The AI Build-Out Is Running Into Real-World Limits
top of page
Search

The AI Build-Out Is Running Into Real-World Limits

  • Writer: Girish Appadu
    Girish Appadu
  • Dec 10
  • 2 min read

Everyone is talking about AI software and algorithms, but the real constraint lies in physical infrastructure. While 2025 US data-centre plans look massive on paper, most are not actually being built. Many projects are stalled due to power shortages, slow transmission lines and financing complications.


Key challenges holding back AI infrastructure:


  • Construction delays: Only about one in ten planned data centers is actually under construction.

  • Power supply issues: Many projects face multi-year waits to connect to the grid.

  • Limited generation capacity: Factories producing power equipment are booked out for years.

  • Slow transmission lines: New high-voltage lines can take 6–10 years to build.

  • Financing arrangements: Some deals make demand appear higher than it really is.


The bottom line: 


Announcing data centres does not guarantee they can operate. Without electricity and proper connections, these facilities remain non-functional. Yet AI-related stocks often trade as if these constraints do not exist, creating a disconnect between market expectations and reality.


Why investors should pay attention:


  • Building AI infrastructure is slower and more expensive than many anticipate.

  • This delay creates a short-term opportunity to capitalise on the gap between AI demand and available data-centre capacity.

  • The window for this opportunity is roughly 2026 to 2028.


The limiting factor isn’t AI technology. It is power and space:


  • Running large AI systems requires vast amounts of electricity and computing equipment.

  • Even tech giants such as Microsoft, Amazon, Google, and Meta cannot build new facilities quickly enough to meet demand.


Emerging opportunities for investors:


  • GPU-sharing networks: Companies are using spare computing power from personal PCs, universities, and businesses to meet demand.

  • Cheaper and faster: These networks avoid billion-dollar data centres and can scale incrementally.

  • Serving smaller players: Startups, independent developers and emerging markets gain access to computing power they otherwise could not afford.


Investment insight:


  • Upside: Networks can profit during the shortage and establish long-term credibility.

  • Downside: Returns diminish as large data centres come online.

  • Strategy: Treat this as a temporary arbitrage play. There is high potential upside with limited risk if positions are sized carefully.


Conclusion: 


AI demand is surging, data-centre capacity is limited and the 2026–28 period presents a unique window of opportunity. The key question for investors is whether you are positioned to benefit from this temporary gap before the market normalises.


ree

bottom of page