AI Data Centers Are Squeezing Fiber Supply: What It Means for Business Installations
Optical-fiber spot prices are rising sharply, global supply is tight and some of the world’s largest network operators are reserving cable years in advance. But a fivefold increase in one fiber category does not mean your $50,000 business installation suddenly costs $250,000. Here is where the shortage could actually hit enterprise fiber projects — and when waiting until 2027 may be the more expensive bet.
There are telecom headlines businesses can safely ignore. This probably is not one of them. On October 19, 2026, TrendForce reported that the global optical-fiber market has moved into a significant supply deficit as AI data centers, data-center interconnects and high-capacity backbone construction consume more physical fiber. Some spot-market categories are now being quoted at multiples of their 2025 levels. [1]
The New Numbers Are Hard to Ignore
TrendForce’s October 19 analysis describes a market where demand has increased faster than fiber manufacturers can expand production.
Inventory has reportedly fallen from roughly four months in 2024 to approximately one month in 2026. Some leading suppliers were already booked into the first quarter of 2027. TrendForce expects the shortage to persist until at least 2028. [1]
Not All Fiber Went Up the Same Amount
The eye-catching numbers refer to particular finished optical-fiber categories in the spot market — not a national index of enterprise circuit prices.
Bend-insensitive fiber used extensively inside dense data-center environments shows the largest reported increase. Fiber used in AI backbone and data-center interconnect applications is also under strong pressure, while TrendForce estimates mainstream telecom fiber spot quotations at approximately 3.7 times their previous level. [1]
Why AI Is Suddenly Consuming So Much Glass
Data-center construction accounts for an estimated 18% to 20% of worldwide fiber demand in 2026. TrendForce expects that share to jump to approximately 30% to 35% in 2027. [1]
That matters because a modern AI campus consumes fiber twice: enormous quantities are required inside and around the campus, while equally serious capacity is required to connect campuses, clouds and metro or long-haul backbones.
The Biggest Network Buyers Are Already Locking Supply
The strongest evidence that operators are taking physical-fiber availability seriously is not an analyst forecast. It is the contracts they are signing.
Verizon signed a multiyear, multibillion-dollar agreement covering more than 80 million miles of high-density optical fiber and connectivity products for delivery from 2027 through 2032. [4]
AT&T entered a multiyear fiber-and-cable supply agreement in September 2026 worth more than $3 billion as broadband and AI-related network demand grows. [6]
Meta’s multiyear agreement supports optical fiber, cable and connectivity for its expanding U.S. AI data-center infrastructure. [7]
Zayo expanded its supply relationship to secure a large portion of the cable needed for its plan to add 15,000 route miles through 2030. [5]
A 3.7× Fiber Price Does Not Mean a 3.7× Construction Bill
This distinction is important enough to do the arithmetic.
The Fiber Broadband Association’s latest U.S. deployment study puts median labor-and- material cost at approximately $18 per foot for underground construction and $8 per foot for aerial construction. Labor represents roughly 72% of underground cost and 64% of aerial cost. [2]
An intentionally extreme thought experiment
Imagine that every material used in an underground fiber build doubled overnight while labor stayed unchanged.
Because materials represent only about 28% of the FBA underground benchmark, the modeled $18-per-foot build rises to roughly $23 per foot — an increase around 28%, not 100%.
If every material tripled, the modeled cost rises to roughly $28 per foot — about 56% above the starting point.
Actual fiber glass is only one part of the materials bucket, alongside conduit, handholes, hardware and other components. The exercise is therefore not a forecast; it simply demonstrates why multiplying an entire installation quote by TrendForce’s spot-fiber multiplier would be mathematically wrong.
Meanwhile, Finished Bandwidth Is Still Getting Cheaper
The other side of the market is almost the mirror image.
Lightyear’s 2026 U.S. quote data puts median 1 Gbps DIA at approximately $901 per month and median 10 Gbps DIA at about $1,987. Its year-over-year analysis shows pricing continuing to compress at common DIA speeds, especially 10 Gbps. [3]
Which Business Projects Are Most Exposed?
| Project | Fiber Shortage Exposure | Reason |
|---|---|---|
| Existing on-net 1G DIA | Low | Fiber already reaches the building. The customer is primarily buying electronics and service capacity rather than a new outside-plant build. |
| Existing on-net 10G upgrade | Low–Moderate | Often more dependent on electronics, port availability and carrier network capacity than new fiber construction. |
| New building lateral | Moderate–High | Requires actual construction, fiber, conduit access, splicing, engineering and crew availability. |
| Off-net DIA | High | The carrier may have to extend its network to the property before service can be delivered. |
| Existing dark-fiber route | Moderate | Existing strands reduce construction exposure, although scarce route capacity and stronger demand can still influence commercial negotiations. |
| New dark-fiber extension | High | Physical construction and cable are central to the project economics. |
| New campus fiber system | High | Large strand counts and multiple routes create direct exposure to fiber, conduit and installation capacity. |
| Colocation cross-connect | Low | Short facility-based fiber runs are unlikely to resemble a multi-mile outside-plant construction project. |
The Building Being “Near Fiber” Is Not Enough
Businesses often hear that a carrier has fiber on the street and assume installation risk is minimal.
The difference between fiber passing the property and fiber terminating inside the telecom room can include permitting, trenching or boring, conduit access, building entrance work, splicing, traffic control and inside-building extension.
Lightyear’s current quote analysis confirms that off-net locations requiring construction tend to fall toward the expensive end of the DIA market. [3]
Should Businesses Lock In Fiber Installations Now?
Start Now: Off-Net Locations
If a project requires trenching, boring or a new building lateral and service is needed in 2027, earlier carrier engineering gives more time to absorb material, permitting and construction delays.
Start Now: Dark-Fiber Construction
A new dark-fiber route is directly exposed to physical construction economics. Obtain route design and firm pricing earlier than you would for a simple bandwidth upgrade.
Do Not Panic: Existing On-Net Service
If multiple carriers already serve the property, keep competing them against one another. Component scarcity does not erase carrier competition or falling bandwidth cost per bit.
Separate Construction From Bandwidth
The ideal deal may lock the physical build while preserving upgrade rights, competitive MRC and flexibility to increase bandwidth later.
Lock the Expensive-to-Recreate Part, Not Necessarily Every Price
This is probably the most useful procurement strategy to take from the shortage.
Physical construction is difficult to recreate. A trench, conduit path, building entrance and fiber lateral represent location-specific infrastructure.
Bandwidth, however, continues to become cheaper per bit as network electronics improve. A business therefore does not necessarily want to freeze today’s bandwidth economics for five years merely because fiber cable is getting more expensive.
Eight Contract Points Worth Negotiating Now
| Contract Item | What to Ask |
|---|---|
| Construction price validity | How long is the quoted non-recurring construction charge protected? |
| Material escalation | Can the carrier increase construction charges after signature because materials became more expensive? |
| Construction cap | Is the customer’s maximum contribution contractually fixed? |
| Ready-for-service date | Is the date merely an estimate or is there a contractual delivery commitment? |
| Upgrade rights | Can 1G become 10G, or 100G become 400G, without restarting an entirely new term? |
| Existing route confirmation | Is the fiber already installed, or does the order depend on a future build? |
| Cancellation exposure | What construction or third-party expenses become non-refundable after order acceptance? |
| Annual service escalator | Does the monthly charge rise automatically even if market bandwidth pricing falls? |
Waiting Six Months Can Cost More Than a Better Future Bandwidth Price Saves
This is the calculation businesses often miss.
Waiting may produce a cheaper recurring bandwidth price. But an off-net project can simultaneously face a larger construction contribution, a tighter delivery schedule and the operational cost of keeping temporary or inadequate connectivity in place.
That means the correct equation is not simply:
It is: “Will the bandwidth savings exceed higher installation cost plus the economic cost of delaying the network?”
Projects That Can Still Afford to Wait
Competitive On-Net Buildings
If several carriers already terminate in the building and no outside construction is required, there is much less reason to sign early solely because global fiber spot prices increased.
Bandwidth-Only Upgrades
An existing 1G connection that can be electronically upgraded to 10G may benefit from continued bandwidth price compression without meaningful outside-plant exposure.
Flexible Move Dates
A project with no firm opening or migration date can retain negotiating leverage rather than paying a scarcity premium simply to guarantee an arbitrary schedule.
Projects I Would Move Earlier
One More Risk: Everyone Orders Early at the Same Time
Long-term agreements can stabilize supply for the buyers who secure them, but they can also remove product from the spot market.
TrendForce expects LTAs to become increasingly common as operators try to guarantee access to cable. Verizon, AT&T and Zayo are already demonstrating that behavior. [1][4][5][6]
For ordinary enterprise customers, the likely effect is not that a carrier tells you, “We have no fiber anywhere.” It is more subtle: fewer cheap construction exceptions, longer engineering windows, tighter quote validity and more scrutiny around expensive off-net builds.
The 2027 Procurement Playbook
| Timing | Action |
|---|---|
| 12+ months before need date | Identify locations likely to require construction. Check which carriers are truly on-net and request preliminary route validation. |
| 9–12 months | Obtain competing quotes with construction charges separated from monthly recurring service. |
| 6–9 months | Confirm route, construction contribution, price validity, permitting assumptions and targeted ready-for-service date. |
| 3–6 months | For construction-sensitive orders, the project’s physical delivery risk should now outweigh attempts to save a few dollars on monthly bandwidth. |
| Under 3 months | If new construction is still required, build a temporary connectivity fallback rather than assuming the permanent fiber will arrive exactly on schedule. |
TrendForce’s price multiples describe selected optical-fiber products in the component market. They should not be interpreted as forecasts that U.S. business fiber installation charges, DIA rates or dark-fiber lease prices will increase by the same percentages. Every enterprise project remains dependent on carrier footprint, construction scope, existing conduit, route length, labor, permitting, contract term and local competition.
Install Now vs Wait 6 or 12 Months
Use an actual carrier quote to test the tradeoff. The model allows construction costs to rise while recurring bandwidth pricing falls — exactly the mixed environment businesses may face heading into 2027.
| Scenario | Future Installation | Future Monthly Service | Delay / Temporary Cost | Total Modeled Cost | Difference vs Now |
|---|
The calculator uses hypothetical growth assumptions supplied by the user. It is not forecasting actual carrier construction inflation or fiber service pricing. Enter a negative service-price change if you expect bandwidth pricing to continue falling. Delay cost can represent temporary circuits, lost productivity, postponed occupancy, cloud migration delay or another business-specific consequence.
