Optical Fiber Price Surge: What Is Driving It and How Long Could It Last?

The current optical fiber price surge is no longer a niche procurement issue. In China, market coverage in early 2026 described G.652D moving from below RMB 20 per fiber-kilometer in late 2025 to above RMB 35 in January 2026, with spot quotations later moving past RMB 50 and in some cases toward RMB 60. G.654E also moved sharply higher, with quoted ranges rising from roughly RMB 130–140 to RMB 170–180, and some spot quotations reported materially above that level.

This matters because optical fiber remains core infrastructure, not a marginal input. China’s 2025 telecom statistical release said national optical-cable route length reached 74.99 million kilometers by year-end, while CRU-linked market commentary put 2025 global fiber shipments at roughly 662 million fiber-kilometers. A price move at this layer affects telecom networks, broadband rollout, data-center expansion, industrial connectivity, and public procurement.

The current optical fiber price surge refers to a structural supply-demand imbalance in which new AI-related data-center demand, specialty-fiber demand, and slow upstream capacity response are pushing fiber prices higher. It is not just a routine telecom cycle rebound, because the new demand is more fiber-intensive, more specification-sensitive, and harder to satisfy quickly.

For years, the fiber business was heavily shaped by operator-led build cycles: backbone networks, FTTH, and mobile-network expansion. Those cycles could be large, but they were still recognizably cyclical. CRU noted that by mid-2025 China Mobile’s large optical-cable tender still reflected weak domestic conditions and persistent oversupply from earlier years, with implied fiber pricing around RMB 18.85 per F-km including VAT. That is an important baseline, because it shows how quickly the market shifted from oversupply psychology to scarcity psychology.

By late 2025, the demand structure had changed. CRU described AI-driven data-center investment as the strongest growth driver in the optical fiber and cable market during 2025, while traditional telecom demand softened in several markets. In other words, this is not simply “another telecom upcycle.” It is a market in which new compute infrastructure is changing what kinds of fiber are needed, where they are needed, and how urgently buyers want to secure them.

The shift is visible not only inside data centers, but also between them. DCI, or data-center interconnect, matters because AI does not live inside a single building. Training clusters, storage systems, backup sites, and geographically distributed compute resources all increase the need for high-capacity optical links. CRU said data-center applications would account for roughly 5% of total global optical cable demand in 2025, a small share in absolute terms but already large enough to alter the balance in a market that had previously been dominated by telecom deployment.

The more important point is not the starting share, but the growth rate and product mix. LightCounting said AI created a new wave of demand for optical connectivity between 2023 and 2025 and expected that growth impulse to continue through 2030. Some more aggressive market commentary has projected a much larger late-2020s share for data-center and DCI-related fiber demand, but the exact percentages should be treated as scenario estimates rather than settled facts.

This is the key transmission mechanism behind the current price spike. G.652 remains the standard workhorse single-mode fiber family for mainstream telecom deployment, while G.654 is defined by ITU-T as a very low-loss, cut-off shifted single-mode fiber optimized for use in the 1530–1625 nm region and suited to long-distance digital transmission. When high-value projects pull more low-loss fiber into AI backbones and DCI links, they do not just make G.654E more expensive. They also redirect upstream manufacturing attention away from mainstream products.

The table summarizes reported market moves described in Chinese trade and financial coverage.

AI changes fiber demand because it changes interconnect density. Corning has stated that generative-AI-enabled data centers require over 10x more optical fiber than traditional data-center networks. That is consistent with broader market commentary describing AI clusters as dramatically more fiber-rich because east-west traffic inside the compute fabric becomes far more intense, and because high-performance fabrics require many more optical pathways per rack, row, pod, and site.

That is why even a moderate change in the data-center share of total demand can still move the whole market. The issue is not only volume. It is volume multiplied by density, multiplied by performance sensitivity, multiplied by urgency. AI infrastructure consumes more fiber, but it also tends to favor lower-loss or more carefully optimized links, which tightens the supply picture disproportionately.

In technical terms, G.654 sits in a different positioning from G.652. ITU-T defines it as loss-minimized and optimized around the 1530–1625 nm operating region, which is why it is closely associated with long-haul terrestrial and submarine transmission. In commercial terms, that means it is well placed wherever buyers care deeply about loss budgets, span economics, or premium long-reach performance. AI-related backbone buildout and DCI do not automatically mean every link becomes G.654E, but they clearly increase the demand pull for low-loss fiber categories.

That helps explain why G.654E pricing moved sharply at the same time as G.652D. A market that once treated low-loss fiber as a more specialized category is now seeing more capital directed toward applications that justify paying for that performance. Once manufacturers see stronger margins and more urgent buying in that segment, the knock-on effect on mainstream allocation becomes hard to avoid.

North America matters because hyperscaler capex is now large enough to influence supply chains directly. In January 2026, Corning and Meta announced a multiyear agreement worth up to USD 6 billion for fiber-optic cable to support Meta’s U.S. data-center buildout. Corning’s own 2025 results showed USD 6.274 billion in full-year Optical Communications net sales, which means the Meta commitment is not a symbolic order. It is large enough to illustrate how AI buyers are increasingly locking in supply at the top of the market.

Broadband policy adds another layer. The U.S. BEAD program provides USD 42.45 billion to expand high-speed internet access. That is not the same as a simple “100% fiber mandate,” and it should not be described that way. But it does reinforce the broader point: U.S. demand for fiber-related infrastructure is being supported by both hyperscaler AI investment and large public broadband programs. When those forces overlap, global supply becomes more exposed to North American buying behavior.

The “AI is the only reason” story is too simple. Another incremental demand channel comes from fiber-guided FPV drones. ITU-T G.657 defines bend-loss-insensitive single-mode fiber, and the G.657.A2 subcategory is appropriate for a minimum design radius of 7.5 mm while remaining compliant with G.652.D transmission and interconnection properties. That makes it attractive wherever fiber must be wound tightly, handled roughly, or deployed in a space-constrained format.

Battlefield reporting in 2026 described fiber-guided drones operating over distances of up to 50 kilometers, specifically because fiber control links are resistant to jamming. Whether one focuses on exact per-mission spool length or not, the engineering logic is clear: this is a consumable, specialty-fiber application that did not matter much to the mainstream cable market a few years ago, but it now absorbs real manufacturing attention.

Once specialty demand becomes meaningful, the question is no longer just “how much fiber is produced?” but “what kind of fiber is produced, and at what manufacturing efficiency?” Market commentary around G.657.A2 has repeatedly linked the recent price surge to new defense demand and to lower effective throughput than standard telecom fiber. Even where precise numbers vary by producer and line configuration, the direction of the effect is consistent: specialty fiber can consume more scarce upstream capability per unit of mainstream-equivalent demand.

This mapping combines ITU fiber definitions with current market reporting on AI infrastructure and fiber-guided drones.

When buyers see prices jump, the natural question is why manufacturers do not just turn on more output. The answer is that full drawing-line utilization is not the same as easily expandable supply. Supply-chain reporting and industry commentary in 2025–2026 repeatedly identified a “perfect storm” in which AI demand, policy-driven broadband buildout, and trade frictions were tightening fiber availability, especially in the U.S. market.

The deeper issue sits upstream. In practice, the industry can debottleneck some downstream processes faster than it can add robust upstream capability. That is why a market can appear operationally “full” without having a credible path to a near-term supply reset.

The true structural bottleneck is often the fiber preform stage, not just the draw tower. Multiple industry sources describe preform manufacturing as the more technically demanding and capital-intensive step in the chain. That matters because producers burned by earlier oversupply and price wars do not usually race to add large new upstream capacity at the first sign of better pricing. They tend to wait for confirmation that the demand shift is durable.

That historical context helps explain why supply response looked slow even though AI had already become a visible theme before 2026. A market can correctly perceive demand growth and still respond too late if recent memory is dominated by price compression, oversupply, and weak utilization. In fiber, that behavioral lag matters almost as much as the physical bottleneck.

Short-term price spikes can sometimes be solved by faster procurement or extra shifts. Preform scarcity is different. If the upstream process is the hard constraint, then a price increase does not automatically create a quick supply cure. That is why the current market feels more structural than opportunistic. Even the buyers who believe pricing will eventually stabilize still have to plan around a period in which upstream conversion cannot instantly catch up with upgraded demand.

The constraint picture above synthesizes current supply-chain reporting, CRU market framing, and public company disclosures.

G.652D is not the most glamorous fiber in the market, but that is precisely why it sits at the center of the price shock. It is the broadest-application product, the volume anchor for conventional network deployment, and the category most exposed when premium demand and specialty demand pull on the same upstream resources. When the market tightens, the workhorse product often becomes the most visible casualty.

The pressure on G.652D does not require G.652D demand itself to become extraordinary. It is enough for G.654E to capture more premium allocation and for G.657.A2 to absorb more specialty capacity. Once both happen at the same time, mainstream supply can tighten even if total industry output has not collapsed. That is why G.652D becomes the “price pressure point” in a structurally mismatched market.

A disciplined answer is that the current cycle looks too structural for a quick snapback. CRU described AI-driven data-center investment as a defining growth driver in 2025, while LightCounting expects AI-related optical-connectivity growth to continue through the decade. Corning’s large Meta commitment reinforces the same signal from the buyer side: this is not a one-quarter restocking event.

Several forces can keep prices high simultaneously: continued AI cluster buildout, more DCI spending, public broadband programs, and ongoing specialty-fiber pull from military applications. On top of that, operator-side procurement in China is already showing stress, with emergency cable tenders requiring repeated price-limit increases or multiple rounds before completion. That kind of behavior is exactly what one would expect in a market where supply is no longer comfortably elastic.

Some market forecasts go further and argue that a material global supply gap could persist into 2026 and beyond. Those projections should be treated as forecasts, not facts, but they do align with the broader logic of a market constrained by upstream preform response and product-mix competition.

No responsible forecast should pretend that duration is certain. Fiber pricing depends on whether hyperscaler capex stays elevated, whether premium-fiber orders keep crowding out mainstream allocation, whether public broadband projects accelerate or slip, and how quickly upstream capacity actually comes online. The most defensible judgment today is not “prices will stay high for exactly X months,” but rather that the conditions for a fast reversion are not yet obvious.

Downstream buyers feel the squeeze before the market reaches any formal equilibrium. In March 2026, reporting based on China Telecom Sunshine Procurement disclosures described emergency optical-cable tenders that failed, reopened, and only cleared after meaningful upward revisions to bid caps. That is not just a price story. It is a risk story for operators, EPC contractors, and integrators who budgeted projects under very different fiber assumptions.

When supply is uncertain and spot quotations keep moving, forward buying and inventory building become rational, even if they worsen the tightness. Buyers are not only reacting to today’s price. They are buying against the risk of tomorrow’s non-availability. That is one reason markets can overshoot during structural transitions: defensive procurement becomes part of the demand surge itself.

Paradoxically, a shortage in conventional fiber can also slow enthusiasm for newer fiber technologies. When mainstream budgets are already under pressure, adoption of newer and more expensive categories such as hollow-core or advanced multicore concepts may be delayed outside the highest-value use cases. The technology roadmap does not disappear, but commercial adoption becomes more selective when the industry is still fighting over conventional capacity.

The most useful way to understand the current optical fiber price surge is not as a single-cause event. AI matters, but so do DCI growth, premium-fiber allocation, specialty-fiber demand from FPV drones, and the slow response of upstream preform capacity. In that environment, G.652D becomes the most visible pressure point not because it is the most advanced fiber, but because it is the market’s workhorse.

The broader lesson is that optical fiber is no longer priced only by the old telecom cycle. It is increasingly being priced by the intersection of AI infrastructure, specialty applications, and upstream manufacturing rigidity. That is why the current rally looks structural, and why any expectation of a quick normalization should be treated with caution.

• Why are G.652D fiber prices rising so sharply?

  Because G.652D sits at the center of mainstream network deployment, it feels the strongest pressure when premium low-loss fiber and specialty bend-insensitive fiber compete for the same upstream resources. Recent Chinese market coverage showed G.652D moving from sub-RMB 20 levels in late 2025 to above RMB 35 in January 2026 and above RMB 50 in subsequent spot quotations.

• How is AI data center growth affecting optical fiber demand?

  AI data centers use far more optical connectivity than traditional facilities. Corning has said generative-AI-enabled data centers require more than 10x as much optical fiber as traditional data-center networks, and CRU has described AI-driven data-center investment as the strongest growth driver in the optical fiber and cable market during 2025.

• Why does G.654E demand matter for the broader optical fiber market?

  Because G.654-type fiber is positioned for lower-loss, long-reach, performance-sensitive applications. When AI backbones and DCI links pull more of that product into the market, manufacturers have stronger incentives to prioritize premium output, which can indirectly tighten the availability of mainstream G.652D. (ITU)

• How do FPV drones increase demand for G.657A2 optical fiber?

  Fiber-guided FPV drones create a new specialty-fiber consumption channel. G.657.A2 is attractive because it is bend-loss-insensitive and suitable for tighter handling conditions, while battlefield reporting in 2026 described fiber-guided drones operating over distances of up to about 50 km to resist jamming.

• Why can’t fiber manufacturers expand capacity quickly when prices go up?

  Because the real bottleneck is not just downstream drawing capacity. Industry reporting consistently points to upstream preform manufacturing as the slower, more capital-intensive stage. That means price signals can arrive faster than credible new capacity.

• How long could the current optical fiber price surge last?

  There is no precise universal answer, but the current setup does not look like a short-lived fluctuation. AI investment remains strong, public broadband programs continue to support fiber deployment, specialty-fiber demand has added a new pressure channel, and procurement stress is already visible in operator tenders. That combination argues for caution against expecting a fast reset.