The Unsung Foundation of the AI Revolution: A Deep Dive into Copper Clad Laminates (CCLs)

The High-Tech Bedrock of the AI Brain

Imagine a futuristic metropolis built exclusively for Artificial Intelligence (AI). High-performance processors like NVIDIA's GPUs or Google's TPUs are the towering skyscrapers, the core of its computational power. Memory modules are the sprawling residential districts, storing vast oceans of data. The millions of intricate copper traces on a Printed Circuit Board (PCB) are the superhighway network connecting everything.

In this grand urban plan, what is the role of the Copper Clad Laminate (CCL)? It is the very land and bedrock upon which this entire city is built. The quality of the CCL directly determines whether the superhighways can handle light-speed traffic without collapsing, whether the skyscrapers can be built tall and stable, and whether the entire city can operate at peak efficiency without overheating and shutting down. It may seem like an unremarkable "board," but it is the fundamental substrate of the entire electronic system, and its performance dictates the capabilities, reliability, and cost of AI hardware.   

A Copper Clad Laminate is a composite material created by impregnating a reinforcing material (like glass fiber cloth) with a resin (such as epoxy), which is then clad on one or both sides with copper foil and laminated under high temperature and pressure. Its core functions are twofold: first, to provide robust mechanical support for all electronic components, and second, to serve as the canvas upon which conductive traces are etched, building the bridges for electronic signal transmission.   

The central thesis of this article is this: the unprecedented computational demands brought by AI are not just challenging the limits of chip manufacturing but are fundamentally reshaping the requirements for foundational materials like CCLs. The revolution ignited by AI has sparked a fierce technological arms race in the field of materials science, creating a highly differentiated and extremely valuable new market segment. In this global race, Taiwan's supply chain is playing a pivotal and indispensable role.   

The AI Spectrum: Differentiated Demands from Cloud Training to the Intelligent Edge

The term "AI" encompasses a vast range of applications, from massive cloud data centers to tiny IoT devices. Each domain places distinctly different demands on the underlying hardware and its foundational material, the CCL. Understanding these differences is the key to grasping the pulse of the CCL market.

The Cloud Titans: AI Training and High-Performance Computing (HPC)

The Challenge: Training cutting-edge AI models, such as Large Language Models (LLMs), is an incredibly resource-intensive process. It involves repeatedly processing enormous datasets through models composed of hundreds or even thousands of parameters. This means servers must be packed with multiple high-performance GPUs, like the NVIDIA H100 or GB200, and made to work in concert with extreme efficiency. The key to the entire system's success is the ability to transfer massive amounts of data between GPUs and between GPUs and memory at lightning speed. Any delay or error could render millions of dollars in training costs worthless.   

CCL Requirements:

• Ultra-Low Signal Loss: To ensure data integrity over ultra-high-speed interfaces like PCIe 5.0/6.0 or NVIDIA NVLink, the CCL material's Dissipation Factor (Df) must be exceptionally low. Any energy loss (attenuation) or waveform distortion during signal transmission can lead to computational errors, potentially crippling the entire system.   

  
  

• High Layer Count & Density: The designs of AI server motherboards and GPU accelerator modules (like the Universal Baseboard, UBB, and OAM modules) are exceedingly complex. PCB layer counts often exceed 20 layers to accommodate all necessary routing in a limited space. This demands that CCL materials possess excellent dimensional stability and can withstand multiple high-temperature, high-pressure lamination cycles without warping or delaminating.   

  
  

• Superior Thermal Management: These servers are veritable power hogs, generating astonishing amounts of heat during operation. The CCL material must have high thermal conductivity and a very high Glass Transition Temperature (Tg) to effectively dissipate heat away from the chips and prevent the PCB from bending or delaminating at high temperatures, which would lead to component failure.   

  
  

• Impeccable Reliability: AI training servers are required to run 24/7 under continuous high load. The CCL material must be extremely reliable and resistant to long-term failure risks like Conductive Anodic Filament (CAF) formation. CAF is a phenomenon where tiny conductive pathways form between glass fibers in high-temperature, high-humidity environments, causing short circuits.   

  
  

The Front Lines: AI Inference Workloads

The Challenge: Inference is the stage where an AI model is put into practical application, such as a chatbot generating a real-time response or a video platform recommending content you might like. The core objectives for inference tasks are low latency (fast response time) and high efficiency, as these requests can occur billions of times a day, making the cost and power consumption of each computation critically important.   

CCL Requirements:

• Balanced Performance and Cost: While inference servers also require high-speed signal transmission, they don't necessarily need the most expensive, lowest-loss materials used in top-tier training clusters. The market focus shifts to achieving the optimal balance between performance, power consumption, and unit cost, especially for large-scale deployments.   

  
  

• Optimized for Throughput: The CCL material needs to support designs that maximize "inferences per second per watt." This still implies a need for low-loss materials, but perhaps a tier below the "ultra-low loss" or "extreme-low loss" grades used for training, such as a "very low loss" material.

• Equally High Reliability: Like training servers, inference servers deployed in data centers must operate continuously. Therefore, the material's thermal stability and long-term reliability remain essential considerations.   

  
  

AI Everywhere: Edge Devices

The Challenge: Edge AI refers to running inference directly on the device where data is generated, such as the decision-making computer in an autonomous vehicle, a smart camera in a factory, or the smartphone in your hand. The operating environments for these devices are extremely demanding, facing multiple constraints of power, physical space, and harsh conditions (e.g., high temperatures, vibration).   

CCL Requirements:

• Extreme Power Efficiency: Low-loss CCLs are equally important here because less signal energy loss translates to lower power consumption, which is critical for battery-powered devices or those with limited cooling space.   

  
  

• Compact Form Factors: The drive for miniaturization in edge devices has fueled the demand for High-Density Interconnect (HDI) PCBs. HDI technology utilizes CCLs that support microvias and extremely fine traces. Furthermore, flexible CCLs (FCCLs) based on materials like Polyimide are indispensable for applications that need to bend or conform to irregular shapes, such as wearable devices.   

  
  

• Durability and Reliability: Automotive and industrial applications require CCLs with high mechanical strength, resistance to vibration, and the ability to maintain stable performance under extreme temperature and humidity fluctuations.   

  
  

• Cost-Effectiveness: For mass-produced consumer electronics, the cost of the CCL is a significant part of the overall budget, necessitating a smart trade-off between performance and price.   

  
  

This differentiation in application requirements is not static; it is driven by profound industry forces. First, AI hardware development exhibits a "symbiotic upgrade cycle." When NVIDIA launches a GPU like the GB200 with a staggering 1.8 TB/s of bandwidth , its immense power cannot be harnessed without PCBs and CCLs capable of carrying such high-speed signals without severe degradation. This compels CCL manufacturers like Elite Material, ITEQ, and TUC to accelerate the development of next-generation low-loss materials (e.g., M8, M9 grades). The availability of these advanced materials, in turn, gives chip designers like NVIDIA the confidence to plan even faster next-generation processors, creating a co-dependent, spiraling evolution of chips and materials.   

Second, the value of a CCL is not uniform but forms a "value spectrum" directly correlated with the "value of the data" it processes. In a multi-billion dollar AI training center, the primary goal is to shorten training time, as each extra day represents enormous costs in electricity and computing resources. In this context, the added expense of using the most premium "extreme low loss" CCL is negligible compared to the cost of a failed or delayed training run. Conversely, for a mass-market smart home device, the value of a single inference is minuscule, making cost-effectiveness the top priority. Thus, a standard or mid-loss CCL is the logical choice. This demonstrates that CCL selection is not just a technical decision but an economic one, creating a clear value hierarchy from the data center (high-value data, high-price CCL) to the edge (low single-instance value, cost-sensitive CCL).   

Finally, the traditional dichotomy of a powerful cloud and a weak edge is blurring, giving rise to an emerging category of "high-performance edge computing." Applications like autonomous vehicles  or on-premise enterprise AI servers , while at the "edge," need to perform complex real-time computations comparable to cloud servers. This creates a hybrid demand: they require data-center-grade high-performance CCLs (low loss, high speed) combined with the robustness of traditional edge devices (high-temperature resistance, anti-vibration) and form factor constraints. This has opened up a lucrative and technically challenging new battleground for CCL suppliers.   

A Guide to High-Performance CCLs: Matching Materials to Applications

To meet the stringent demands across the AI spectrum, the CCL industry has developed a sophisticated system for grading materials. Understanding the language of this system is key to translating application needs into specific product choices.

Decoding the Language of Loss: Dk, Df, and Tg Explained

To evaluate the performance of a CCL, the industry focuses on three core physical parameters that collectively determine signal transmission quality and material stability.

• Dielectric Constant (Dk or Er​): This can be understood as an indicator of how much an electrical signal's speed is "slowed down" as it passes through the material. For high-speed digital signals, a lower and more stable Dk value across different frequencies is crucial. It ensures precise signal timing and makes it easier to control circuit impedance, preventing signal distortion from reflections.   

  
  

• Dissipation Factor (Df or tanδ): This parameter measures the "leakiness" of the material—how much signal energy is absorbed by the material and converted into heat during transmission. The lower the Df value, the less energy the signal loses (attenuates). For applications like AI servers that require long-distance, high-frequency transmission, Df is the single most critical performance metric.   

  
  

• Glass Transition Temperature (Tg): This refers to the temperature at which the resin in the CCL transitions from a rigid "glassy" state to a softer "rubbery" state. A higher Tg value means the material is more heat-resistant, which is vital for withstanding the immense heat generated by high-power chips and for surviving the 260°C+ temperatures of lead-free soldering processes. A high Tg ensures the PCB does not warp or delaminate at high temperatures, maintaining its structural integrity.   

  
  

The Ladder of Speed: The Evolution from FR-4 to M-Grade

Based on these key parameters, especially the Df value, CCL materials form a clear performance pyramid. Server platform giants like Intel have also proposed an "M-scale" material grading system, providing a clear upgrade path for the entire industry.

• Standard Loss: Representative material is FR-4, with a Df typically greater than 0.01. This is the cornerstone of consumer electronics—inexpensive but incapable of meeting the high-speed demands of AI servers.   

  
  

• Mid Loss: Performance is between standard and low loss, used in some non-critical server components or cost-sensitive networking equipment.
  

• Low Loss (LL): Df is approximately between 0.005 and 0.01. This is the entry-level material for server and networking applications, corresponding to Intel's M4 server platform grade.   

  
  

• Very Low Loss (VLL): Df is further reduced, corresponding to the M6 grade. This was the choice for previous-generation server platforms and some current mainstream networking equipment.
  

• Ultra Low Loss (ULL): Df is less than 0.005. This is the core material for current mainstream AI servers (e.g., platforms with NVIDIA H100/A100 GPUs), corresponding to M7 and M8 grades.   

  
  

• Extreme Low Loss (ELL) / Super Low Loss (SLL): This is the apex of the CCL pyramid, boasting the lowest Df values. It is engineered for the most cutting-edge applications, such as next-generation AI accelerators (like the NVIDIA Blackwell GB200) and 800G/1.6T ultra-high-speed switches, corresponding to M9 and higher grades.   

  
  

Practical Application and Material Matching

By mapping the aforementioned application requirements to these material grades, we can create a clear, practical guide:

• AI Training Servers (e.g., NVIDIA DGX Platform):

  • Key Boards: GPU Universal Baseboard (UBB), OAM accelerator modules, high-speed switch boards.   

  • CCL Requirement: These boards have layer counts of 20 or more and must use Ultra Low Loss to Extreme Low Loss grade materials (M7, M8, M9+). Industry examples include Elite Material's EM-892K, ITEQ's M6/M7/M8 series, and TUC's high-end ThunderClad series.   

    
    

• High-Speed Networking Equipment (800G/1.6T Switches):

  • The Challenge: As the "central nervous system" of the data center, switches connect all AI servers. Their signal integrity requirements are even more stringent than those of the servers themselves.

  • CCL Requirement: Must use the highest-performing Extreme Low Loss materials available (M9+). This is a key battleground for top-tier suppliers like TUC and Elite Material. Panasonic's Megtron 8 from Japan is also a leader in this segment.   

    
    

• General Purpose & Inference Servers:

  • Key Boards: CPU motherboards.

  • CCL Requirement: Layer counts are typically between 12 and 18. They use Low Loss to Very Low Loss grade materials (M4-M6 grades). This is a more cost-sensitive but high-volume market.   

    
    

• Edge AI (Automotive Radar & ADAS):

  • The Challenge: Must operate at extremely high frequencies (e.g., 77 GHz) and requires exceptional environmental reliability.

  • CCL Requirement: Uses special high-frequency materials, typically based on Polytetrafluoroethylene (PTFE). The core requirement is a stable Dk value over a wide temperature range. Rogers Corporation's RO3000/RO4000 series is the undisputed leader in this field.   

    
    

Behind this technological evolution, Intel's M-Grade classification has acted as a market "pacemaker." When a platform leader like Intel or NVIDIA announces that its next-generation product (e.g., a platform supporting PCIe Gen 6) will require M7-grade materials , it sets a clear, unified target for the entire supply chain. This signal cascades down: PCB manufacturers must validate their M7 process capabilities, and CCL suppliers must prepare sufficient capacity for M7-grade products. This creates predictable, wave-like upgrade cycles that can be tracked by investors and analysts. It simplifies complex material physics into an easy-to-understand grade, facilitating specification by OEM customers and product positioning by suppliers. The focus of competition then shifts to who can provide the best M7 (or M8, M9) material in terms of performance, cost, and capacity.   

Furthermore, an often-overlooked detail is that a CCL's "loss" comes not only from the resin and glass fiber but also from the surface roughness of the copper foil. At frequencies reaching tens of GHz, the signal travels primarily along the surface of the copper trace (the "skin effect"). A rough copper surface increases the effective path length of the signal, thereby increasing loss. Therefore, achieving "ultra-low loss" performance requires not just an advanced resin formulation but also pairing it with extremely smooth copper foil (e.g., VLP - Very Low Profile Copper). This adds another layer of complexity and cost to high-end CCL manufacturing and highlights the close collaboration needed between CCL makers and specialized copper foil suppliers. It is a dual solution of "advanced dielectric + smooth copper."   

The Global Ecosystem: Mapping the High-End CCL Supply Chain

The high-end CCL market is not a perfectly competitive open market but a specialized domain with extremely high technological barriers and a high concentration of players. Understanding its global supply chain structure is essential for discerning the industry's power dynamics.

Building from the Ground Up: Key Upstream Raw Materials

The performance foundation of a CCL depends on three key upstream raw materials, the supply of which is also highly concentrated.

• Specialty Glass Fiber Fabric: As the "skeleton" of the CCL, high-end CCLs require low dielectric constant (Low Dk) glass fabric to reduce signal loss. Its supply is dominated by a few Japanese and Taiwanese manufacturers. For example, Japan's Nittobo is a key supplier to TUC , while Taiwan's Taiwan Glass and F.C.F. Fishery also play important roles. Recent shortages in high-end glass fabric have become a major bottleneck, limiting CCL production and driving up prices.   

  
  

• High-Performance Resins: As the "flesh" of the CCL, the chemical formulation of the resin is the core intellectual property of each CCL manufacturer, directly determining the material's electrical and thermal properties.   

  
  

• High-Speed Copper Foil: As mentioned, smooth VLP copper foil is critical for reducing high-frequency loss. This is another highly specialized upstream market.

The Titans of CCL: A Tour of the Global Giants

Besides Taiwanese firms, the global market is co-dominated by several American and Japanese giants, who are the primary competitors or partners of their Taiwanese counterparts.

• United States:

  • Rogers Corporation: The undisputed leader in the high-frequency materials space, specializing in radio frequency (RF) applications for automotive radar, aerospace, etc. Its PTFE-based RO3000 and RO4000 series are industry benchmarks. Rogers has also entered the high-speed digital market with its XtremeSpeed™ RO1200™ series, targeting top-tier servers.   

    
    

  • Isola Group: Another major player with a broad product line. Its Tachyon, I-Tera, and TerraGreen high-speed digital materials are precisely targeted at the server, networking, and 5G markets that drive AI development.   

    
    

• Japan:

  • Panasonic: A giant in the electronic materials field. Its "MEGTRON" series (especially Megtron 6, 7, and 8) is a world-leading brand for ultra-low loss materials, widely used in high-speed servers and switches. It is the number one rival to Taiwan's "Big Three" in the highest-end market segment.   

    
    

  • Resonac (formerly Hitachi Chemical): Another significant Japanese supplier of high-performance materials, also holding a solid position in the global high-end market.   

    
    

The Source of Demand: Architects of AI Hardware

The development pace of the entire CCL industry is ultimately set by its downstream end customers.

• Cloud Service Providers (CSPs): Google, Amazon (AWS), Microsoft, Meta, etc. They are not only the largest buyers of AI servers but are also increasingly designing their own custom AI chips (ASICs). This is reshaping the traditional supply chain, creating new opportunities and challenges for CCL manufacturers.   

  
  

• AI Chip Design Companies: NVIDIA is the 800-pound gorilla in the market. The specifications for its GPU platforms (like Hopper and Blackwell) have become the de facto standard for CCL material requirements. AMD and Intel are also formidable forces in the market with their GPU and accelerator products.   

  
  

A deeper analysis of the global supply chain reveals a highly concentrated "high-tech oligopoly." The number of manufacturers capable of producing the high-end, low-loss CCLs required for AI is very small, consisting mainly of the aforementioned Panasonic, Rogers, Isola, and Taiwan's Big Three. Developing a new resin formulation with the target Dk/Df properties, ensuring it can be mass-produced with stability, and passing the rigorous qualification processes of giants like NVIDIA or Google requires years of time and enormous R&D investment. This creates extremely high barriers to entry, allowing the few leading players to compete for the most lucrative orders within this "walled garden" and giving them significant pricing power during raw material shortages.   

Meanwhile, geopolitical considerations are profoundly reshaping this supply chain. Major Taiwanese CCL manufacturers like TUC  and ITEQ  have announced plans to build factories in Thailand. Their official reasoning is to align with the "supply chain diversification plans of European and American OEM and PCB customers"  and to respond to "changes in the global supply chain". This clearly reflects the "China Plus One" strategic trend. To ensure supply chain resilience, customers are demanding that their suppliers reduce their dependence on a single geographic region. This is not just business expansion but a strategic geopolitical maneuver aimed at mitigating risk and aligning with the priorities of their key Western customers. While this increases operational costs, it makes them more attractive long-term partners.   

Taiwan's Pivotal Role: A Superpower in the CCL Industry

In the global race for AI hardware, Taiwan is not just the center for wafer fabrication and chip packaging; it also plays an indispensable role in the upstream CCL materials sector. The companies known collectively as Taiwan's "CCL Big Three"—Elite Material Co. (EMC), ITEQ Corporation, and Taiwan Union Technology Corp. (TUC)—form the core of the global high-end CCL supply chain.

A Deep Dive into Taiwan's "Big Three"

• Elite Material Co., Ltd. (EMC):

  • Market Position: EMC is widely recognized as the global leader in halogen-free environmentally friendly substrates and is a key supplier in NVIDIA's AI server supply chain. Its status as an exclusive or primary supplier for certain NVIDIA platforms gives it a significant market advantage and pricing power.   

    
    

  • AI Product Focus: The company's EM-890K and EM-892K series are ultra-low loss materials designed specifically for high-speed applications like AI, 5G, and HPC. The EM-891K targets even higher-end applications with per-lane speeds of 40 to 50 Gbps.   

    
    

  • Core Strategy: To leverage its deep ties with the market leader (NVIDIA) and its long-term technological expertise in high-performance green materials to continuously solidify its leading position in the AI server market.   

    
    

• ITEQ Corporation:

  • Market Position: ITEQ is a manufacturer with an extremely broad product line, covering everything from standard-loss to extreme-low-loss grades. It is a major supplier to multiple server platforms, with customers including NVIDIA (for L40S servers) and several cloud giants with their own custom ASIC chips.   

    
    

  • AI Product Focus: The company is actively shipping M6, M7, and M8 grade high-speed materials to AI GPU/ASIC accelerator card customers. Its M9 grade material, corresponding to 1.6T switch speeds, is already being sampled by major end customers for qualification, showing that its technology is keeping pace with the market's cutting edge. Its IT-988G is an    

    
    

    ultra-low loss material for 100G/400G solutions , and the IT-968 is another high-end ultra-low loss product.   

    
    

  • Core Strategy: To pursue a customer diversification strategy, with business spanning NVIDIA, CSP custom chips, automotive electronics, and network communications to mitigate risk. Actively advancing the construction of its Thailand plant is a key step in its global expansion.   

    
    

• Taiwan Union Technology Corp. (TUC):

  • Market Position: TUC has long been a powerhouse in the traditional server and switch markets. In recent years, it has emerged as a strong challenger, aggressively penetrating the top-tier AI server market. According to data from the research firm Prismark, TUC's high-speed CCL revenue ranked third globally in 2023, with a market share of about 16.3%.   

    
    

  • AI Product Focus: The company has successfully entered the AI ASIC server projects of a major US-based cloud service provider. Shipments of its M7 and M8 grade high-end CCLs are steadily increasing, and it is also actively developing M9 grade products. Its "ThunderClad" and "PegaClad" product series cover the full spectrum from low-loss to extreme-low-loss applications, serving both high-speed digital and RF markets.   

    
    

  • Core Strategy: To leverage its strong technical capabilities, possibly combined with a more competitive pricing strategy , to aggressively capture market share in the highest-end segments, especially in 800G switches and custom AI ASIC servers. Its new plant in Thailand is designed specifically to support complex, high-layer-count boards (30+ layers), laying the foundation for future growth.   

    
    

Competitive Dynamics and Strategic Outlook

The competition and cooperation among Taiwan's Big Three collectively shape the landscape of the global high-end CCL market.

• The Battle for 800G/1.6T Switches: The switch market is the most profitable and technically demanding battleground. Success here requires the most advanced M9+ grade material technology, and TUC and EMC are considered strong contenders in this arena.   

  
  

• The Opportunity in Custom ASICs: As cloud giants like Google and Amazon increasingly adopt their own custom chips, this opens new doors for CCL makers who are not the primary suppliers to NVIDIA. TUC is believed to have achieved significant market share growth in this emerging area.   

  
  

• The Constraint of Upstream Materials: The shortage of high-end glass fabric is a severe challenge faced by all CCL manufacturers. Whoever can secure a stable supply of upstream raw materials is likely to come out ahead in the market competition over the next few years. This makes the relationship between CCL makers and raw material suppliers like Nittobo extremely strategic.   

  
  

In this oligopolistic market, a "second source" strategy has become a key path for challengers to get ahead. EMC established its leadership by becoming the primary supplier for NVIDIA's flagship AI servers. However, large customers, for risk mitigation reasons, are highly averse to a single-supplier situation. This creates a huge opportunity for competitors. ITEQ and TUC are actively positioning themselves as qualified second sources for NVIDIA while striving to become the primary suppliers for other platforms (like the L40S or CSP custom ASICs). By proving their technical capabilities and mass production reliability on the most demanding products, they can build market credibility and potentially leap from "second source" to "primary source" in the competition for the next generation of products. This "path from second to primary" is a crucial strategic route in the high-end CCL market.   

On the surface, Taiwan's Big Three may seem to be making similar products, but a deeper analysis reveals they each have a unique "corporate personality" and strategic posture. EMC is the "Incumbent Leader," fully leveraging its market leadership and deep integration with top GPU designers. Its core strategy is to defend its existing territory and maintain its technological lead. ITEQ is the "Diversified Generalist," maintaining an extremely broad product portfolio across multiple markets like automotive, servers, and consumer electronics. Its strategy is to spread risk and avoid over-reliance on a single customer or market. TUC plays the role of the "Aggressive Challenger," appearing to use its technical strengths in the switch market and more flexible pricing to attack the highest-end market segments, aiming to seize share from the incumbent leader, especially in the booming custom ASIC server space. Understanding these different strategic positions is crucial for predicting future market share shifts.

Conclusion and Future Outlook

The wave of AI is reshaping every corner of the technology industry at a breathtaking pace. As a foundational material for this revolution, the importance of the Copper Clad Laminate (CCL) is becoming increasingly prominent. From the cloud to the edge, different AI applications present distinct yet equally demanding requirements for CCLs, driving rapid evolution in materials science and creating a global supply chain dominated by a few technology giants, in which Taiwanese manufacturers play an indispensable core role.

Key Takeaways and Summary Table

To clearly present the core findings of this report, the following table systematically organizes and maps AI applications, hardware requirements, CCL specifications, and key suppliers. For investors or industry strategists, this table provides a macroscopic view, offering an at-a-glance understanding of the entire industry landscape—which markets have the highest demands, what grade of materials are needed, and which companies are well-positioned in their respective lanes.

AI Application Area	Key Hardware	Key CCL Requirements	Required Loss Grade (Intel M-Scale)	Representative CCL Products	Key Global Suppliers	Key Taiwanese Suppliers
Cloud: AI Training (Cutting-Edge)	NVIDIA GB200, Custom ASICs	Extreme speed, >24 layers, very high thermal reliability, lowest signal loss	Extreme Low Loss (ELL) / M9+	Next-gen Megtron, New EMC/TUC/ITEQ series	Panasonic, Rogers	EMC, TUC, ITEQ
Cloud: AI Training/HPC (Current Mainstream)	NVIDIA H100/A100, AMD MI300X	High speed, >20 layers, very low signal loss, high Tg	Ultra Low Loss (ULL) / M7-M8	Panasonic Megtron 8, EMC EM-892K, ITEQ IT-988G, TUC ThunderClad 4, Isola Tachyon	Panasonic, Isola	EMC, ITEQ, TUC
Cloud: High-Speed Networking	800G / 1.6T Switches	Absolutely lowest signal loss, highest signal integrity, >30 layers	Extreme Low Loss (ELL) / M9+	TUC ThunderClad 4HZ, ITEQ M9 (sampling), Panasonic Megtron 8	Panasonic	TUC, EMC, ITEQ
Cloud: General Purpose/Inference Servers	Intel Xeon (Eagle Stream), AMD EPYC	Balance of cost & performance, 12-18 layers, good thermal reliability	Low Loss (LL) to Very Low Loss (VLL) / M4-M6	Isola I-Speed, TUC TU-872 SLK, ITEQ IT-170GRA1TC	Isola, SYTECH	TUC, ITEQ, EMC
Edge: Automotive Electronics	ADAS Processors, 77GHz Radar	High reliability, high-frequency performance, wide temp range, anti-vibration	High-Frequency Materials (PTFE-based) / Low Loss	Rogers RO4000/RO3000 Series, TUC PegaClad	Rogers, Panasonic	TUC, ITEQ
Edge: High-Performance (Industrial, On-Prem AI)	NVIDIA Jetson, Compact GPU Systems	Compact (HDI), power efficient, high reliability	Low to Mid Loss, special form factors	General Low/Mid Loss HDI materials	Isola, SYTECH	All major Taiwanese suppliers
Edge: Consumer IoT	Low-power MCUs, SoCs	Cost-effectiveness, miniaturization, low power	Standard Loss (FR-4)	Standard FR-4	Kingboard, SYTECH	N/A (not a high-end market)

The Road Ahead: Future Trends in AI and CCLs

Looking forward, the symbiotic evolution of AI and CCLs will be driven by several key trends:

• The Pursuit Beyond 1.6T: The race for faster network transmission speeds is endless. With the development of 224G and even higher-speed SerDes technology, the demands on material loss will reach unprecedented extremes, continuously driving innovation in materials science.
  

• The Rise of Co-Packaged Optics (CPO): When traditional electrical interconnects hit physical bottlenecks in bandwidth and power consumption, CPO technology—which involves packaging optical components directly with silicon chips on the same substrate—will become mainstream. This will present entirely new challenges for substrate materials, requiring them to possess excellent electrical and optical properties simultaneously, giving rise to new hybrid material solutions.
  

• The Wave of Sustainability and Halogen-Free: Increasingly stringent global environmental regulations and corporate emphasis on ESG (Environmental, Social, and Governance) will continue to drive market demand for green materials like halogen-free substrates. In this regard, manufacturers like Elite Material, which have long cultivated the halogen-free market, will have a sustained competitive advantage.   

  
  

• The Supply Chain as Strategy: The realities of geopolitical risk and upstream material bottlenecks have elevated the importance of supply chain management to an unprecedented strategic level. In the coming decade, the winners in the CCL industry will be determined not only by their R&D prowess but also by the resilience of their supply chains, the wisdom of their global footprint, and their strategic relationships with key upstream partners.

In conclusion, the Copper Clad Laminate has transformed from a traditional passive component into a key enabling technology driving the AI revolution. This transformation, led by data, computation, and speed, is pushing the CCL industry to the pinnacle of technology and the forefront of the market as never before.