While the market focuses on the high growth within the AI computing power supply chain, some seemingly "traditional" industries are quietly approaching an inflection point as their own supply is being squeezed by AI demand. Fiber optic cables represent one of the most typical cases of this trend. Recent developments indicate that signals of shortage and price increases in the fiber optic industry are becoming increasingly clear. Guotai Haitong stated in a February 5th report that a structural contradiction has emerged for mainstream domestic G.652D loose tube fiber, characterized by "difficulty securing supply with cash even at high prices." Delivery lead times are lengthening. After prices recovered in the second half of 2025, they climbed further in January 2026. Upcoming operator centralized procurement and pre-holiday stockpiling ahead of the Spring Festival may sustain this upward price trend.
The core incremental demand stems from AI data center internal connectivity and Data Center Interconnect (DCI). Citing CRU, Guotai Haitong reported that global data center demand for fiber optic cables is projected to increase by 75.9% year-over-year in 2025. They further estimate that the proportion of fiber demand driven by AI-related applications could rise from less than 5% in 2024 to 35% by 2027. "Volume locking" by overseas players is also intensifying supply tightness. Corning signed a long-term supply agreement with Meta valued up to $6 billion and will expand production to support Meta's AI data center construction. For investors, fiber optics is seen less as a perfect replica of the memory market, but rather as having characteristics of a "mini-memory" cycle. In a February 11th report, Shenwan Hongyuan Research positioned fiber optics as an "upside option combining traditional sector recovery with short-term AI business progress expectations." They believe the industry's supply has been consistently declining since 2023, confirming a cyclical bottom. However, with the sector's Price-to-Book (PB) ratio at historically high levels, potential returns are expected to rely more on the realization of fundamental improvements and structural upgrades, rather than simple valuation expansion alone.
The memory market provides a template for how high demand can "squeeze out supply" and spill over into traditional product categories. Shenwan Hongyuan Research's analysis suggests the core driver of the 2025 memory market rally was structural changes in AI training demand, which prioritized production capacity allocation towards high-margin products like HBM. This squeezed out capacity for traditional DRAM and NAND. Concurrently, AI inference workloads boosted broad-based demand for memory, leading to a dramatic shift in supply-demand dynamics that resulted in a double benefit from both price increases and valuation expansion. This mechanism is summarized by the research firm into four principles of "high-demand spillover": First, earnings-driven gains come from an accelerated inflection point due to squeezed supply, while valuation-driven gains stem from recovery expectations when initial valuations are at or below historical medians. Second, spillover rallies have a floor; they require confirmation of a clear cyclical bottom, controllable downside demand risk, and sufficient supply rationalization. Third, high success probability is not inherent; it requires additional alpha factors such as pricing power, production expansion bottlenecks, and the potential to enter high-growth sectors. Fourth, valuations in spillover-driven segments struggle to sustainably break above historical averages for extended periods, and are more prone to high-level volatility once value propositions diminish.
On the supply side, the long expansion cycle for preform and capacity shifts towards higher-value products are constraining traditional supply. The "supply rigidity" of fiber optics forms the foundation of the current price increase narrative. Guotai Haitong notes that the core bottleneck in the fiber optic supply chain lies in the production of fiber preforms, which have a lengthy expansion cycle of 2 to 3 years. Long-tail preform capacity from the previous industry downturn has been largely eliminated. Incumbent manufacturers are generally cautious about capacity expansion, leading to slow growth in global preform capacity and overall supply inelasticity. Simultaneously, capacity is structurally shifting towards products with higher demand and profit margins. Shenwan Hongyuan Research points out that conventional capacity is being directed towards high-margin products catering to AI data centers and other emerging demands, thereby squeezing traditional fiber capacity. Given the long lead time for upstream preform expansion, there are bottlenecks in adjusting overall capacity allocation. Guotai Haitong also states that leading manufacturers, while ensuring contractual commitments, are continuously shifting conventional G.652D capacity towards higher-profit products like ultra-low loss G.654.E, G.657.A2, multi-core, and hollow-core fibers. This creates a structural imbalance characterized by "insufficient high-end capacity and squeezed traditional capacity."
Price increases are already materializing, evidenced by G.652D shortages driving up spot prices and extending delivery times. The direct manifestations of these increases are seen in loose tube fiber prices and delivery cycles. Guotai Haitong reports that the loose tube market is predominantly G.652D. Weak demand in the first half of 2025 put pressure on prices, which bottomed out after June 2025. In the second half, a significant increase in overseas demand for products like G.657.A2 prompted capacity switching, leading to reduced G.652D supply and longer lead times. By Q4 2025, significant price hikes and supply tightness emerged for mainstream G.652D standard fiber, with inventory at some major manufacturers insufficient to fulfill orders, forcing them to source externally. Price-wise, Guotai Haitong states that G.652D loose tube fiber quotes had climbed to 24.0-25.0 RMB per core-kilometer by the end of 2025, representing a cumulative annual increase of over 50%. By January 2026, the latest spot prices exceeded 30 RMB per core-kilometer, approaching 40 RMB, effectively more than double the fiber equivalent prices seen during the 2025 operator centralized procurement rounds. They believe prices may continue rising due to impending telecom operator procurements and heightened pre-Spring Festival inventory building. Acceptance of these price increases by major domestic customers would further validate the underlying market logic.
On the demand side, the competitive race for computing power is pulling fiber optics from a "telecom cycle" towards a "data center cycle." While the cyclical nature tied to traditional telecom investment persists, AI is reshaping marginal demand. Guotai Haitong highlights that smart computing centers have rigid requirements for ultra-high bandwidth and low-latency transmission, making them the core engine for fiber demand growth. A typical 10,000-GPU cluster may require tens of thousands of core-kilometers of fiber just for internal server interconnectivity. A single smart computing center's fiber demand can be several times, or even more than ten times, that of a traditional data center. Data reinforces this trend. Citing CRU, Guotai Haitong reports the 75.9% projected YoY increase in global data center fiber demand for 2025, noting the North American market is experiencing the fastest growth, driven by data centers. Guolian Minsheng, in a February 2nd weekly report, mentioned that high-quality fiber products from leading companies are in short supply, with lead times for some orders extending to several months. The convergence of factors like AI data center construction is causing concentrated upward demand across various fiber types. The approximately 2-year expansion cycle for fiber preforms exacerbates the supply-demand mismatch.
The potential "upside options" lie in overseas expansion, volume commitments from tech giants, and the industrialization progress of advanced fibers like hollow-core. The fiber optic narrative is not solely about price increases but also a transition from focusing on "volume" to emphasizing "quality." Guotai Haitong notes sustained growth in demand for specialty and multi-mode fibers. Hollow-core fiber, with advantages like lower latency and higher capacity, is seen as a key future direction for intra-data center links. They disclosed progress from several manufacturers: Changfei's hollow-core fiber has achieved a minimum loss of 0.05 dB/km. Changfei, together with Hengtong Optic-Electric, won a Guangdong Unicom procurement project for hollow-core fiber, involving a total purchase of 82 core-kilometers at a unit price not exceeding 33,000 RMB per core-kilometer. Zhongtian Technology participated in a hollow-core fiber pilot with UAE operator e&, with test results showing approximately 30% lower transmission latency compared to traditional fiber. Internationally, Guotai Haitong reports that both Microsoft and AWS are exploring hollow-core fiber. AWS VP of Core Networking, Matt Rehder, suggested hollow-core fiber might eventually replace traditional fiber as the standard for long-haul backbone networks, though supply is currently insufficient. Demand certainty is also reflected in orders and capital expenditure: both Guotai Haitong and Guolian Minsheng referenced Corning's up-to-$6-billion supply agreement with Meta. Guolian Minsheng also cited Tencent Technology reporting Meta's disclosed 2026 capital expenditure plan of $135 billion. Overseas markets provide another avenue of growth for Chinese manufacturers. Guotai Haitong reported that combined exports of optical cables, fibers, and preforms reached 41,000 tonnes in December 2025, up 6.5% YoY, with an export value of $450 million, up 41.8% YoY. For full-year 2025, combined exports totaled 454,000 tonnes, up 11.1% YoY, with a total export value of $4.475 billion, up 44.1% YoY. They view overseas expansion as a significant profit growth driver for industry players.
Regarding the question of becoming the "next memory market," short-term probability has increased, but success hinges more on valuation and sustainability. From a mechanistic perspective, fiber optics shares key elements with the memory market: a concentrated surge in AI demand, supply-side bottlenecks in preform expansion, capacity shifts squeezing traditional product supply, price increases already evident in both prices and lead times, and additional upside potential from high-end products and international markets. Based on this, Shenwan Hongyuan Research judges that fiber optics has a "high short-term probability of success," viewing the potential entry of high-end products into the North American market as one such upside option. However, unlike memory, the fiber optic sector's starting valuation is not low. Shenwan Hongyuan Research explicitly cautions that the sector's PB ratio is at historically high levels, largely due to previous market pricing of certain companies' optical connectivity businesses. This implies that any sustained rally depends more on the actual realization of price increases and a continuous strengthening of fundamental visibility. If price hikes become more widespread or demand visibility fluctuates, the sector is more likely to transition from a trending uptrend to high-level volatility. On the risk side, Shenwan Hongyuan Research highlights potential pitfalls including slower-than-expected AI technological progress, industry supply expansion exceeding expectations, and demand falling short of forecasts. Guotai Haitong also warns of risks related to technological updates and intensifying competition, raw material price volatility, and fluctuations in market demand. For investors, the key variables determining whether fiber optics can replicate a "memory-style" rally will be whether price increases can be sustained through centralized procurement pricing cycles, whether supply bottlenecks remain effective, and whether high-end products and overseas expansion can extend the industry's prosperity from a phase of "shortage" to one of lasting "structural upgrade."