Nvidia Directly Engages Co-Tech to Secure Long-Term HVLP4 Copper Foil Capacity
Miles Bennett
Nvidia has invited Taiwan copper-foil maker Co-Tech to the U.S. to negotiate long-term HVLP4 supply; only four producers worldwide can deliver the material reliably, and a 15%–25% supply gap is forecast for 2026 — the AI server buildout is now being bottlenecked by upstream materials.
Why is Nvidia negotiating directly with a copper-foil maker?
The bottleneck for AI compute expansion is shifting from chips to more basic materials — HVLP4 copper foil (an ultra-low-profile foil used as the base substrate for high-speed signal PCBs) is a critical input for next-generation AI servers.
Co-Tech Chairman Frank Lee disclosed on June 8 that Nvidia invited the Co-Tech team to the U.S. to discuss long-term capacity planning.
This means → Nvidia is bypassing PCB middlemen to lock upstream material directly, the same playbook end-customers used during the earlier T-glass fiberglass-cloth shortage.
How large is the supply gap?
Only four producers worldwide can supply HVLP4 copper foil reliably. Co-Tech is the only one based in Taiwan.
Lee estimates a 15%–25% supply shortfall in the HVLP4 market in 2026, widening further in 2027.
In plain terms = high barriers and long lead times mean supply cannot keep pace with surging AI demand — in the near term, money alone cannot buy enough volume.
What does this deal mean for Co-Tech?
If talks succeed, Co-Tech would become the second copper-foil supplier — after Mitsui Kinzoku — to operate a direct-consign model with Nvidia (supplier stages inventory at the customer's warehouse; settlement follows actual usage).
Co-Tech's HVLP4 foil is already used in the switch trays and compute trays of Nvidia's Vera Rubin servers.
This means → Co-Tech upgrades from "a supplier to Taiwan CCL makers" to "a direct Nvidia supplier," gaining both pricing power and order visibility in one step.
Can the capacity ramp keep up?
Current combined HVLP3 + HVLP4 monthly capacity stands at 100 tonnes, roughly 15%–20% of revenue. The plan is to reach 150 tonnes (+50%) by end of 2026.
A third factory is set to start production in Q1 2027. HVLP4 capacity ramps in phases through Q4, lifting total high-end foil capacity to 600 tonnes and targeting a revenue share above 50%.
This reflects a full strategic pivot from commodity foil to high-end AI materials — Co-Tech is betting on a multi-year supply-deficit window.
What do pricing and financials show?
Foil processing fees have risen 5%–10% year-to-date, with room for further hikes in H2 depending on supply-demand. Order visibility now extends to 2028–2029.
2026 revenue is expected to rise quarter by quarter to a full-year record. The H1/H2 split is roughly 40:60, with a gross-margin target of about 30%.
For reference, 2025 consolidated revenue was NT$7.88 billion (~US$250.3 million), gross margin 22%, net income NT$1.062 billion, EPS NT$4.21. In plain terms = gross margin jumping from 22% toward 30% is driven entirely by the rising share of high-end foil.
When does the ASIC platform catch up?
ASIC platforms — custom-designed AI chip solutions — still mostly use HVLP3 foil today. The upgrade to HVLP4 is expected to accelerate in 2027.
This means → current HVLP4 demand comes mainly from Nvidia GPU servers. Once ASIC platforms switch over, demand steps up again, and the supply gap could widen beyond today's estimates.
Content is for reference only, not financial advice.