AI Packaging Demand Widens ABF Substrate Supply-Demand Gap; Citi Sets ¥30,000 Target Price for Ibiden
Claire Weston
Citi set Ibiden's target price at ¥30,000, arguing that AI GPUs and hyperscaler ASICs are repricing high-end ABF substrates from cyclical material to advanced-packaging infrastructure — with Morgan Stanley projecting a 22% supply gap by 2030.
Why is a substrate suddenly worth this much?
AI GPUs, hyperscaler ASICs — custom AI chips designed by cloud giants — and server CPUs are pulling high-end ABF substrates (multilayer circuit boards that carry chips) in three directions at once.
This means → ABF substrates are no longer a cyclical material that rises and falls with chip cycles; they are being repriced as advanced-packaging infrastructure.
Citi targets ¥30,000, BofA ¥29,800. The two agree on direction; the debate is whether profits can keep getting revised upward through FY3/29–FY3/31.
How large is the supply gap?
Ibiden's capacity by FY3/29 is roughly 2.8× the baseline, but AI server and ASIC demand could approach 5×.
In plain terms = factory expansion cannot keep pace with downstream orders.
Morgan Stanley projects a 22% ABF shortfall by 2030. This means → pricing power hinges not on "how much ABF prices can rise" but on who controls the capacity.
Which three assets drive profits?
Ono plant serves GPU substrates; Gama plant serves AI ASICs and silicon bridges — interposer layers connecting multiple chips. Both anchor the AI upside.
Existing CPU substrate lines recovering utilization from 60–70% to 80–90% will be the first to steepen the profit curve, starting in FY3/27.
Mid-term targets: FY3/31 electronics revenue of ¥800 bn, group operating profit above ¥300 bn; long-term framework points to FY2030 net sales of ¥1 tn at a 30% operating margin.
What is each bank buying?
UBS buys the supply-gap thesis. BofA buys CPU utilization recovery and silicon-bridge ramp. Macquarie buys the EMIB-T process moat — an embedded bridge packaging technology that competitors struggle to replicate. Citi buys the valuation re-rating driven by ~30% annual AI-server growth.
This means → the market has no directional disagreement; the real debate is whether profits can keep rising from FY3/29 through FY3/31.
Customer prepayments convert expansion risk from a balance-sheet issue into a delivery issue — financially sound, but delivery cadence becomes the new core risk.
What signals should trigger caution?
Four cool-down signals: weakening AI ASIC demand, worsening T-Glass (a high-end glass core material) supply bottlenecks, declining customer prepayments, and CPU utilization recovery falling short.
In plain terms = ¥30,000 already discounts the good news; the stock is not sitting in undervalued territory.
If these indicators deteriorate first, the valuation premium will compress quickly. But a sharp drop with fundamentals unchanged is a different kind of opportunity altogether.
Content is for reference only, not financial advice.