A Steeper Curve: The Computing Power "Pricing Power" in PCB Q3 Reports
As of October 19, 13 companies in the SW PCB industry chain have disclosed preliminary Q3 earnings increases, and 4 have released formal Q3 reports. Shengyi Technology's... wait, no—WUS Printed Circuit (Huadian) expects net profit attributable to shareholders of 5.450 billion to 5.700 billion yuan for the first three quarters, up 100.54% to 109.74% year-on-year, with Q3 alone at approximately 2.53 billion to 2.78 billion yuan, up more than 50% quarter-on-quarter. The steeper growth stems from simultaneous rises in capacity utilization, product mix, and prices: Morgan Stanley estimates that the PCB value per rack for Nvidia's Rubin rack has risen from about $35,000 for GB300 to $117,000, an increase of 233%, with high-multilayer boards of 18 layers and above expected to grow 62.4%. Capacity utilization at leading high-end production lines is above 90%, with orders locked through 2027, but 17 listed PCB companies are expanding capacity through private placements; after concentrated capacity release in 2027-2028, scarcity will turn into competition.
Body · AI translation
Summary
1. Thirteen companies issued positive profit warnings and four released formal financial reports, as PCB makers write AI computing power from an industry story into their income statements. But more worth reading than "net profit doubled" is why the growth rate suddenly steepened in the third quarter, and how much longer it can stay steep;
2. The sharp change in the Q3 slope stems from simultaneous rises in capacity utilization, product mix and prices; incremental demand has shifted from area to unit price, and high-end boards offer ample elasticity.
A single-quarter "jump" that cumulative figures conceal
As of October 19, 13 listed companies in the SW PCB industry chain had disclosed third-quarter profit pre-increase announcements, and four had delivered formal third-quarter reports. High growth is nothing new in this sector; what is new is the shape of the growth curve. Through the first three quarters it was still a straight line sloping upward, and then in the third quarter the slope suddenly changed.
Wus Printed Circuit expects net profit attributable to shareholders of 5.450 billion to 5.700 billion yuan for the first three quarters, up 100.54% to 109.74% year on year. Backing out from the half-year report, the third quarter alone comes to about 2.53 billion to 2.78 billion yuan, versus 1.681 billion yuan in the second quarter, a quarter-on-quarter increase of more than 50%. Kinwong Electronic expects third-quarter profit of 912 million to 1.089 billion yuan, up 205.46% to 264.74% year on year. Jinlu Electronics has an even steeper slope: 124 million to 140 million yuan for the quarter, a year-on-year increase of 1,266.91% to 1,442.76%.
We pull out the single-quarter figures because cumulative numbers dilute the change. In the four formal third-quarter reports from Sihui Fuji, Kingbrother, Welgao and Junya Technology, revenue growth ranged from 17.65% to 65.41%, while net profit growth was 101.67%, a swing to profit, 161.43% and 259.40%, respectively.
On the demand side, the numbers must first be read for "quality." Prismark expects global PCB output value of $95.78 billion in 2026, up 12.5% year on year; shipments in the same year grow only 5.9%. The gap between the two numbers is the nature of this cycle: the increment is not in area but in unit price. High-multilayer boards of 18 layers and above are expected to grow 62.4%, while ordinary boards grow less than 3%. Morgan Stanley broke down the bill of materials for Nvidia's Rubin rack: PCB value per rack rises from about $35,000 for the previous-generation GB300 to $117,000, an increase of 233%; board layer counts are pushed from 8 to 12 layers to 18 to 30 layers, and evolving toward 40 layers and above. It is still the same board, but what is being sold is already something else.
Supply is tightening in tandem. Copper-clad laminate accounts for more than 30% of PCB cost, and its upstream raw materials have all run short this year: electrolytic copper has held firm at 109,600 yuan/ton; 7628-type electronic cloth has risen more than 70% year to date, with another 1.5 yuan/meter added in October, and G75 electronic yarn is up 3,000 yuan/ton; the full-year shortfall in HVLP high-end copper foil is estimated by the industry at 15% to 25%, and lead times for high-end fiberglass cloth have exceeded 30 weeks. Kingboard Laminates has issued eight rounds of price-increase letters this year, raising prices another 15% on October 13 for laminates and prepregs using thin cloth of 2116 and below; Panasonic raised prices from September 1, with some products up as much as 30%.
Upstream price increases have always been a double-edged sword for the midstream. In the first half, PCB makers absorbed costs without passing them through smoothly, and profits were squeezed in the middle; in the second half, as leading lines run at full capacity and orders stretch to year-end, bargaining power has returned to the board makers. Jinlu Electronics disclosed that the average unit price of orders in the first half rose nearly 30% year on year, and Sunda Technology implemented structural price increases on circuit board products. This is the real mechanism behind the third-quarter profit "jump": not that demand suddenly increased, but that the same capacity began shipping at new prices and in a new mix.
The 2027 "watershed," and the other half of the price increases
The weight of the mix can be measured in material grades. Taking the average price of M2-grade copper-clad laminate as 1x, M4 is about 1.4x, M6 about 2.4x, M8 jumps to 6.5x, and M9 about 12.1x. Copper-clad laminate makers therefore benefit earlier: Shengyi Technology expects net profit attributable to shareholders of 6.460 billion to 6.870 billion yuan for the first three quarters, up 164% to 181% year on year; Goworld expects 340 million to 360 million yuan, up 90.78% to 102%, and the company explicitly attributes this to two lines: rising volume and price of copper-clad laminate and optimization of the printed board mix. Further upstream, China Jushi expects net profit of 5.136 billion to 5.393 billion yuan for the first three quarters, up 100% to 110% year on year. Profit is distributed along "material grade," not evenly along the length of the industry chain.
That is precisely why the industry shows a K shape rather than a broad rally. The other side of the 13 profit pre-increases is growth of less than 3% for ordinary boards, and small and mid-sized manufacturers having almost no bargaining room between copper prices and price-increase letters. The boom is real, but it falls only on the capacity that can obtain high-grade materials and pass platform certification.
Automotive electronics is another curve that is less steep but steadier. The PCB area used per new energy vehicle is about six times that of a fuel vehicle; 800V high-voltage platforms, intelligent driving domain controllers and automotive Ethernet have pulled up thick-copper boards, high-frequency boards and automotive-grade HDI together, and global automotive electronics PCB output value is expected to exceed $12 billion in 2026, up more than 20% year on year. It cannot explain the third-quarter jump, but it does explain why some companies not deeply tied to computing-power customers have not seen their reports stall.
The third-quarter reports answer "can it be delivered"; the fourth-quarter reports and next year must answer "can it be retained."
Order visibility is indeed not low right now: capacity utilization at leading companies' high-end lines is above 90%, orders are generally locked through 2027, the AI PCB supply-demand gap is about 17%, and lead times have stretched to 12 to 20 weeks. But capacity is also assembling rapidly. Since the start of this year, 17 listed PCB companies have disclosed private placement plans or completed private placements, almost all directed at high-order HDI, high-multilayer and optical module substrates.
This is where the "watershed" lies. The construction cycle for high-end lines is 18 to 24 months, and customer certification takes another 12 to 18 months; these two thresholds keep 2026 tight. But by 2027 to 2028, when this batch of capacity is ignited at the same time, scarcity will turn into competition. By then, what determines profit will no longer be "whether there are AI orders," but three things: whether M8 and M9 grade materials can be obtained stably, whether new products can be designed into leading platforms' design lists, and whether gross margins can be defended when prices fall back.
Another easily overlooked variable is customer structure. Giants such as Nvidia have already locked up 2027-2028 capacity in advance; such long-term orders are a "moat" during scarcity, but in a surplus period they may become a price ceiling.
In a price-increase cycle, inventory also briefly plays the friend of profit: the difference between low-priced raw material inventory and the new selling price for the month falls directly into gross margin. But the same inventory becomes an adversary when prices turn. This explains why leading manufacturers sign long-term orders to lock in volume while pressing inventory turnover tighter.
Final thoughts
So in addition to growth rates, the third-quarter reports deserve three more lines of numbers: the quarter-on-quarter direction of gross margin, which determines whether the price-increase dividend stays in their own hands or merely passes through; the match between operating cash flow and net profit, which tests whether growth is "paper wealth"; and the pace of converting construction in progress into fixed assets, which foreshadows when depreciation will start eating into profit. The third-quarter reports are powerful evidence, but they prove only the past nine months.