Recently, the public account has changed its rules, and articles are no longer pushed in chronological order. If you don’t want to miss articles from the “Thinking Summary Account“, please set it as a “star” and frequently click “See”. Thank you for your support!
This public account is a backup account for the “Thinking Summary Society”, channel ↓↓↓
Content on both sides is inconsistent, please do not miss the wonderful contentAll content comes from the internet. If there is any infringement, please contact the backend for deletion. Thank you for your understanding.
Full Text Summary
With the surge in AI computing power demand, the PCB (Printed Circuit Board) industry is facing unprecedented opportunities and challenges. The rapid development of fields such as AI servers, GPUs, and ASICs has significantly increased the demand for high-performance PCBs, driving the PCB market towards the application of high-density, multi-layer, and new material technologies. Market analysis indicates that the PCB industry is experiencing a dual growth in supply and demand, with value assessments showing a continuous rise in industry value and a clear trend of technological upgrades, indicating that domestic and foreign PCB companies will accelerate production expansion to meet market demand. However, the efficiency of capacity release has not fully kept pace with the accelerated pace of AI demand, suggesting that the PCB industry will continue to develop over the next 1 to 2 years, with value further increasing. Additionally, the application of HDI, high multi-layer boards, PCB backplanes, and new materials is continuously optimizing PCB performance to meet increasingly complex circuit design requirements, injecting new vitality into the PCB industry.
Trends of Quantity and Price Rise in the PCB Industry Driven by AI Computing Power
The conference focuses on the development trends of the PCB industry over the next 1 to 2 years against the backdrop of strong AI computing power demand, analyzing the supply-demand gap and technological directions, emphasizing the logic of quantity and price increase, and optimistic about the continuous development of the PCB industry and opportunities for market share increase under the trends of technological and AI upgrades.
Analysis of PCB Market Demand and Technological Evolution from 2025 to 2027
The report predicts that the PCB market demand will reach $5.6 billion by 2025, maintaining rapid growth until 2027. The increase in AI computing power and the growth in ASIC GPU demand are driving the market, especially the application of high multi-layer boards and high-frequency high-speed materials. New technologies such as the Kewo mainboard will replace traditional interconnection methods, driving market expansion. The report provides a detailed analysis of market demand, technological evolution, and future growth points.
GPU Market Forecast and PCB Value Analysis
The dialogue discusses in detail the GPU market forecast, including the shipment volume and price changes of the H series, B series, GD200, and GD300, analyzing the impact of the transition from the H series to the GD300 architecture on PCB value, emphasizing the dual-driven growth model of quantity and price, and the enhancement of value through modular design.
Forecast of PCB and ASIC Demand Surge from 2024 to 2026
The dialogue elaborates on the significant growth of PCB and ASIC demand in the coming years, especially between 2024 and 2026, where NVIDIA’s PCB demand is expected to surge from $850 million to $2.64 billion. At the same time, the capital expenditure of the four major North American cloud service providers has been significantly increased, with total capital expenditure expected to reach $366 billion in 2025, a nearly 50% year-on-year increase, and a further growth of nearly 20% in 2026. This growth is mainly driven by the acceleration of AI commercialization, which continues to drive the demand for computing power infrastructure and ASICs.
Analysis of ASIC Market Growth and PCB Demand
The report points out that the ASIC market size reached $55.7 billion in 2018, with significant growth in the shipment volume of ASIC chips from manufacturers like Google and Amazon, driving up PCB demand. It is expected that the shipment volume of ASICs and MDs will reach 8.1 million units in 2023, with PCB market demand reaching $4.5 billion, and the market for network communication equipment should not be overlooked.
Analysis of PCB and Optical Module Market Driven by AI and Supply-Demand Forecast
The dialogue provides a detailed analysis of the growth trends of the PCB and optical module markets driven by AI, predicting that the AI-related PCB market size will reach $5.6 billion by 2025 and $10 billion by 2026. The demand for optical modules is expected to grow from $1.6 billion to $3 billion, with a lower unit price but a large shipment volume. On the supply side, domestic PCB companies are accelerating production expansion, but factories in Southeast Asia may experience delays in capacity release due to supply chain issues, and a supply-demand gap is expected to persist. High-end materials such as HDI and high multi-layer boards are in high demand, with suppliers from Taiwan dominating, but capacity allocation may affect server applications. In the future, PCB manufacturers have broad space for new material applications, with rapid market growth, but attention must be paid to capacity matching and supply chain stability.
Discussion on HDI and High Multi-Layer Board Technology Routes and Backplane Connection Solutions
The discussion covers the advantages and disadvantages of HDI and high multi-layer boards in terms of heat dissipation, signal transmission, and cost, pointing out that high multi-layer boards may become the preferred choice for new technology introduction due to their low cost, high yield, and good heat dissipation performance. At the same time, the trend of backplane connections shifting from copper cables to PCBs is analyzed, emphasizing the requirements for rigid backplanes in high-density integration and signal transmission, predicting that this will bring a market increment of $2 billion, but copper cables still have irreplaceable advantages in high-speed signal integrity and medium-distance transmission.
Development and Market Analysis of PCB Substrate Technology
The report delves into the development of PCB substrate technology, shifting from traditional ABF substrates to SLP substrates, analyzing the market landscape and future trends, pointing out that Taiwanese companies dominate the market, while mainland Chinese companies have lower production capacity. The advantages of PCB substrates over HDI are discussed, including narrower line spacing, faster signal transmission speeds, and lower losses, while challenges in line width reduction and CTE thermal expansion coefficient solutions are also raised. The report also looks forward to material innovations, such as the application prospects of Ma9 materials, PTFE, and quartz cloth, emphasizing the importance of new materials in enhancing PCB performance, and suggesting that investors conduct in-depth research based on the report’s content.
Q&A Review
In the context of strong AI computing power demand, what market trends in the PCD segment of the PCB industry are worth noting?
In the current environment of strong AI computing power demand, the PCD industry shows a logic of simultaneous quantity and price increase. From a supply-demand perspective, in the next 1 to 2 years, due to the accelerated growth of AI demand and the lagging efficiency of PCD capacity release, a supply gap is expected in the market. By 2026, the PCB industry is expected to have a market increment of up to $10 billion, with overall value increasing faster than the global AI server growth rate, especially in terms of quantity growth speed. Therefore, we remain optimistic about the continuous development of the PCB industry and the opportunities for market share increase for related companies under the trends of technological and AI upgrades.
What are the main contents covered in the in-depth report released by Zhongning Technology’s team on the PCB sector?
The in-depth report is mainly divided into three parts. First is the estimation of market demand, predicting that the PCB industry demand will reach $5.6 billion by 2025, about $2.8 billion in 2024, and is expected to reach $10 billion by 2026, showing a trend of doubling for three consecutive years. Second is the sorting of the supply side, discussing potential future changes, including technological evolution and decision factors for manufacturers when choosing PCBs. Finally, the report provides a detailed analysis of the PDD industry driven by AI computing power, pointing out that it is overall in a logic of simultaneous quantity and price increase, and conducts in-depth research on fields such as GPUs, ASICs, switches, and optical modules that have a driving effect on PDB.
From the perspectives of quantity and price, what factors have driven the growth of PDB demand?
From the perspective of quantity, the main drivers are the increase in GPU and ASIC shipment volumes, influenced by AI demand, rapid growth in computing power, and the application of large models, with cloud manufacturers promoting ASIC deployment. It is expected that GPUs and ASICs will become important driving forces for PDB demand growth. At the same time, the demand for switches and optical modules will also see significant growth. From the perspective of price, on one hand, AI has increasingly high requirements for PCBs, including performance, power consumption control, and the use of high-frequency high-speed materials (such as Ma9, Xingbu, PDFE, etc.), driving up the price of single boards; on the other hand, the application of new processes and technologies, such as the promotion of Kewo mainboards, gradually replacing copper cables or expansion boards, further drives the increase in PCB value.
What are the main growth engines for the AITDB market news in 2026 and beyond?
The main growth engine is the continuous expansion of market size and the replacement of other interconnection methods. It is expected that the market size will continue to grow, and the growth rate will not be limited to 2026, with significant growth also possible in 2027.
How is the GPU market size forecast conducted?
For the GPU market size forecast, we based it on the shipment data of the H series and B series this year (approximately 1.6 million units for the H series, and a total of about 4 million units for the B series and GD300), predicting that next year, with the further delivery of GD300, the total GPU shipment volume will reach 6.04 million units, and with partial shipments of the R series, the total shipment volume is expected to be around 6.76 million units, which is an increase of about 1 million units compared to 2025, representing a growth of about 20% to 30%.
How is the price forecast conducted? What is the specific composition of the PCB value for H series and B series servers?
We conducted detailed calculations for each generation of products. For example, the single GPU value for H series, B series, and GD200 to GD300 servers under the open architecture is $183, $240, $301, and $413 respectively, with an expected growth of 30% for each generation. This is mainly due to factors such as increased area, heat dissipation power consumption, and performance improvement, as well as changes in price structure. The single GPU corresponding PCB value for H series servers is about $183, mainly composed of OM and UVB architecture, involving Ma Liu materials, 5G HDI, and 26 to 28-layer high multi-layer board components. The B series servers have some architectural changes, but overall still adopt similar PCB requirements, with adjustments in area and materials, with a single GPU corresponding PCB value of about $240.
What significant changes does GD200 bring to PCB demand as a new generation server?
GD200 has significant architectural changes from the B series, no longer using the OM plus UBB method, but instead adopting a more integrated configuration. The GPU mainboard no longer requires an OM small acceleration card but is directly soldered onto the mainboard, increasing the demand for Ma Ba materials’ HDI, with the PCB value for a single GD200 server significantly increasing to about 2224 RMB ($309), mainly reflected in the upgrades of the GPU mainboard and other high-density interconnection components.
What changes occurred in the 300 mainboard at the GDC conference? Why did it shift from integration to modularization and then back to integration?
A significant change in the 300 is the shift from an integrated mainboard to modular supply, introducing the XN park module to provide similar GPU functionality, using OM and UVB architecture. Although an attempt was initially made for modular assembly, it ultimately reverted to integration, mainly because this solution is superior in performance and cost-effectiveness, directly placing the GPU on the OM and forming large-scale integration through modular assembly. Additionally, regular CPU and BGA packaging technology were adopted, resulting in changes in overall PCD value, such as UBB carrying two CPUs and four S-type m pop connectors, shifting from direct soldering on community boards to UBD plus OM design, leading to an increase in single cabinet value from 200 to 214,000 RMB, and a 34% increase in single machine PU’s PP value to 2973 RMB, approximately $413.
What are the main reasons for the increase in quantity and price?
The increase in quantity and price mainly stems from the simple division of a large board into HDI within OM plus UVB, increasing the overall PCB demand, achieving a transition from integration to modularization, resulting in an approximate 34% increase in quantity and price.
What are the forecasts for GPU single machine PU in the coming years and the reasons behind them?
It is expected that GPU single machine PU will grow by 30% after the initial production delivery of GD300 starts in August, with the forecasted single machine PU price around $550, representing a growth of 30% to 40% compared to existing solutions. This growth is mainly reflected in the upgrades of PCB architecture and materials, including the potential use of Ma Jiu materials, which have not yet entered mass production and are expected to be significantly more expensive than existing materials.
What is the forecast for NVIDIA’s PCB demand from 2024 to 2027?
It is expected that NVIDIA’s PCB demand will reach $850 million in 2024, double to nearly $1.53 billion in 2025, and again double to nearly $2.64 billion in 2026, with very rapid growth, and significant growth is still expected in 2027.
What impact does the acceleration of AI commercialization have on the demand for S (possibly data centers or servers)?
The acceleration of AI commercialization has driven continuous increases in capital expenditure for data centers, such as Google raising its 2025 KPI from $75 billion to $85 billion, with total KPIs for the four major North American tech companies expected to reach $366 billion in 2025, a nearly 50% year-on-year increase. The demand for S (servers) has significantly increased due to AI transforming into hardware and computing power infrastructure, especially the demand for ASICs and PCBs will continue to grow.
What is the forecast for Broadcom and Marvell in the ASIC market?
Marvell predicts that the ASIC market size will reach $55.7 billion in 2023. Taking Google as an example, its seventh-generation TPU VCP (I5) adopts a new chip stacking and connection method, significantly increasing PPT usage. This year, Google expects to ship about 2 million units, with an expected shipment of 3.3 million units next year, showing rapid growth, with V6 being replaced by V7, and PPT usage continuing to increase.
How is Google’s annual procurement demand growing?
Google’s annual procurement demand is expected to grow from $1 billion this year to nearly $2 billion next year, and considering the price increase for each generation from V6 to V7, the growth rate may exceed expectations.
What are the differences in strategic strength between ASIC and GPU? What are Amazon’s expectations for PCB and chip shipment volumes?
Compared to GPUs, ASICs have slightly weaker strategic strength, addressing power consumption and energy demand through PPT, and may adopt higher-level PCBs and design structures similar to those used in Amazon’s 2.5 or 3rd generation chips. It is expected that Amazon will ship about 1.5 million chips this year, reaching 2.2 million next year, with significant demand for overall PCB area and volume, with related benefiting suppliers including Business Electronics.
What are the consequences of not adopting HBM design for GPUs?
Not adopting HBM will require more LPDDR5 or GDDR memory next to the GPU, increasing chip size. For example, the application of Meta’s MTIA architecture in actual products will lead to an increase in PCB area.
What are the expected GPU shipment volumes for Microsoft and Meta this year and next year? What is the overall growth situation and unit price for S and AMD shipments?
This year, it is expected that Microsoft and Meta will ship nearly 200,000 units combined, reaching 1 million units next year. For AMD, this year focuses on MI300 and MI350, transitioning to M400 next year, with an expected shipment of 700,000 units this year and 1 million units next year. The overall S (possibly referring to GPUs) shipment volume is expected to be 4.5 million units this year, growing to 8.1 million units next year, with a unit price calculated at $550, leading to a PCB demand of $2.5 billion this year and reaching $4.5 billion next year.
What is the demand situation for optical modules, switches, and communication equipment for PCBs?
This year, the demand for PCBs from optical modules and switches is $1.6 billion, expected to grow to nearly $3 billion next year. The unit price of optical modules is relatively low, but the demand for PPT is high, and as the rate upgrades, the value is also increasing, making the overall market size considerable.
How is the expansion situation of PCB companies on the supply side?
Starting from the third quarter, the KPI of PCB companies has shown significant month-on-month growth, and it is expected to reach a total output value of 25 to 30 billion RMB by 2025. The KPI for Q2 2024 is about 3.8 billion RMB, which has grown to 4.5 billion in the third quarter, and even reached 5.286 billion in the fourth quarter, with this month-on-month growth trend continuing to accelerate.
What is the market expectation for the capacity release of newly produced high-end products?
The market holds an optimistic view on the capacity release of newly produced high-end products, expecting that based on a 1:2 output value calculation, there may be nearly 40 billion RMB of capacity release. However, whether such capacity can be achieved is uncertain, as factories in Southeast Asia may face challenges in the yield ramp-up cycle for high-end products, localization support, and infrastructure (such as power and logistics efficiency), which may lead to lower-than-expected capacity release from overseas factories.
How long will the current supply-demand imbalance last?
We believe that the market will continue to experience supply-demand imbalance at least until this wave of production comes online. According to estimates, the output value demand is around 20 billion RMB, and if produced at a 1:1 ratio, it would be 20 billion RMB in output value, and if produced at a 1:2 ratio, it would reach 40 billion RMB in output value, which far cannot meet the demand of $10 billion. Additionally, considering the potential risks in the Southeast Asian supply chain and the lagging efficiency of capacity release compared to demand growth, the supply-demand gap is expected to persist in the medium term for 1 to 2 years.
What is the demand and supply situation for high multi-layer and HDI materials?
Currently, due to the expansion plans of PCB manufacturers, the demand for high multi-layer and HDI materials continues to grow. Especially for high-end HDI materials, their production capacity is relatively tight, and major suppliers in Taiwan may allocate capacity to other fields such as satellites and military applications, which may not all be used in the server field. Additionally, the processing difficulty of HDI materials is high, with strict yield requirements and high costs. Despite these issues, due to their advantages, they will still be widely used in high-end products.
What are the advantages and challenges of HDI technology in 5G applications?
HDI technology has high processing precision and yield requirements under the conditions of 5G HDI deep width, and it involves many steps, making the process difficult and costly. Its advantages include supporting 3mm thick copper and heat dissipation holes, with better heat dissipation performance than traditional technologies; at the same time, the transmission rate is relatively high, but it is prone to signal placement and insertion loss at high frequencies. Therefore, in high-end technology route decisions, it is necessary to balance the three-dimensional indicators of heat dissipation, signal transmission, and cost.
What is the trend of introducing new materials such as Ma Jiu and TDFE in the PCB field?
In the future, the application of difficult materials such as Ma Jiu or TDFE may first be introduced in the form of high multi-layer, as the combination of Ma Jiu and HDI may lead to lower yields. Additionally, the next-generation solution from NVIDIA will shift the rigid backplane from an oblong structure to an open structure, requiring high-density integration and signal transmission performance, and it is expected to first use Ma Jiu and high-layer solutions, becoming a new increment in the PCB market.
What are the development trends in the TCB and substrate markets?
TCB (Copper Substrate) is gradually replacing the modified version as a new way of transitioning to CMOS, with a high market share for SLP-type substrates. Taiwanese manufacturers dominate the SLP market competition landscape, while mainland manufacturers such as Deep Blue Circuit also have production capacity, but their share is relatively low. With technological advancements, TCB requires narrower line widths, faster transmission speeds, and lower losses, but also faces challenges such as CTE thermal expansion coefficients and dielectric loss, increasing the requirements for copper-clad laminate materials, such as Ma Jiu materials, which will be more widely used, driving up PCD value.
Why is PTFE in PCB materials receiving attention, and what are its advantages?
PTFE is receiving attention due to its excellent electrical performance and thermal stability, capable of long-term operation in ultra-low and high-temperature environments without electrical breakdown, especially effective in reducing pressure risks under high-layer pressure, making it a potential mainstream PCB material, as seen in some of NVIDIA’s solutions.
What predictions can be made about the new generation of materials that may be adopted in next-generation electronic products?
The new generation of materials may be a combination of Ma9 and PTFE, expected to be applied in the next-generation Ruby or ASIC solutions, with the main suppliers being two companies from Japan and Beili Hua and Zhongtai Technology from mainland China under Taishan Fiberglass.
—
In the era of AI, I hope to be of help to the big players.For detailed summaries, foreign research reports, and domestic and foreign audio, please join the Knowledge Planet to view.For specific advantages, please click“Introduction to Summary Planet”
