In mid-September, the State Administration for Market Regulation (SAMR) released a batch of national standards for core quantum devices. Lithium niobate, explicitly named in the "Test Methods for Performance of Periodically Poled Lithium Niobate Quantum Frequency Conversion Devices," immediately became the focus of renewed market attention.
01. Lithium Niobate: The "Universal Crystal" in the Optical World
Lithium niobate is a synthetically grown optical crystal, referred to in the industry as "optical silicon." It simultaneously possesses a rare combination of properties: a prominent electro-optic effect, where changes in voltage alter the refractive index, enabling high-speed "writing" of electrical signals onto light waves; a broad transmission window, covering from visible light to the mid-infrared; and a high nonlinear optical coefficient, allowing light to directly "interact" with light to achieve frequency multiplication, conversion, and entanglement. Surface acoustic wave (SAW) filters, electro-optic modulators, laser frequency doublers, and quantum frequency converters—products spanning communications, consumer electronics, and frontier physics—are all built upon this single crystal.
In fact, this is not the sudden emergence of a new material: according to publicly available industry data, SAW filters in the RF front-end of mobile phones consume billions of lithium niobate wafers annually, and lithium niobate modulators in backbone network coherent optical modules have been in service for years. However, in its traditional form, it has always existed as a bulk material discrete component, and the market prices it more like a mature functional material rather than a growth track.
The true industrial variable occurred after "thin-film" processing. Traditional processes fabricate lithium niobate into bulk material devices, which are size-constrained and incompatible with chip manufacturing processes, remaining at the discrete component stage for a long time. Thin-film lithium niobate (TFLN), on the other hand, uses ion slicing and wafer bonding to attach a single-crystal film only hundreds of nanometers thick onto an insulating substrate, forming a substrate structure similar to the semiconductor SOI (Silicon on Insulator) process. Thin-film processing brings about an order-of-magnitude leap: the bandwidth of electro-optic modulators is pushed from tens of gigahertz to over 110 GHz, the driving voltage drops to around 1.5 volts, and power consumption and integration are simultaneously improved, perfectly meeting the demands of next-generation optical modules and photonic integration.
Previously, modulator chips for optical modules were dominated by indium phosphide (InP)-based EMLs and silicon photonics solutions. The former was constrained by the capacity bottleneck of InP substrates, while the latter hit a bandwidth ceiling at single-wavelength rates above 200G. TFLN entered the fray with higher electro-optic bandwidth, becoming the most material-performance-abundant among the three routes for optical chips. A single piece of lithium niobate thus simultaneously stepped onto the demand ends of both AI optical interconnects and quantum technology.
02. National Standards Endorsement and Resonance of Industrial Demand
On September 18, the SAMR (Standardization Administration of China) concentrated on promoting the formulation of a batch of national standards for core quantum devices. It newly released two standards: "Technical Specifications for Quantum Squeezed Light Sources" and "Test Methods for Performance of Periodically Poled Lithium Niobate Quantum Frequency Conversion Devices," and initiated four new standard plans, including continuous-variable optical quantum entanglement sources, diamond nitrogen-vacancy center materials, and solid-state spin quantum magnetometers.
Specifically, for quantum states to enter and exit fiber optic networks, and to achieve entanglement and conversion between light sources of different wavelengths, high-precision frequency conversion is required. Periodically poled lithium niobate (PPLN) is the mainstream device for this step. The standard establishes a unified benchmark for its test methods, meaning that for the first time, the quantum industry has directly referenceable national specifications from device selection to acceptance and delivery. For the lithium niobate industry chain, this signals that, following optical communications, another major customer group has completed its "standardized entry."
In addition, another track is also being catalyzed. The optical interconnect upgrade for AI clusters is pushing 1.6T optical modules toward large-scale volume production, with single-wavelength200G becoming the mainstream rate. After silicon photonics solutions hit the bandwidth ceiling, TFLN modulators have become a popular choice due to their higher electro-optic bandwidth and lower driving voltage. Backlog orders at leading manufacturers are generally scheduled through the end of the year.
Over the past few years, controversies in the quantum industry have focused on "whether prototypes can represent capabilities," with core indicators entirely self-reported by companies and lacking a benchmark for cross-validation. Quantum frequency conversion devices being the first to obtain nationally standardized test methods is equivalent to bringing the most upstream device, which is also common to all routes, into the verifiable scope first. Referring to the previous experiences of the National Essential Medicines List and the 3C certification for charging piles, every time a standard covers a specific link, procurement in that link shifts from customization and empiricism to batching and standardization. For leading enterprises that possess compliant capacity, this is a favorable development at the structural level.
From a technical perspective, this standard also mutually corroborates the progress of domestic research. In May this year, USTC's (University of Science and Technology of China) "Jiuzhang 4.0" optical quantum computer achieved Gaussian boson sampling with 1,024 squeezed state inputs and the participation of 3,050 photons, solving specific problems 54 orders of magnitude faster than the world's fastest supercomputer. Its optical path system extensively uses PPLN devices to complete frequency conversion and entanglement manipulation. In quantum key distribution networks, the wavelength of single photons must be adapted to the low-loss window of optical fibers, which also relies on such devices for efficient frequency conversion. Every upgrade in device performance elevates the entanglement quality and transmission distance of the entire machine. The standard unifies this performance language, giving the upstream and downstream of the industry chain a common blueprint for dialogue for the first time.
03. The Slow Craft of Crystal Growth: Bottlenecking the Entire Chain
The supply elasticity of lithium niobate is extremely low. Crystal growth relies on the Czochralski method, where compositional uniformity, optical uniformity, and defect control directly determine the yield of downstream devices. It takes years of ramp-up for a production line to go from construction to stable batch supply. The core of the thin-film process is ion slicing and wafer bonding, with process parameters measured in nanometers. Due to the different thermal expansion coefficients of lithium niobate and the substrate, film thickness drift, bubbles, and voids during high-temperature bonding are the main killers of yield. Between a single sample showing beautiful performance and stable shipment of three consecutive batches lies the most difficult stretch of the entire industry chain. Downstream customer certification cycles often take years; optical module manufacturers' validation of modulator chips involves reliability, temperature cycling, and lifespan tests, and once passed, suppliers are rarely changed. The global market has long been dominated by manufacturers such as Sumitomo Electric in Japan, and the pricing power for high-purity grades has not been in domestic hands for years.
The progress of domestic substitution has significantly accelerated after 2025. Tiantong Inc. has prepared 12-inch optical-grade lithium niobate wafers, becoming the only manufacturer globally, apart from Sumitomo in Japan, to master this size technology. In the thin-film wafer segment, Jinan Jingzheng uses ion slicing technology to rank among the top three globally in shipments of 4-inch products, while purchasing over 80% of its bulk materials from domestic crystal manufacturers; a domestic closed-loop of the chain is taking shape. Capacity expansion schedules are also advancing intensively: new capacities for crystals and wafers from domestic manufacturers are concentrated to land from the second half of 2026 to 2027. If the ramp-up proceeds as scheduled, the domestic wafer self-sufficiency rate will rise accordingly. Weaknesses are equally apparent: over 90% of the crystal raw material, niobium pentoxide, relies on imports, mainly from Brazilian mining giants, with only a few domestic enterprises capable of small-scale production; the purification process for high-purity grades still has a purity gap compared to Japanese manufacturers. The shortcoming of upstream raw materials is a lesson that must be made up in the next stage of this chain.
04. From Crystals to Devices: Domestic Positioning Largely Completed
In the upstream crystal and substrate segment, Tiantong Inc. (600330) is a listed enterprise in China that has achieved mass production of 8-inch optical-grade lithium niobate crystals and wafers. Its annual crystal capacity is about 1.3 million to 1.4 million pieces. The 12-inch wafers have been prepared, positioning for the next-generation demand of co-packaged optics (CPO). The company also operates in photovoltaic and magnetic material businesses. The revenue contribution of lithium niobate is still ramping up, but as a core domestic supplier of large-size upstream wafers, its positioning value continues to rise with downstream volume growth. CASTECH Inc. (002222) supplies high-purity nonlinear optical crystals, serving laser and quantum optical devices. NSIG (688126) has its holding subsidiary layout in lithium niobate single-crystal thin-film substrates, with optical-grade products continuing to ship in batches. The unlisted Jinan Jingzheng holds the core ion slicing process, has received investment from industrial capital, and is advancing its IPO process.
Midstream devices represent the thickest layer of value. Advanced Fiber Resources (AFR) (300620) started with fiber optic devices. Through the acquisition of an overseas established lithium niobate business and years of independent R&D, it has, according to publicly available industry data, become one of only three companies globally and the only one in China to mass-produce ultra-high-speed TFLN modulators. Its 70GHz and 110GHz products are delivered in batches, with high-end product yields exceeding 90% and a domestic market share of over 90%. Accelink Technologies (002281) covers single-photon detectors and quantum optical modules, serving as a core upstream partner for quantum computing. Tengjing Technology (688195) supplies precision quantum optical components, receiving research visits from 86 institutions within the year, making it one of the most institutionally watched targets in the quantum computing concept. HGTECH (000988) is laying out quantum dot lasers and single-photon sources.
Downstream applications span three directions: optical modules and optical interconnects are undertaken by module leaders such as Innolight (300308); quantum frequency conversion and key distribution correspond to complete machine enterprises such as QuantumCTek (688027); while LiDAR and AR optics represent a third curve still in incubation (the above sorting does not constitute investment advice).
05. Orders Scheduled Through Year-End: Initial Signs Emerging at the Financial Statement Level
Demand-side data is turning from expectations into orders. The gross profit margin of AFR's TFLN modulators is maintained in the range of 48% to 55%. Backlog confirmed orders exceed CNY 1.6 billion, with schedules locked through the end of 2026. Capacity planning by the end of 2026 is upgraded to 1 million units per year, and further to 2 million units in 2028. Single-wavelength 200G products have passed the certification of overseas leading computing power manufacturers. Modulators for 1.6T optical modules are already supplied in batches, and 3.2T products are in the sample validation stage. Tiantong Inc. has signed a long-term supply agreement with Innolight, with an annual supply scale reaching the level of 300,000 pieces, and 8-inch wafers are in the capacity ramp-up period. Market calculations show that the global supply and demand gap for TFLN wafers is about 40%, and the price per piece once rose to CNY 28,000, an increase of nearly 7 times compared to before.
The transmission chain can also be clearly calculated. Modulator chips account for about 10% to 15% of the total cost of optical modules. The impact of wafer price increases on the overall module cost is limited, and downstream absorption is smooth. This means that upstream shortages can be directly translated into profits in the crystal and device segments without worrying about demand being crushed by prices. Although new scenarios such as LiDAR and AR optics currently have limited scale, they all rely on lithium niobate devices to improve ranging accuracy and optical processing capabilities, providing a third layer of demand redundancy for this chain beyond optical communications and quantum. Although the contribution of quantum is still small, the direction is clear: after the implementation of national standards, device selection by research institutions and industrial customers has a unified basis. Procurement shifts from sporadic customization funded by research grants to verifiable standardized products. The significance of this curve lies in opening up incremental growth, rather than current scale.
06. Outlook: Two Market Trends Sharing One Production Line
Institutional judgments focus on the duality of demand.
Research reports from multiple brokerages list TFLN as the upstream intersection of AI optical interconnects and quantum technology, believing its value lies in "one process, two markets." Optical modules provide current revenue elasticity, while quantum and LiDAR provide long-term valuation imagination.
Space calculations also support this judgment: CAICT (China Academy of Information and Communications Technology) expects the domestic quantum computing industry scale to exceed CNY 20 billion in 2026 (in RMB terms). Statistics from ICV show that the global quantum computing industry scale is expected to grow from about USD 5 billion in 2024 to nearly USD 220 billion by 2030. The statistical scope and currencies of the two sets of data are not the same; the former looks at the domestic market, while the latter is on a global scale, so they should not be directly compared. The iteration cycle of the AI optical module market is faster, with rate upgrades advancing almost on a yearly basis.
The calm side is equally clear: TFLN modulators account for about 10% to 15% of the total cost of optical modules, leaving limited room for price increase transmission. The valuations of leading device companies have already priced in relatively full capacity expectations. If the ramp-up or validation pace falls short of expectations, volatility will arrive before the financial statements. Alternative routes such as silicon-based heterogeneous integration are still advancing, and the landscape is far from finalized.
But for this industry chain, the fact that two market trends share one production line is in itself a scarce situation. AI determines how high it can fly, quantum determines how long it can fly, and standards and orders determine how steadily it flies. When national standards hand over quantum procurement from empirical judgment to benchmarks, and when the iteration of optical modules pushes modulator bandwidth to the limit, this thin crystal is being tested by the demands of two eras simultaneously.