AI PCB and Substrate Thesis: GPU, CPU, NIC and CCL Demand Are One System Bottleneck

The market often frames AI hardware as a sequence: GPU first, then memory, then substrates. That is only partly right. AI infrastructure is now a rack-scale system made of GPUs, CPUs, DPUs, NICs, switch ASICs, memory modules, power boards and low-loss CCL. Every chip expansion needs a board. This sector thesis connects Samsung Electro-Mechanics, Daeduck Electronics, Doosan Electronic BG, Kolon Industries and Pamicell to the same system-level bottleneck.

Toss / Viva Republica IPO — Why the Honest Valuation Range Is $5B–$15B, Not Just $10B: A Neutral Anatomy of the Coordinate-Definition Event

Part 5 was the conservative endpoint of new-coordinate definition (company defining, market discounting). Part 6 was the aggressive endpoint (market pricing, company catching up). Part 7's Viva Republica (Toss) goes one step further — a pre-listing case where the coordinate itself hasn't been accounting-closed yet. This piece refuses to declare a single 'base case'. Considering the five institutional walls of US listing — public F-1 not yet filed, ADR/FPI discount, Korean financial-regulation exposure, secondary-share liquidity complexity, and the standard 30–40% IPO discount — the possible price spectrum spans $5B to $15B. Reuters' '$10B+' headline sits at one point in the upper-end region, reachable only when all five walls are cleared and the super-app narrative is convinced. Inverting the four scenarios into the series' implied-cost-of-equity matrix reveals that the same company can land anywhere from one endpoint to the other.

KakaoBank (323410) — The Korean Bank the Market Already Prices for ROE 11%: A 4.2% Implied Cost of Equity, the Mirror of Korea Investment Holdings

Part 5's Korea Investment Holdings was 'a company defining a new coordinate while the market applies a conservative discount.' Part 6's KakaoBank is precisely its mirror image — a company actively claiming a new coordinate ('mobile financial platform + AI-native bank + ROE 15%'), with the market already pricing that claim in. Implied cost of equity ~4.2% — far below all five other companies in the series and the polar opposite end of Korea Investment Holdings's 17.3%. Two ends of the same 'new coordinate definition' chapter. The company pulling the market vs. the market pulling the company. The accounting meaning of the asymmetry.

Korea Investment Holdings (071050) — A Fifth Coordinate Candidate That Doesn't Map onto Any of the Four Existing Peaks: 'Capital-Operations Platform'

Series Parts 1–3 framed Meritz, Kiwoom, and KB as three peaks. Part 4 traced Shinhan's transit toward KB's coordinate. Part 5's Korea Investment Holdings (071050) maps onto none of them cleanly. Equity ₩12.1T → outstanding short-term notes (발행어음) ₩21.5T against a ₩24.2T cap (89% utilization) → IMA balance ₩1.9T → combined funding capacity ₩36.3T. That 5-step 'fund → asset-create → operate' flywheel exists nowhere else in the series. ROE 18.5% × PBR 1.07× = implied cost of equity 17.3% — the highest of all five. The 5.8 pp gap is not a 'mispricing' — it is the accounting evidence that 'a new coordinate is being defined'. The speed at which it gets recognized depends on two self-stabilizing mechanisms: governance and shareholder-return policy.