BofA Raises ABF Substrate Shortage Forecast, Supply-Demand Gap Widening to 19% by 2028
Nashnova编辑部
Bank of America raised its 2028 ABF substrate supply-gap estimate to 19%, driven by surging server-CPU demand that is tipping the industry from a buyer's market into a seller's market where pricing power flows through to gross margins.
Why does the gap keep getting revised upward?
BofA's August 19 report lifts the 2026–2028 ABF substrate — a high-end board that carries chips and routes signals — supply gap to 7%, 14%, and 19%, each 3–4 percentage points above the prior model.
The key driver is server CPUs. BofA now projects their share of total ABF demand at 17%→21%→21%, up from 13%→14%→14% previously — a 7-point jump by 2027.
This means → ABF demand no longer rides on GPUs alone. Server CPUs, custom ASICs, and network-switch chips form parallel growth lines, making the gap more resilient to any single-platform delay.
Chips are getting bigger — why does that make substrates scarcer?
AI chips are growing in die area, I/O count, and power draw. Substrates must match with larger area, more layers, and finer trace density.
In plain terms = as chips get more complex, the same equipment produces fewer qualifying substrate units — demand rises while per-unit capacity falls, so the gap steepens faster than chip shipment growth alone.
This reflects a mechanism easy to overlook: the shortage is driven not just by "more volume" but by "each piece is harder to make."
How concentrated is capacity — and can expansion close the gap?
The top seven makers hold roughly 97% of 2026 nameplate ABF capacity: Ibiden 21%, Unimicron 20%, Kyocera 16%, Shinko and Nan Ya PCB each ~11%, Kinsus and Samsung Electro-Mechanics each ~9%.
Unimicron is expanding most aggressively — BofA expects it to add over 30% capacity per year in 2027–2028. But high-end AI substrates require equipment delivery, process tuning, customer qualification, and yield ramp; generic capacity cannot substitute directly.
This means → expansion plans and persistent shortages can coexist — capacity under construction is not capacity available. Alternative technologies such as glass substrates are, in BofA's view, a variable closer to ~2030, not a confirmed supply source for 2026–2028.
How much have margins improved — and who is fastest?
In Q2 2026, all three Taiwanese substrate makers saw clear margin gains: Nan Ya PCB's gross margin rose 890 basis points quarter-on-quarter, Unimicron 680 bps, and Kinsus 490 bps.
Nan Ya's outsized improvement ties directly to its 50%+ share of high-end network-switch substrates and a customer base with limited competition.
BofA lifted Unimicron's 2028 revenue forecast by 12.6%, gross-margin estimate by 1.7 percentage points, operating-profit forecast by 20.9%, and EPS forecast by 20.8% — the most balanced upgrade across metrics, reflecting its scale and technology mix.
Whose 2028 EPS revision carries the most upside?
BofA raised 2028 EPS estimates for Unimicron, Nan Ya PCB, and Kinsus by 21%, 18%, and 25% respectively. Kinsus gets the largest revision.
Kinsus currently derives only ~5% of revenue from AI CPUs; BofA expects that to reach ~15% by 2027, with existing equipment upgrades contributing roughly 7–8 percentage points of gross-margin improvement from 2028.
This means → Kinsus has the highest leverage precisely because its base is lowest — a jump from 5% to 15% is far easier to deliver than incremental gains from an already elevated share.
Why do BofA's two 2030 models rank the stocks differently?
Top-down model (fixed unit pricing; demand growth and share as variables): Kinsus's base-case net profit rises 96% vs. old 2028 estimates, 155% in the bull case — the highest leverage, driven by a low share base and large AI ramp potential.
Bottom-up model (shipment CAGR fixed at 15%; ASP CAGR of 5%–25% as variable): Unimicron and Nan Ya PCB base-case net profits rise 78% and 71%; Kinsus only 39% — pricing leverage favors the two with greater scale and higher-end product mix.
In plain terms = if the shortage is monetized mainly through share redistribution, Kinsus has the most upside; if mainly through price increases, Unimicron and Nan Ya benefit more directly. High-end lead times, long-term contract renegotiations, and margin trends over the next two quarters will determine whether the seller's market moves from forecast to operating reality.
Content is for reference only, not financial advice.