NVIDIA Q1 FY27 delivered revenue of $81.6B, Data Center revenue of $75.2B, and Non-GAAP EPS of $1.87, beating consensus across the board. Q2 guidance of $91.0B ±2% also exceeded market expectations by a wide margin. The numbers alone offer no evidence of an AI CAPEX peakout. That said, the stock dipped roughly -1% in after-hours trading immediately after the report — not because of a miss, but because expectations were already elevated. The Korean takeaway is not simply 'buy HBM.' AI server bottlenecks are migrating from memory to packaging, power integrity, PCB/substrates, and data center power infrastructure. The primary beneficiaries remain SK Hynix and Samsung Electronics on the memory side, but fresh alpha is more likely to emerge from Samsung Electro-Mechanics silicon capacitors, HBM equipment, AI server substrates, and data center power infrastructure.
Samsung Electro-Mechanics disclosed a KRW 1.557T (~USD 1.035B) silicon-capacitor supply contract with a major global customer on May 20 — covering Jan 1, 2027 to Dec 31, 2028, an implied KRW 779B/year (~6.9% of FY25 revenue). The point is not 'one big order.' It is that SEMCO is being reclassified from an 'MLCC / camera-module / substrate company' into an 'AI-package power-integrity component company.' The market reacted: May 20 close +7.50% at KRW 1,061,000; May 21 pre-market +4.52% at KRW 1,109,000. But the stock had already run +56% from April 20 to May 20, so chasing is inefficient. Holders: Hold. New buyers: Wait — either a pullback to KRW 850–900K or a 2027E EPS consensus lift to KRW 28–30K. Next checkpoints: named customer, application position (Top / Land / Embedded), recognition pace, margin, and second / third design wins.
Bucketing iComponent as a 'display parts stock' gets the thesis wrong from the start. In 1Q26, revenue was ₩9.74B, with COATED PET (barrier coating film) at ₩8.06B — 82.8% of total. The company's identity has already changed. FY2025 revenue of ₩34.65B (YoY -6.6%) produced operating profit of ₩6.04B (+132.6%) and OPM of 17.4% — a clear mix-improvement signal where profits surged despite falling sales. 1Q26 repeated the same structure: revenue ₩9.74B, OP ₩1.66B, OPM 17.0%. Market cap ~₩42.1B, TTM PER ~6.0x. On raw numbers alone, this is obvious undervaluation. However, four unverified variables remain: no formal sell-side report within 90 days, undisclosed revenue breakdown by customer and product, inability to isolate FX effects, and the June 29 EGM treasury share disposal plan. The core question is whether 'COATED PET at 80% mix + OPM in the 17% range repeats in 2Q26.'
Fundamentally, Pamicell is already an electronic materials company. In 1Q26, low-dielectric electronic materials revenue reached ₩26.0 billion, 71% of total revenue, with OPM in the mid-30% range. KRX also changed its industry classification from basic pharmaceutical substance manufacturing to electronic component manufacturing as of 2026-05-04. But the two-week flow data from 5/4 to 5/18 says the market has not yet reclassified it. Daeduck Electronics rose +22.7% with foreign + institutional net buying of ₩64.8 billion, while Pamicell fell -17.9% with only ₩1.6 billion of combined buying. Its price correlation is closer to Doosan and ST Pharm than Daeduck. The business has changed into electronic materials, but the stock is still trapped in biotech inertia.
The two-week flow picture for Pearl Abyss does not yet qualify as a supply-demand turn. From May 4 to May 18, the stock fell 21.7% from ₩58,500 to ₩45,800, with foreigners and institutions net selling approximately ₩30.1B and ₩26.4B respectively. That said, after the May 13 forced-selling episode, foreigners and program trading posted partial recoveries, and short-selling share fell from its peak. The core battle is between foreign bargain re-entry and continued selling by institutions, private funds, and securities firms. This is a flow-bottom test, not a flow turn. A genuine reversal requires holding ₩45,500–₩46,000, reclaiming ₩47,200, and seeing durable buying from trust funds, pension funds, and insurers.