🚀 The Invisible Hurdle of the AI Era: The Fiber Optic Cable Crisis!

[Süeda Asil]

When thinking about AI infrastructure, power and chips usually come to mind first. However, fiber optic cables, which connect thousands of AI chips and link data centers to the outside world, are an overlooked but critical part of this equation.

🚨 Fiber Optics: The Hidden Bottleneck of AI Development​

This is a big mistake! Fiber optics have become one of the biggest constraints on AI infrastructure, and most companies are not managing this well. An AI-optimized data center requires significantly more fiber than a traditional one because AI chips need to be more densely interconnected to function as a single system.

• Demand has sharply increased in the last two years.
  
• Prices have exploded.
  
• Lead times have lengthened.
  

Some large tech companies are signing multi-billion dollar supply contracts directly with fiber manufacturers. This industry, with a supply chain built for stable demand, has never seen such interaction before, and the sudden shock is revealing planning gaps that need to be addressed.

⏳ Structural Mismatch: Not a Temporary Glitch!​

Fiber manufacturing is a precise and capital-intensive process. Building new capacity typically takes 18 to 24 months. This means that money committed today will not relieve the market until 2027 at the earliest. The constraint is not capital; it's time and the physics of manufacturing.

This timeline puts suppliers in a difficult position. Dell'Oro Group's estimate of $1.7 trillion in global data center capital expenditure by 2030, and Goldman Sachs' estimate of approximately $5.3 trillion for hyperscalers by 2030, point to years of growth rather than a short surge.

However, there is a real debate about the risk of overbuilding. No one can say for sure whether demand will plateau for years or partially reverse. Suppliers are being asked to commit hundreds of millions of dollars to an 18-24 month build cycle despite uncertain demand and limited planning transparency.

📉 Where Does Planning Go Wrong?​

Several issues make managing the underlying mismatch even more challenging:

• Inconsistent Planning Among Hyperscalers: When you ask five hyperscalers for their 12-month build plans, you get five different answers, not just in numbers but also in the level of detail. This inconsistency makes it difficult for suppliers to form a coherent demand outlook.
  
• Project-Based Demand vs. Standard Product Economics: Data center builds are large, one-off, timeline-driven projects, while fiber manufacturing operates on standard lines designed for stable output.
  
• Evolving Technical Requirements and Constrained Allocation: The specific types of fiber customers need are still evolving. This means suppliers building capacity risk having the wrong capability. Furthermore, every unit of fiber allocated to a hyperscale AI contract cannot be offered to a broadband, telecom, or industrial customer.
  
• Manual, Labor-Intensive Processes: Much of cable manufacturing and termination is still done by hand. Automation could help meet demand, but this transition itself is a significant undertaking on top of everything else straining the supply chain.
  

None of these are fatal on their own. But taken together, they suggest that customers and suppliers are not on the same page about true demand, which is the gap collaborative planning aims to close.

🤝 Solution: Not Just More Capacity, But Collaborative Planning!​

None of these problems are solved by investment alone. They are solved by better decisions made together, starting from a common set of information. Integrated business planning (IBP) and integrated tactical planning (ITP) are designed to close this gap.

First, internal discipline must come. A company cannot extend collaborative planning to its customers or suppliers if its own IBP and ITP are not working well. Once this foundation is solid, the same approach can be extended externally. This way, a data center customer and a fiber supplier can operate from the same demand signal, the same capacity constraints, and the same risk outlook.

This allows both parties to do what they are currently struggling with: remove recurring demand from forecasts, make clear trade-off decisions on allocation, time capacity investment against an honest view of how long demand will last, align technical roadmaps to harmonize new features and new manufacturing capabilities, and present a single, accurate picture to sales teams that can actually be committed.

💡 Key Takeaway​

While compute and power grab headlines, fiber is a quieter constraint and may be more at risk: building new capacity takes longer than the industry's confidence in how long today's demand will last. Forecasters disagree, and suppliers cannot wait for this to resolve itself.

The answer is not a bigger bet that the boom will continue. It's about building true planning discipline within each company first, then connecting it across the customer-supplier relationship so that both parties can make timely, joint decisions on service, cost, cash, and profitability as the picture changes.

Whichever way demand goes, companies with this discipline will be the ones that can adapt without being stranded.