AI PCB equipment and drill-bit supply chain Report Interpretation
Bank of America argues that the Vera Rubin and Rubin Ultra AI-server cycles will materially increase PCB content, layers and material complexity. It sees Dtech and Han's Laser as direct beneficiaries and rates both Buy.
Summary
Bank of America argues that the Vera Rubin and Rubin Ultra AI-server cycles will materially increase PCB content, layers and material complexity. It sees Dtech and Han's Laser as direct beneficiaries and rates both Buy.
- Rubin PCB content value could be about 3x that of GB200/GB300, with a further doubling possible for Rubin Ultra.
- The AI PCB drill-bit market is projected to grow at a 96% CAGR in 2026-28E to RMB14bn in 2028E.
- The PCB drilling-equipment market is projected to grow at a 12% CAGR in 2025-29E to US$2.7bn.
- M9 material could require more than 5x as many drill bits as M7.
- The report initiates Dtech with Buy and reiterates Buy on Han's Laser.
Report Interpretation
Overview
The report examines how faster AI-server upgrades are increasing PCB content and technical complexity, creating demand for PCB drilling equipment and consumable drill bits. It identifies Dtech as a direct drill-bit beneficiary and Han's Laser as a leading equipment beneficiary.
Core views
Bank of America argues that strong AI-cloud capital expenditure and faster GPU/ASIC upgrade cycles are accelerating an AI PCB upgrade cycle. Relative to GB200/GB300, the Rubin platform could raise PCB content value by about 3x, with Rubin Ultra potentially doubling that value again. The report links this increase to a broader range of boards, including a 44-layer HLC midplane in the Vera Rubin compute tray, additional ConnectX NIC and BlueField DPU boards, more HDI and HLC boards, and likely migration from M8/M8.5 to M9 copper-clad laminate. A Rubin Ultra backplane remains uncertain, but could add further content upside from 2027 onward. Vera Rubin racks require 54 more PCBs than GB300 racks before considering the effect of higher layer counts. The technical upgrade raises equipment intensity. AI servers generally require 20-40+ PCB layers versus 12-18 layers in general servers. Vera Rubin's Bianca board rises to 6+12+6 HDI layers from 5+12+5 in GB200/GB300, while switch-tray boards rise to 32 HLC layers from 22. Rubin Ultra backplanes could exceed 78 layers. More HLC raises demand for mechanical drilling of through-holes, while more HDI and finer blind vias raise demand for CCD mechanical drilling and laser drilling. M9 materials are more heat-sensitive and use harder Q-glass and advanced copper foils, making ultra-fast laser drilling more relevant. The report estimates drilling equipment, which accounts for 21% of the PCB equipment market, will grow at a 12% CAGR over 2025-29E to US$2.7bn, compared with 6% CAGR over 2020-24. It also notes that leading Chinese PCB makers' 2026 capex more than doubled year on year and should remain high in 2027-28E despite moderating from peak levels. Han's Laser is the report's preferred equipment beneficiary. It highlights the company's edge in CCD drilling equipment, potential upside from ultra-fast laser drilling as M9 adoption advances, sufficient capacity and product breadth, and scope to gain share from global peers through product quality and shorter lead times. The report also cites its Apple supply-chain position and possible 3D-printing-equipment opportunities. It reiterates Buy, with a new RMB185 price objective based on the average of RMB175 from DCF and RMB195 from a 50x 2027E P/E valuation. The report views drill bits as a more sustainable exposure than cyclical equipment because they benefit from continuing PCB shipment growth and increasingly demanding board designs. It forecasts the Nvidia-related AI PCB drill-bit market to grow at a 96% CAGR in 2026-28E and reach RMB14bn by 2028E. Its drivers are projected AI rack-volume growth, including likely doubling in 2026E and about 50% growth in 2027E; the new 44-layer compute-tray midplane; rising HDI/HLC adoption; and possible Rubin Ultra backplane and M9 adoption. The report explains that thicker multilayer boards and smaller holes require higher-aspect-ratio and coated drill bits. Drill bits for Rubin and Rubin Ultra may reach up to 50:1 aspect ratios, compared with below 20:1 for conventional bits. A >50:1 drill bit carries an ASP of about RMB20 per unit, more than 10x a conventional RMB1-2 bit. Dtech's blended drill-bit ASP rose 49% year on year in 1Q26 and 36% in 2Q26, while gross margin improved by 15.2 percentage points and 17.6 percentage points to 53.3% and 57.8%, respectively. Coated drill bits represented 35% of global sales revenue in 2025 and command materially higher prices than uncoated products. M9 is central to the drill-bit demand case. The report says its Q-glass has more than 99.9% silica content, increasing abrasive wear and shortening coated drill-bit life to 100-200 holes, versus 500-800 for M8 and about 1,000 for M7. This implies roughly 5x drill-bit demand compared with M7-era designs. HLC boards also require step drilling, which uses multiple bit lengths to create deep, high-aspect-ratio holes and increases bits consumed per hole. Bank of America initiates coverage of Dtech with Buy. It describes Dtech as the world's largest PCB drill-bit manufacturer, with 29.2% global market share in 2025 and supply relationships with more than 70 of the global top-100 PCB makers and nine of the top 10. The report cites direct leverage to AI PCB growth, resilient 2026-28E earnings visibility, and vertically integrated manufacturing that can support faster capacity ramps and share gains. Its HK$430 Dtech-H price objective is based equally on DCF and P/E. The DCF uses 3% terminal free-cash-flow growth and 13.2% WACC; the P/E approach applies 45x 2027E EPS, reflecting a 0.5x PEG against estimated 90% earnings CAGR over 2026-28E. Dtech-A's RMB506 objective is based on a 35% premium to Dtech-H, in line with its historical average.
Analysis framework
The report traces AI-server architecture changes into PCB volume, layer-count and material upgrades, then maps those changes to drilling-equipment and drill-bit demand. It supports the transmission with market-size forecasts, technical comparisons across server generations, supplier positioning, operating indicators and DCF/P/E valuation for the covered companies.
Methodology notes
AI server design upgrades are mapped through PCB specifications to equipment and drill-bit demand.
The report explains how additional boards, higher layer counts, HDI/HLC adoption and M9 materials increase manufacturing complexity and consumption of drilling equipment and bits.
Drill-bit market growth is driven by higher unit demand and higher average selling prices.
The report separates demand from more racks, boards and drilling steps from value uplift caused by high-aspect-ratio and coated drill bits.
DCF valuation for Dtech and Han's Laser.
The report discounts projected cash flows using stated terminal-growth and WACC assumptions to derive fair values.
Target P/E multiples applied to 2027E earnings.
The report uses earnings multiples, including Dtech's PEG-linked 45x 2027E P/E, alongside DCF to set price objectives.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- Dtech Technology (301377 CH / 1377 HK)Covered PCB drill-bit manufacturer positioned to benefit from AI PCB growth.
- Strengths
- World's largest PCB drill-bit manufacturer; 29.2% global market share in 2025; vertically integrated platform; exposure to high-aspect-ratio and coated bits.
- Comparison
- The report states Dtech held the largest global market share, ahead of Jinzhou at 21% in 2025.
- Risks
- Slower AI capex, slower AI PCB specification upgrades, drill-bit competition, and raw-material-price or high-end tungsten-steel availability fluctuations.
- Han's Laser (002008 CH)Covered PCB-equipment supplier positioned to benefit from higher HLC/HDI adoption and M9-related ultra-fast laser drilling.
- Strengths
- Strong CCD drilling position, ultra-fast laser-drilling progress, broad product range, capacity, and potential share gains from global peers.
- Comparison
- The report sees an opportunity to gain share from global peers with stronger CO2 laser-drilling positions, including Mitsubishi.
- Risks
- Slower consumer-electronics demand, slower handset-manufacturing upgrades, weaker PCB-segment growth or margins, new competition, and slower progress in ultra-fast laser-drilling equipment.
Key data
- Rubin PCB content valuec.3x versus GB200/GB300Bank of America estimate; Rubin Ultra PCB content could double again.
- Vera Rubin incremental PCB volume54 more PCBs per rackVersus GB300, excluding the impact of additional layers.
- PCB drilling-equipment market12% CAGR over 2025-29E; US$2.7bn by 2029ECompared with 6% CAGR over 2020-24.
- AI PCB drill-bit market96% CAGR over 2026-28E; RMB14bn by 2028ENvidia-related market estimate.
- M9 drill-bit consumptionc.5x versus M7Coated drill-bit life falls to 100-200 holes for M9 versus about 1,000 for M7.
- Dtech global drill-bit market share29.2% in 2025Ranked No. 1 globally by the report.
- Dtech 2Q26 drill-bit gross margin57.8%Up 17.6 percentage points year on year; blended drill-bit ASP rose 36% year on year.
Impact & implications
The report expects AI-server architecture, material and layer-count upgrades to broaden demand for PCB production tools while raising recurring drill-bit consumption and unit value. It identifies Han's Laser as an equipment beneficiary of CCD and ultra-fast laser drilling, and Dtech as a direct beneficiary of rising drill-bit volume and ASP.
Risks
- AI capital expenditure could grow more slowly than expected.
- AI PCB specification upgrades could progress more slowly than expected.
- Competition in PCB drill bits could intensify.
- Raw-material prices could fluctuate or high-end tungsten steel could become less available.
- Consumer-electronics demand could weaken.
- PCB-segment growth or margin performance could fall short of expectations.
- Competition from new entrants could intensify for Han's Laser.
- Progress in ultra-fast laser-drilling equipment could be slower than expected.