The GenAI Infrastructure Crunch Requires Technological Independence

Forbes

6 days ago

Tomás O'Leary is CEO & Founder of Origina, helping enterprises extend the life of mission-critical IT systems.
Our existing digital infrastructure is cracking under pressure it was never designed to handle, while the foundation beneath our digital economy gives way.
Standing in yet another overheated data center recently watching graphics processing units (GPUs) consume electricity at industrial scales, I had the same realization I had at the Open Compute Project summit in early 2025: we're building a house of cards on quicksand.
Data Centers At A Breaking Point
The numbers tell a story of reckless ambition divorced from reality. The Center for Strategic and International Studies projects U.S. data centers will need 84 gigawatts of power by 2030, a 2,000% increase from today's 4 gigawatts. That's equivalent to adding the entire electricity demand of Texas to the grid, just for AI workloads.
Meanwhile, data centers already consume 3% of global electricity, a figure that is expected to rise sharply. We're demanding infrastructure transformation at warp speed while equipment lead times stretch from months to years. Generators that once arrived in 20 to 30 weeks now take up to two years. Companies seeking data center capacity can face 24-month waiting lists.
Lawrence Berkeley National Laboratory projects that data centers could consume upwards of 12% of U.S. electricity by 2028. Further, Morgan Stanley research found that data centers are on track to generate 2.5 billion tons of CO2 emissions by 2030.
Capacity Hoarding Accelerates
The dynamics of resource control in the data center industry are shifting rapidly. Hyperscalers such as Amazon, Microsoft and Meta are on track to take up over 60% of available capacity by 2029. Many are locking down entire campuses before they even come to market, accelerating a race that leaves smaller players struggling to secure access.
This consolidation raises uncomfortable questions about digital equity. As a handful of firms gain outsized influence over the infrastructure powering AI and cloud innovation, startups, enterprises and even public institutions are increasingly forced to compete on unequal footing for the same finite resources.
How AI Drives Explosive Growth
Generative AI represents the most visible driver of this crisis. Boston Consulting Group (BCG) estimates that GenAI will account for 60% of data center power growth through 2028. Each new AI iteration consumes exponentially more power, water and computational capacity than the last.
However, there are innovators proving that more strategic approaches not only exist, but work. DeepSeek achieved comparable AI performance at 6% of typical costs—$6 million versus the industry standard of $100 million—exposing the fundamental inefficiency at the heart of current approaches.
This isn't isolated cleverness. Companies across several sectors are discovering that targeted solutions often outperform the latest GPU-hungry models.
Traditional Workloads Compound The Crisis
What makes this crisis particularly maddening is that much of the demand driving infrastructure to breaking point shouldn't exist. Some 65% of today's data center power is fueling traditional enterprise workloads, according to BCG. Workloads that continue to grow as companies are forced to upgrade.
While companies today are working to build sustainable data centers, much of that work is being drowned out by avoidable demand, created by a tech culture that treats every upgrade as urgent, even when it's not.
Every pointless software upgrade, every vendor-mandated migration and every compliance-driven refresh adds to infrastructure strain. Companies waste enormous resources on changes that deliver zero business value while competing for the computational capacity needed for actual innovation.
Consider a media company we worked with recently. Faced with a deprecated crypto standard, their initial instinct was an expensive vendor upgrade. Instead, we worked with them to develop a new standard into their existing system, maintaining compliance without the resource waste. By understanding where unnecessary upgrades can be avoided, companies can break free from the cycle and redirect limited compute resources toward real innovation.
Strategic Solutions For Planetary Survival
The math is brutal. Every increase in workloads brings us closer to the moment when our digital infrastructure can no longer support our ambitions, yet the vendor-consulting complex continues pushing the same playbook: bigger models, more hardware and endless upgrades.
The environmental cost is staggering, the economic waste enormous and the strategic risk profound. We're creating artificial scarcity in computational resources while demanding that infrastructure rebuild itself around unsustainable consumption patterns.
Some companies are already adapting. They're optimizing existing systems rather than constantly upgrading. They're choosing efficiency over raw performance. They're building solutions that work within infrastructure constraints rather than demanding the impossible.
This represents more than cost savings—it's about technological independence. Companies that can solve problems without depending on the latest infrastructure arms race will prove more resilient when capacity constraints bite.
True optimization begins with leadership fostering a culture that values operational discipline and challenges the impulse for constant upgrades. By embedding continuous improvement and sustainability into technology decision making, companies can extend the life and effectiveness of existing systems while reducing environmental impact. This mindset shift from expansion to resilience is essential for thriving amid infrastructure constraints.
The pattern repeats everywhere: businesses question whether they need AI or just better data processing, whether they can pilot with existing infrastructure and what the true total cost of ownership looks like compared to alternatives.
The Infrastructure Reckoning Arrives
Early failures are visible in power constraints, GPU shortages and capacity limitations worldwide. The infrastructure reckoning isn't coming—it's already here.
While others wait two years for data center capacity or compete for scarce GPU resources, efficiency pioneers understand that the future belongs to those who achieve more with less, not those who simply scale existing waste.
The question isn't whether GenAI will transform business. Instead, it's whether businesses will transform themselves to thrive within physical constraints rather than chase computational fantasies. Choose efficiency over excess, sustainability over waste and genuine innovation over artificial obsolescence. Our digital future depends on it.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?