Artificial Intelligence (AI) has experienced a dramatic transformation, evolving from a specialized area of research to the cornerstone of modern technology. Today, AI is driving everything from personalized recommendations and self-driving cars to sophisticated financial algorithms and cutting-edge medical diagnostics.

As these applications grow more advanced and data-intensive, the need for resilient and scalable digital infrastructure has become critical. I had the chance to attend ITW 2025 earlier this month, and one of the standout discussions was the keynote panel, “AI Infrastructure: Can Our Networks and Data Centers Handle the Surge?” — a conversation that brought together some of the brightest minds in the tech industry, including Scott Mills, SVP of Product & Engineering at Digital Realty, to tackle this exact question.

The panel delved into the core of what it'll take to power the next wave of AI innovation. With applications growing more complex and compute-hungry by the day, the panelists didn’t shy away from the tough issues — power, connectivity, scalability, and the massive infrastructure overhaul ahead.

The industry is moving towards 800G interfaces, with both hyperscalers and smaller operators adopting these advanced networking technologies.

Solving the infrastructure puzzle: insights for IT leaders

Here are five takeaways from the panel — each one revealing a key piece of the infrastructure puzzle that IT leaders must solve to unlock AI’s full potential.

1. Power and connectivity are critical
AI infrastructure requires massive power capabilities, with facilities needing 50-100 megawatts or even gigawatt-level power. Connectivity is equally crucial — without a robust network infrastructure, even the most powerful data center becomes useless. Extensive fiber connections and strategic power entitlements will ensure that facilities can operate at their full potential.

2. Edge computing is emerging
AI applications are driving a shift towards distributed computing. The goal is to move computing closer to end-users to reduce latency, with the potential for data centers to be located within 40 miles of major population centers. Real-time applications like autonomous driving are pushing this trend, requiring faster, more localized computing resources. This shift isn't just about proximity but also about creating a more distributed and resilient computing ecosystem.

3. Data center evolution
The industry is transitioning from traditional CPU-based architectures to GPU-centric designs. This involves increasing computational density, with cabinets now supporting up to 150-300 kilowatts, compared to 50 kilowatts just a decade ago. The focus is on maximizing computational efficiency, reducing physical footprint, and improving overall performance. This evolution is crucial for meeting the growing demands of AI workloads and ensuring that data centers remain agile and scalable.

4. Networking transformation
Network architectures are being simplified and redesigned to support AI workloads. This includes integrating advanced technologies like coherent optics directly into Ethernet switches, reducing networking gear, and increasing bandwidth. The industry is moving towards 800G interfaces, with both hyperscalers and smaller operators adopting these advanced networking technologies. This transformation isn't just about speed, but about reducing the complexity and cost of networking gear, making it easier to scale and manage AI infrastructure.

5. Self-powered infrastructure
Many data center companies are developing their own power generation capabilities. This includes exploring green solutions, microgrids, battery storage, and alternative energy technologies. The goal is to reduce dependence on external power grids, control energy costs, and ensure a reliable power supply for increasingly demanding AI infrastructure. This shift towards self-powered infrastructure is a game-changer, aligning with the growing need for sustainable and resilient data center operations.

The panel illuminated more than just the technical requirements for AI — it revealed a paradigm shift in how we think about digital infrastructure. What began as a conversation about data centers and network capacity quickly evolved into a broader vision for intelligent, self-sustaining ecosystems, where power, connectivity, and compute are deeply interwoven.

This convergence of AI, energy, and networking technologies marks a transformative moment for the industry. No longer is it enough to simply scale hardware or add bandwidth. True leadership in the AI era will come from those who can innovate across the full spectrum — from generating reliable, sustainable power, to designing agile, distributed networks, to deploying compute at the edge and beyond.

The message from the panel was clear: the future of AI isn’t just written in code. It’s built on the infrastructure that supports it. As we move into this next chapter, success will depend on our ability to reimagine how we generate, distribute, and utilize digital resources in a world where intelligence must be everywhere — and instantly accessible.

At Digital Realty, we're at the forefront of this transformation. With PlatformDIGITAL®, we're helping organizations build the agile, scalable, and sustainable infrastructure needed to meet the demands of tomorrow. Our solutions are designed to future-proof organizations, ensuring they can leverage the full potential of AI while maintaining operational efficiency and environmental responsibility.

Want to learn more about becoming data and AI ready? Download this ebook to learn more.

Tech Talk: Inside the Ashburn Data Center Market: AI Demand, Power Strategy, and What It Takes to Deploy at the Center of It All