AI Infra Summit Highlights: Cadence's Unique Design for AI and AI for Design

The AI Infra Summit 2025 was a great experience that left attendees buzzing with inspiration! From September 9 to 11, the event became a melting pot of innovation as industry leaders, visionaries, and experts came together to explore the latest advancements in artificial intelligence (AI) infrastructure.

The summit was more than just a gathering—it showcased innovative developments, highlighted by inspiring keynote speeches, insightful panel discussions, and interactive demonstrations that provided a glimpse into AI's transformative potential. The summit's success was evident in the packed auditoriums, engaged audiences, and dynamic discussions that took place across every session.

Throughout the three-day summit, Cadence made a significant impact, delivering comprehensive insights across multiple presentations, panels, and demonstrations. The event highlighted our commitment to driving innovation through design for AI and AI for design: our role in building out the AI infrastructure and in using AI to improve our own solutions for designers.

Chuck Alpert's Visionary Keynote: Design for AI and AI for Design

At the summit, Cadence AI Fellow Chuck Alpert addressed how AI has evolved from being just a workload to becoming an indispensable tool in designing the chips, systems, and data centers that power AI itself. His keynote, titled "Design for AI and AI for Design," shed light on the growing computational demands of hyperscale AI, the trillion-dollar investment cycle in infrastructure, and the transformative role of agentic AI in design workflows. According to Alpert, the compute demand for AI is doubling every year, with each new generation of systems consuming 1.5 to 3 times more power. He also predicted that in the next three years, 90% of chips will be designed using AI technology.

Alpert believes that this rapid growth necessitates innovative solutions, including advancements in 3D packaging, liquid cooling, and photonic integration—areas where Cadence is making significant strides by extending its EDA expertise into multiphysics simulation and digital twin modeling. He emphasized that at Cadence, leveraging AI in our solutions is part of our DNA and central to our mission. By accelerating simulations with custom hardware and integrating agentic AI, we empower our customers to achieve breakthrough productivity and innovation.

Track Presentations and Panel Discussion from Cadence

The summit showcased two captivating presentations and a panel discussion from Cadence, each diving deep into the world of AI-driven methodologies and AI infrastructure design and optimization.

Elevate Data Center Management to Maximize ROI: Insights from Sherman Ikemoto

On Wednesday, Sherman Ikemoto, Group Director at Cadence, delivered a presentation titled "Simulation-Driven Data Center Design, Operations, and Risk Management." During the session, Ikemoto highlighted Cadence';s partnership with NVIDIA, aimed at addressing simulation needs for next-gen AI factories, and shared that GB200 NVL72 DGX SuperPod is now part of the Cadence Reality Digital Twin Platform library. The new model of NVIDIA's high-performance accelerated computing platform will enable data center designers and operators to easily deploy the world's leading AI accelerator in the buildout of AI factories. Cadence's Reality Digital Twin Platform is the first and only solution that allows users to model AI factories and data centers with high accuracy to a specified service-level agreement—accurately accounting for constraints such as cost, space, energy, cooling, and environmental impact—all before physical implementation.

Ikemoto pointed out some real challenges we face with data center performance and costs, sharing how capacity goes underutilized, leading to serious economic impacts—$700 billion globally! He further dug into what causes these performance hiccups, explaining that it boils down to the variability within fixed facility infrastructures. He highlighted that while design assumptions tend to be generic and fixed, the realities of operations can be all over the place, which can affect available capacity and the reliable operation of IT services. He also referenced the Uptime Institute's Data Center Management Maturity Model, suggesting that to hit Level 4 maturity, we really need physics and AI-driven integrated data center infrastructure management (DCIM) systems with some automation.

Ikemoto then moved on to explain the advanced features of the Cadence Reality Digital Twin Platform, which supports predictive simulations to optimize power and cooling systems while improving operational strategies. Furthermore, the Cadence Reality Digital Twin Platform plays an important role in assisting data centers in achieving levels 4 and 5 maturity.

Digital Twins in Action: Demonstrated by Paul Harrison

During the summit, Paul Harrison, Senior Principal Application Engineer at Cadence, delivered a demo stage presentation titled "Simulation-Driven Data Center Design, Operations, and Risk Management." He highlighted the increasing complexity of designing and operating data centers in the era of AI factories and introduced the Cadence Reality Digital Twin Platform as a solution to address these challenges.

Harrison detailed the platform's extensive database of library items, which are the digital twin "building blocks" meticulously crafted to mirror the functional behavior of their physical counterparts. Among these library items are AI factory-specific library items and reference designs. The NVIDIA DGX SuperPOD with DGX GB200 systems library items were discussed, given that it is one of the world's most powerful accelerated computing platforms, which now allows AI factory data center designers and operators to accurately model data center performance.

He further enriched his presentation with an example that demonstrated how the platform's predictive analysis capabilities can enhance future planning and improve cabinet power management. In the demo, he showcased the platform's predictive analysis capabilities by simulating the thermal behavior of potential failure scenarios, enabling proactive and informed decision-making.

Panel Discussion on Designing Power-Efficient Chips—From the Edge to the Data Center

On Tuesday afternoon, Badarinath Kommandur, a Cadence Fellow, participated in an in-depth panel discussion alongside Laurent Moll, the chief operating officer of Arteris, and Darren Jones, a distinguished engineer from Andes Tech. The panel discussion sparked an engaging conversation on the critical challenges associated with power optimization across a diverse spectrum of AI applications. Together, they shared insights on potential pitfalls and brainstormed innovative solutions to tackle these complexities.

Booth Presentations: Innovation on Display

Throughout the summit, the Cadence booth served as a hub of innovation, showcasing cutting-edge solutions that address the most pressing challenges in AI infrastructure design. The booth presentations covered four key technology areas that are transforming AI system capabilities.

Ultra-Low-Power UCIe IP with Silicon-Proven Performance at 32G: Best-in-Class Efficiency and Reliability
This demo showcased how chiplet-based architectures can scale compute density while maintaining energy efficiency. An advanced package version, offering even greater integration flexibility, is currently under lab characterization and expected to be ready by early 2026.

HBM3E at 14.4Gbps for AI Data Centers: High-Bandwidth, Low-Latency System Solution
In this demo, a complete system solution with tightly integrated PHY and controller IP, enabling efficient data flow between memory and compute, was presented. This is critical for training large models and running inference at scale, where every nanosecond counts.

224G Long-Reach SerDes for UALink: Enables Scalable AI Interconnect
This demo emphasized Cadence's design IP performance at the AI Infra Summit, showcasing a 45dB channel reach that surpasses the 40dB specification. This impressive capability ensures reliable, high-speed communication across racks and data center fabrics, which is crucial for the substantial data exchange needs of AI accelerators.

The above three Cadence demonstrations were focused on the three fundamental pillars of AI system design: memory bandwidth, high-speed interfaces, and scalable interconnects. For a detailed read, check out this blog post!

Digital Twins for AI Factories: Simplifying Design Complexity with Advanced Simulation
This demo highlighted how Cadence Reality Digital Twin Platform leverages advanced simulation and modeling techniques to drive efficiency. It also introduced vendor-approved behavioral models for AI factory components, which serve as essential building blocks for creating digital twins.

Until We Meet Again at the Next AI Infra Summit!

The AI Infra Summit 2025 reminded us that we’re entering an exciting new chapter in technology. With the growing demand for computing power, significant investments in infrastructure, and innovative AI design tools coming together, we have some incredible opportunities ahead.

Cadence's presence at the summit—from Chuck Alpert's visionary keynote to the practical demonstrations—showcased how the company is leading this transformation. The integration of AI into design workflows isn't just improving efficiency; it's fundamentally changing what's possible in chip and system design.

The summit's success reflects the industry's recognition that traditional design approaches are insufficient for meeting the demands of AI infrastructure. The solutions demonstrated by Cadence and other industry leaders point toward a future where AI-enabled design tools, digital twin technologies, design IP, and advanced simulation capabilities work together to tackle previously impossible challenges.

For engineers, designers, and technology leaders working in AI infrastructure, the insights shared at this summit provide a roadmap for navigating the complexities ahead. The technologies demonstrated aren't just theoretical concepts—they're practical solutions already delivering results in production environments worldwide.

A big thank you to the team that made Cadence's presence at the AI Infra Summit 2025 a success!

Excited about what you saw at AI Infra Summit 2025? Learn more about Cadence AI Solutions.