Three Years to Rebuild the Internet or Lose the AI Race

When the United States built railroads, power grids and interstate highways, it didn’t just expand the old infrastructure — it created entirely new ones. Each leap unleashed decades of growth, strengthened national security and cemented America’s role as a global leader. 

AI presents a similar dramatic inflection point. But the telecom infrastructure we are relying on — the public internet and today’s version of cloud computing — was never designed for the scale, speed and security AI will require. As I outlined in my recent white paper [Cloud 2.0: Because AI Won’t Run on Yesterday’s Internet], unless we act now, the U.S. risks ceding technological and economic leadership to those who solve this challenge first. 

Cloud 1.0, built for SaaS and e-commerce applications, cannot carry the industrial-scale AI workloads already emerging. AI “factories” are being constructed to train and retrain models around the clock, moving petabytes / exabytes of data daily across networks. These patterns demand low-latency, high-capacity, programmable connectivity — capabilities the public internet cannot reliably provide. 

The period around the COVID-19 pandemic first exposed the limits of this system, as surging digital demand tested bandwidth and resilience. AI is magnifying those pressures exponentially. 

My research and analysis by leading firms such as 4MC predict a 10-fold increase in U.S. data-center capacity by 2028, much of it in rural and suburban corridors where space, power and water are more available. This is a once-in-a-generation infrastructure build-out. But unlike railroads or highways, it will unfold in just three to five years, not decades. 

If the U.S. fails to prepare, the bottlenecks will hit fast. The result: constrained innovation, delayed AI deployment, rising costs for business and new vulnerabilities that underpin national security and global economic competitiveness. 

That’s why the conversation can’t remain about more bandwidth at the margins. It must shift to building an entirely new digital backbone. 

Cloud 2.0 will be the infrastructure for the AI century — a purpose-built fusion of cloud architecture and enterprise core networking. Instead of the “flat” internet model of the past 25 years, Cloud 2.0 will establish a high-performance, low-latency programmable fabric backbone designed specifically for AI’s demands.

Dedicated high-speed connections will link premises to data centers and connect clouds directly to one another, bypassing the congestion and unpredictability of the public internet. These connections will be programmable, allowing capacity to scale up or down instantly as workloads require and they will integrate public and private environments seamlessly while embedding security into the architecture from the start. 

At the heart of Cloud 2.0 will be a new data center interconnect core, optimized for both AI training — with its enormous, sustained data flows — and inference, which requires fast, precise responses from the core all the way to the edge. Together, these capabilities will give enterprises and public-sector institutions the ability to move data and deploy AI models with the speed, efficiency and reliability the AI economy will demand. 

This will not replace the public internet; it will supplement it with a high-performance, secure backbone for AI. 

The stakes extend far beyond business efficiency. Cloud 2.0 will underpin industries critical to national security, from defense and energy to financial services and healthcare. Without it, the U.S. risks delays in expanding AI for these vital missions, while other nations that build first gain a decisive global strategic advantage. 

Across the industry, network operators are racing to expand long-haul and metro fiber runs in high-growth corridors, while working closely with hyperscalers and data centers to deliver the high-capacity, low-latency fabric AI workloads will demand. But building Cloud 2.0 cannot fall to one sector. It requires coordinated investment from private industry, public-sector agencies and the financial community—much like past infrastructure revolutions did. That means forward-leaning policies to accelerate fiber deployment, investment incentives for backbone expansion and public-private partnerships to ensure security and resiliency are built in. 

History shows what happens when America leads with vision. From the Erie Canal to the space race, U.S. prosperity has been built on moments when we chose to invest in the infrastructure of the future before the need became unavoidable. AI is our next such moment, and the nation that builds Cloud 2.0 first will shape the innovation economy for decades.  

This is more than a technology upgrade. It’s a strategic imperative for economic competitiveness, national security and global leadership. Let’s invest in building Cloud 2.0 for the AI era. If we do, America won’t just participate in this 21st-century technological revolution; it will lead it.  

Dave Ward is Chief Technology & Product Officer for Lumen Technologies.