[대갤] SK하이닉스, 27년 TSMC 영업이익 추월

Nomura Securities: “SK Hynix 2027 Operating Profit 128 Trillion KRW → Overtaking TSMC”

AI Translated Image

Nomura Securities SK Hynix Outlook 2026 Operating Profit: 99 Trillion KRW (38% upgrade) 2027 Operating Profit: 128 Trillion KRW (46% upgrade) TSMC’s recent operating profit is around 80-90 Trillion KRW.

AI Translated Image

Absolute #1 in HBM3E — Virtually all of Nvidia’s HBM is SK Hynix For Nvidia’s Blackwell and Rubin GPUs, HBM is the most critical performance factor and the biggest cost component. And SK Hynix holds a near-monopoly on the most important HBM3E market share. 2025 HBM3E → SK Hynix virtually sweeps the market Nvidia’s massive orders → All concentrated on SK Hynix In short, as the AI server market grows, Hynix’s profits will grow ‘exponentially.’

Samsung Electronics focusing on HBM4 → Structure where HBM3E competitors vanish Samsung effectively skipped HBM3E and went for the 'Jump straight to HBM4' strategy. While this decision might be advantageous long-term, it resulted in SK Hynix sweeping the HBM3E market solo in the short term. 2025~2026 HBM3E Demand = SK Hynix Monopoly This monopoly is the foundation for the 99T profit in 2026 and 128T in 2027.

AI Translated Image

DRAM/NAND Price Surge + Shipment Increase = ‘Super Cycle of Super Cycles’ Nomura revised the 2026 price outlook for DRAM and NAND as follows: Product | Previous Forecast | Revised Forecast DRAM Price: +38% → +57% NAND Price: +36% → +65% Operating Profit Margin (OPM): DRAM 60~70% NAND 30~40% Basically returning to all-time highs.

AI Translated Image

Explosive growth in Nvidia GPUs until 2027 → Record-breaking order backlog Blackwell Blackwell Ultra Rubin Feynman With Nvidia’s GPU roadmap confirmed through 2027, a massive backlog has formed that won't shrink for 3 years. This is why Nomura went as far as saying they will “surpass TSMC.” Samsung Electronics 2027 Operating Profit 135 Trillion KRW — ‘Google TPU × HBM4’ Leverage Explosion — DS Division alone at 110T, virtually the best performance in company history. Morgan Stanley released the most aggressive profit outlook for Samsung.

AI Translated Image

Morgan Stanley Samsung Electronics Outlook 2027 Total Operating Profit: 135.22 Trillion KRW DS (Semiconductor) Division: 109.896 Trillion KRW Not 100T in revenue, but 100T in *operating profit*. This scale far exceeds the 2018 super cycle.

Why is Samsung Electronics also exploding?

AI Translated Image

1. Google TPU × Samsung HBM4 “Mega Alliance” TPU v8/v9 HBM4 volume → Surge between 2025-2027 Samsung HBM4 supply → 3x increase compared to 2025 2. Nvidia HBM4 Qual Pass Imminent HBM4 is a completely new market, and it's a phase where Samsung has a high chance of catching up or flipping SK Hynix. 3. 2nm Foundry Order Expansion Galaxy Exynos 2600 → Samsung 2nm Tesla AI6 → Samsung 2nm Google TPU → Expanding 2nm/3nm cooperation 4. Memory Price Hike + Supply Constraints Technically and competitively, Samsung is by no means a weak company. Samsung will see the HBM4 + Foundry + DRAM price cycle all rise simultaneously. This structure has more destructive power than the 2017-2018 semiconductor super cycle.

AI Translated Image

TSMC 2026-2027 Outlook — Growing but ‘Limited Speed’ — Structural reasons why it’s hard to keep up with the Memory Super Cycle TSMC is the biggest beneficiary of the AI era, but critical constraints exist. 1. CoWoS Packaging Bottleneck TSMC’s CoWoS production won’t increase significantly beyond 45K/month even in 2025-2027. It's hard for TSMC, which intentionally controls supply, to play the 'volume game.' This bottleneck leads to supply constraints for all major US Big Tech: Nvidia, Apple, Google, AMD. 2. Limits of Foundry Profitability TSMC’s GPM is around 60%, OPM around 50%. They take half the profit just by selling, which is high, but prices don't skyrocket like memory. ASPs don't jump like DRAM (which is basically spot price right now), and margins don't exceed 60% like HBM. So even if TSMC grows, it can't be as explosive as SK or Samsung. This is the basis for Nomura saying “Hynix will surpass TSMC.”

AI Translated Image

2025-2027 Semiconductor Industry Reshuffle — “Samsung × Google TPU” vs “SK Hynix × Nvidia GPU” After 2025, the AI chip industry will split into a duopoly, and the AI hegemony war will begin in earnest. With Amazon AWS Trainium, Apple, and Intel Foundry joining, the world is entering an AI arms race. Literally an AI War. Samsung Camp: Google TPU + Broadcom HBM4 2nm Foundry New Packaging (H-Cube, Fan-out) DS Operating Profit 110T Era SK Hynix Camp: Nvidia GPU + TSMC HBM3E Absolute Powerhouse 2026 HBM4 Mass Production Operating Profit 128T Era

AI Translated Image

“2027 will likely be the historic year where Memory Semiconductors overtake Foundry.” To summarize: SK Hynix: 2027 Profit 128T → Nomura report says they overtake TSMC Samsung: 2027 Profit 135T → Morgan Stanley report says DS alone hits 110T TSMC: Growing, but lacks the speed to keep up with the Memory Super Cycle In short: 2025-2027 is the 'Super Gap' AI chip market created by Nvidia, Google, Samsung, and SK Hynix. In this flow, the TSMC-centric Foundry era is ending, and the 'Return of the Memory Empire' centered on Korean companies is becoming a reality. Source: Taiwan Gallery [View Original]