now engineered to operate as an AI factory. An AI factory is a purpose-built computing environment designed to ...(read more)

How 224G SerDes Unifies Today’s AI Fabrics

HW202512191014 — Sat, 28 Feb 2026 00:00:00 GMT

AI system architects no longer face a binary choice between scale-up and scale-out fabrics. Modern AI platforms require both to be executed efficiently, predictably, and within aggressive schedules. The shift lies in how risk is managed across these ...(read more)

Edge and Micro Data Centers: Powering the Real-Time Digital World

Reela Samuel — Tue, 24 Feb 2026 14:30:00 GMT

The modern world no longer runs on delayed responses. It runs on immediacy. When a self-driving vehicle identifies a pedestrian, when a factory robot adjusts production in milliseconds, or when an augmented reality overlay appears instantly during r...(read more)

Data Center Digital Twins: How Simulation Improves Design and Performance

Veena Parthan — Thu, 19 Feb 2026 07:28:00 GMT

Data center digital twins are transforming data center design from assumption-based planning to physics-backed simulation—well before the first rack is deployed. (read more)

Data Center Operations, DCIM, and Monitoring

Vinod Khera — Thu, 19 Feb 2026 05:21:00 GMT

In today’s digital world, data centers underpin cloud services, streaming, enterprise apps, and e-commerce. Keeping them operational around the clock is complex. Even brief outages can disrupt operations and lead to significant losses. Accordin...(read more)

Choosing the Right Data Center Strategy: Colocation vs Hyperscale vs Enterprise

Veena Parthan — Thu, 12 Feb 2026 03:30:00 GMT

It is essential to understand how colocation capacity planning differs from hyperscale campus design and enterprise data center modernization. Each data center type presents distinct challenges related to power density, cooling architecture, scalability, and operational control. (read more)

AI, GPU, and HPC Data Centers: The Infrastructure Behind Modern AI

Vinod Khera — Tue, 10 Feb 2026 09:12:00 GMT

Artificial intelligence (AI) is stretching compute infrastructure well beyond what traditional enterprise data centers were designed to handle. Modern AI training requires massively parallel compute, low-latency networking, high-throughput storage pi...(read more)

What Is Power Usage Effectiveness (PUE) in Data Centers?

Reela Samuel — Thu, 05 Feb 2026 14:30:00 GMT

Why PUE Still Matters Walk into a modern AI data center, and the first thing you notice is not the servers, but the infrastructure working continuously to keep heat under control. Behind clean aisles and stable ambient temperatures, GPU-dense racks ...(read more)

Data Center Cooling: Thermal Management, CFD, & Liquid Cooling for AI Workloads

Veena Parthan — Tue, 03 Feb 2026 10:10:00 GMT

Data center cooling is now a first-order design constraint, not an afterthought, as AI, hyperscale cloud, and semiconductor workloads drive higher power densities.(read more)