● Korea Claims Global Finance Bridge, CBDC vs Stablecoin Showdown

Can Korea Become the ‘Financial Bridge’ After SWIFT? A Korean-Style Solution Amid the Split Between Stablecoins and CBDCs and an Execution Roadmap for 2025-2026.

This article contains the actual dynamics and constellation between the U.S. stablecoin camp and the China/BRICS CBDC camp.

It examines the structural limitations of the SWIFT network and the decisive trigger for the expansion of on-chain payments in a news format.

It explains why the Bank of Korea’s hybrid CBDC design and the “Hangang Project” are central to a global financial bridge strategy.

An analysis is provided on Ripple’s (XRP) ODL, the practical competitiveness of the new stablecoin (RLUSD), and the three-way competitive scenario involving SWIFT’s on-chain transformation.

This article summarizes, as an execution guide, a Korean-style bridge architecture along with KPIs, a regulatory and governance checklist that are rarely discussed elsewhere.

Global Landscape Update: Stablecoins vs. CBDCs, and SWIFT’s Transformation

The payment and settlement infrastructure is being split between the U.S.-led stablecoin and tokenized deposit camp and the China/BRICS CBDC camp.

SWIFT is expanding its blockchain integration and ISO 20022-based on-chain messaging tests while maintaining its strengths as a sanctions and compliance network.

China’s CIPS (Cross-Border Interbank Payment System) is establishing itself as an alternative to SWIFT for yuan payments, aligning with multilateral CBDC projects like mBridge.

Stablecoins are dramatically reducing remittance fees and payment speeds, rapidly replacing the slow cross-border payments based on intermediary banks.

From an economic outlook perspective, 2025-2026 marks the critical point at which the global financial infrastructure transitions to an on-chain model.

News at a Glance: Key Issue Briefings

Issue 1. The constraints of SWIFT are not due to cost or speed, but rather the network effects of accumulated blacklist and sanction data.

Issue 2. Whether stablecoins can convert these network effects into “on-chain compliance” is the crux of the battle.

Issue 3. With Europe’s implementation of MiCA, the handling of dollar stablecoins within the EEA is restricted by platform, and regional regulatory differentiation is intensifying.

Issue 4. The U.S. major bank consortium’s concept of “tokenized deposit/payment tokens” could have an advantage over non-bank issuers in terms of distribution networks, credit supply, and bundled financial services.

Issue 5. The Bank of Korea’s hybrid CBDC and the “Hangang Project” are designed to interoperate CBDCs, stablecoins, and tokenized deposits within one network, positioning Korea as a financial bridge.

Variables in the Stablecoin Battle: Regulation, Distribution, and Liquidity

Regulation. In the U.S., legislative discussions over stablecoin regulation are ongoing, with the focus on payment stability, reserve transparency, and unified oversight.

Europe, through MiCA, institutionalizes reserve asset requirements and disclosure/approval processes to strictly control handling within the EEA.

Distribution. Tether and Circle have strong advantages in crypto-native distribution networks and global exchange networks.

Banks can reduce friction in adoption by bundling services such as corporate payments, trade finance, credit limits, and cash management.

Liquidity. The key to universal payments lies in the density of on/off ramps and price spread.

If a liquidity alliance is formed among major issuers, banks, and payment companies, exchange costs and settlement risks will be structurally reduced.

Compliance. As on-chain implementations of the travel rule, sanction screening, address reputation, and risk asset routing control become standard, “regulation-friendly stablecoins” become the default choice for institutional adoption.

Positioning of Ripple (XRP) and RLUSD: The Strengths and Realistic Competition of Bridge Currencies

Ripple’s ODL (On-Demand Liquidity) enables real-time currency exchange and payments without pre-funding, significantly reducing inventory capital in cross-border transactions.

Operating XRP as a bridge currency paired with dollar stablecoins (such as RLUSD) broadens the coverage of currency pairs and improves settlement speed.

However, if major stablecoins expand their market share, XRP’s addressable market (SAM) in the simple remittance space might be reduced.

Nevertheless, the technological edge in currency exchange, liquidity, and multi-currency netting ensures that it plays a strategic role in the “niche” between SWIFT’s on-chain transformation and stablecoins.

In conclusion, 2025-2026 is likely to see a phase where SWIFT on-chain, stablecoins, and XRP bridge coexist and compete in a three-way race.

Bank of Korea’s Hybrid CBDC and the “Hangang Project”: Blueprint for a Korean-Style Bridge

Architecture. It is designed to issue interbank settlement assets with wholesale CBDCs, allow commercial banks to issue deposit tokens, and enable interoperation between CBDCs, stablecoins, and tokenized assets on a public network.

Interface. It establishes a “national payment gateway” that integrates ISO 20022 messaging, cross-chain bridges, custody APIs, and on/off ramp standards.

International Interoperability. It assimilates SWIFT’s on-chain messaging, China’s CIPS, and BIS projects (such as mBridge) as a standard conversion layer for interoperability.

Risk. Security in cross-chain bridges, the decentralization limitations of permissioned networks, and conflicts with foreign exchange and capital controls must be addressed early in the design.

Korea’s 12·24·36-Month Roadmap to Becoming a “Financial Bridge”

12 months. In phase one of the Hangang Project, open a sandbox for CBDC-deposit token-stablecoin, and pilot three scenarios for export-import settlements and overseas remittances.

12 months. Permit on-chain currency exchange and mutual interoperability for the travel rule through special provisions in the Foreign Exchange Act and the Act on Reporting and Using Specified Financial Transaction Information, linking with public institutions and the Korea Trade Insurance Corporation for actual transactions.

24 months. Create a bilateral payment corridor with three ASEAN countries under the “K-Finance Bridge” and activate a real-demand testbed for bank-issued won stablecoins.

24 months. Transition the ISO 20022–chain messaging gateway and sanctions/KYC oracle to production to visualize the national backbone network.

36 months. Achieve continuous interoperability among CIPS, SWIFT on-chain, and mBridge, and automate trade finance, exchange rate risk insurance, and VAT settlements on-chain.

36 months. Pilot won-based on-chain payments for public procurement and ODA fund settlements to create policy-driven demand.

Policy and Governance Checklist: A Korean-Style Standard Package

Legislation. Definitions for deposit tokens and payment tokens, reserve regulations, licenses for custody/wallet operators, and on-chain KYC/travel rule mutual authentication regulations are required.

Foreign Exchange. On-chain currency exchange limits and reporting systems, real-time reporting APIs, resident determination based on addresses, and rules for monitoring overseas won must be established.

Security. Multi-signature/TSS for bridges, slashing/guarantee funds, and SOPs for responding to chain reorganization must be standardized.

Data. Public-private joint operation of address reputation, sanctions lists, and PEP data oracles is necessary.

Governance. A dual-led system for technology and regulation under a consortium including the Bank of Korea, the Financial Services Commission, the Financial Supervisory Service, banks, card companies, fintech firms, and exchanges is essential.

Performance Management Through KPIs

Payments. Aim for an average settlement time of T+0.x, an 80% cost reduction, and a failure rate below 0.1%.

Liquidity. Secure 30 currency pairs with an on-chain spread of less than 10bp.

Compliance. Achieve a 95% interoperability rate for the travel rule and less than 10% false alerts.

Transactions. Target an annual payment volume of USD 100 billion on the bridge network and 20% of trade finance processing to be on-chain.

Action Plans for Corporates, Financial Institutions, and Developers

Exporting Companies. Pilot policies for diversifying payment currencies and on-chain currency hedging, and promote onboarding of supply chain partners onto the same network.

Banks. Establish a reserve system for deposit tokens and won stablecoins, create institutional-grade custody, and link on-chain core banking.

Payment Companies. Package on/off ramps, KYC, and sanction screening modules as B2B SDKs.

Developers. Modularize and deploy into the ecosystem an ISO 20022 message-to-chain event conversion adapter, FX AMM, and intelligent router.

Essential Points Not Found in Other News

Korean Revenue Model. If the national gateway charges for payment, currency exchange, and compliance modules, revenue can be diversified through data and verification services even with payment fees under 10bp.

Realistic Path to Internationalize the Won. Instead of becoming an overseas won hub, establishing a “won stablecoin payment corridor” based on trade partners can increment market share without risk.

Technical Architecture Tip. By layering a lightweight L2 payment rollup on a permissioned L1 and enforcing sanctions/KYC oracles at a signal level, the sanction network effect of SWIFT can be replicated on-chain.

Strategic Partnering. Providing interoperability testing with CIPS and mBridge as a neutral gateway positions Korea as an asset of “bridge diplomacy” that is not biased towards either the U.S. or China.

Risk and Scenario Planning

Scenario A. Bank consortium-style stablecoins rapidly absorb corporate payments, while non-bank types focus on investment and trading areas.

Scenario B. Tether and Circle defend their status with liquidity and exchange networks while expanding their lineup of regulation-compliant options within the institutional framework.

Scenario C. Expansion of the mBridge and CIPS corridors leads to a fixed proportion of cross-border trade being handled via CBDCs.

Korean Positioning. By uniting interoperability under a “national payment gateway” during the coexistence of these three scenarios, Korea can maximize its benefits regardless of the outcome.

One-Line Conclusion

The future of global finance after SWIFT is not about “who is faster,” but “who connects better,” and Korea can secure a central role in that connection through hybrid CBDCs and stablecoin interoperability.

< Summary >

The U.S. stablecoin and China/BRICS CBDC camps are splitting, while SWIFT is exploring an on-chain transformation.

The stablecoin war will be decided by the on-chain implementation of regulatory compliance, distribution networks, liquidity, and compliance.

Ripple’s ODL and RLUSD demonstrate strengths in currency exchange and liquidity, setting up a three-way competitive dynamic with SWIFT on-chain and stablecoins.

The Bank of Korea’s hybrid CBDC and the “Hangang Project” lay the strategic foundation for connecting CBDCs, stablecoins, and deposit tokens on one network, turning Korea into a financial bridge.

If a 12·24·36-month roadmap and policy and governance measures are proactively established, Korea can make a leap to become a neutral payment hub in global finance.

[Related Articles…]

Reordering the On-Chain Payment Order After SWIFT

Hangang Project 2.0: Korean-Style Hybrid CBDC Strategy

*Source: [ 경제 읽어주는 남자(김광석TV) ]

– 한국, ‘금융 브릿지’가 된다? 스위프트 시대가 저물고, CBDC와 스테이블 코인이 만드는 새로운 금융 시나리오 | 경읽남과 토론합시다 | 홍익희 교수 2편

---

● Unstoppable AI Bubble

Is It Not an AI Bubble? Global Economic Outlook for 2025~2026 and a Physical AI Investment Map.

This article focuses on three key points.

First, it summarizes in a news style the structural reasons why the market does not easily crash even though it is a “bubble”.

Second, it reinterprets the transition of mega-model competition from GPT and Grok to “Vertical AI” and “Physical AI” from an investment perspective.

Third, it presents a concrete execution scenario in which Korea, with its semiconductor and manufacturing strengths, can become a partner to complement America’s weaknesses.

It is organized so that you can quickly make investment decisions by reading about the global economic outlook, interest rates, inflation, the stock market, and semiconductors all at once.

Breaking News Style Key Briefing.

• After three years, AI is expanding from LLM → LMM → Physical AI.

• According to discussions, GPT‑5 and xAI’s Grok‑4 are rated as top-tier in performance, but AGI has shown limitations in the current approach.

• The profitability of mega-models is unsatisfactory.

• Companies are prioritizing revenue by focusing on ‘Vertical AI’ in sectors such as finance and defense.

• Unlike the dot-com bubble, since the infrastructure and user base are already in place, the period of alternating corrections and upward trends is expected to be prolonged.

• As the data center ecosystem centered on NVIDIA expands, HBM, power, cooling, and substation facilities are expected to benefit simultaneously.

• Jensen Huang’s visit to Korea signals a physical AI manufacturing alliance.

• Korea is emerging as a card that can fill America’s manufacturing gap through its strengths in semiconductors, mass production, and data sovereignty.

AI Development Speed and Technological Track Realignment.

• Following the popularization of LLMs at the end of 2022, penetration has already begun across all industries.

• In discussions, it is summarized that LLM is basic, LMM is for content, agents handle automation, and Physical AI is responsible for powering robotics and vehicles.

• The discourse on AGI has subsided.

• The next steps are ‘monetizing specialized areas’ and ‘hardware robotics’ implementation.

Bubble or Not: The Speed of Capital and the Market’s Endurance.

• The conclusion is “it is a bubble” but “it will not collapse like the dot-com bubble.”

• Reason 1: The user infrastructure is already established worldwide.

• Reason 2: Liquidity expansion and capital recirculation occur rapidly.

• Reason 3: The innovation flywheel is structured so that talent and capital attract each other.

• However, the consolidation of 80% of companies is inevitable.

• An “escalator cycle” characterized by alternating rises and corrections is likely.

Survival Strategies for AI Companies: Focusing on Verticals and Monetizing via Defense and Finance.

• According to discussions, the theme is a transition to “Vertical AI”.

• High-priced SaaS in finance and project-based orders in defense/information are key cash generators.

• Companies like Palantir refine their technology through government contracts before expanding to commercial versions.

• Google and Meta strengthen their stock stories by enhancing productivity solutions through LMM and agents.

• Korean companies need to lower costs by utilizing lightweight, fine-tuned open-source models and combine them with on-site data to create a vertical deep stack.

GPU Economics: Why Has It Become a Battle Over Money.

• Both training and inference incur high costs for GPU usage fees and electricity.

• Discussions explain that training and testing a top-tier model once can cost billions of won.

• Capital strength determines the performance gap and the speed of product releases.

• Data centers are experiencing a simultaneous surge in demand not only for GPUs but also for power infrastructure, cooling, networking, and HBM memory.

• This structure forms a value chain that goes from “NVIDIA → HBM → power & cooling → construction”.

Data Center Ecosystem: Where the Money Is Coming In Now.

• HBM Memory: A structure that requires 10–20 times more volume than GPUs is spreading.
• Power Grid: Demand is rising for power generation, transmission, substations, ultra-high voltage transformers, and switchgears.
• Cooling: With a move to immersion and direct liquid cooling, related components and facilities are growing.
• Networking: Investments in optical modules, switches, and ultra-low latency fabric are increasing.
• Real Estate & Construction: Differentiated points will be large campus data centers and heat recovery solutions.

Inflation pressures increase initial CAPEX, but for market leaders, this creates barriers to entry.

Reasons Why Korea Is at a Disadvantage and Why It Excels.

• Disadvantage: It lacks the capital and GPU reserves to independently develop a hyper-scale LLM.

• Advantage: It excels in semiconductor design and manufacturing, mass production optimization, and rapid on-site adaptation.

• Data Sovereignty: It has strengths in search and communication platforms as well as in healthcare and educational data.

• According to discussions, Korean LLMs begin to show their presence from the lower 20s ranking, and the learning base exists.

• Policy Point: There is a need for a roadmap for approximately 100,000 GPUs for “Sovereign AI” and a public vertical service.

Why Jensen Huang Met with Samsung and Hyundai: A Physical AI Alliance.

• The message is about a manufacturing partnership for mass production in physical AI, such as robotics and autonomous vehicles.

• The intention is to compensate for America’s manufacturing weaknesses with Korea’s mass production and quality control.

• There is also demand for diversification of foundries, considering TSMC’s lead time and pricing issues.

• Hyundai is expected to benefit from synergies with Boston Dynamics.

• Samsung can engage in full-stack collaboration from AI chips, HBM, packaging, to foundry.

The Golden Ticket for Korean-Style Vertical AI.

• Healthcare: An AI doctor based on regular health check-up data is a world-class asset.
• Education: AI tutoring and individualized learning optimization bring significant productivity effects.
• Manufacturing & Quality: On-site data can quickly implement agent automation.
• Finance & Research: There is great potential for the commercialization of domain-specific models and toolchains.

The key is to build a structure that monetizes while maintaining data sovereignty.

Stock Market Strategy: Navigating Through the Bubble.

• The top tier remains the data center cycle centered on NVIDIA.

• The second tier consists of areas with highly visible cash flows such as HBM, packaging, power equipment, cooling, and optical modules.

• The third tier includes vertical AI solutions, SI, and MSP, where order backlogs drive performance.

• In physical AI, the value chain is extensive, covering automakers, robotics, sensors, and actuators.

• When interest rate volatility expands, high-value model stocks reduce leverage, and infrastructure and equipment stocks have relatively higher defensive capabilities.

• A resurgence of inflation may increase power costs and CAPEX burdens, but leading companies will have opportunities to pass on prices.

Risk Checklist.

• Delayed peak in interest rates and a renewed strength of the dollar.
• Bottlenecks in power infrastructure and rising rates.
• Increased US–China regulations and export controls.
• Collapse of profit models due to rapid performance leaps in open-source models.
• Revenue gaps arising from large customers switching to their own models.

Action Plan: Things to Do Right Now.

• Investors: Core focus on the four major areas of the data center ecosystem (HBM, power equipment, cooling, optical modules) and physical AI manufacturing beneficiaries, while approaching vertical AI on an event-driven basis.

• Companies: Rapidly iterate small-scale PoCs and integrate a billing system early by combining lightweight, fine-tuned open-source models with on-site data.

• Policy: A fast-track for power infrastructure, a roadmap for around 100,000 GPUs for Sovereign AI, public data rate schemes, and providing inference credits to startups are necessary.

Key Points Not Found in Other YouTube or News Outlets.

• Korea’s “regular health check-up” data is a globally scarce asset.

• It is one of the few countries with search and messenger platforms, capable of maintaining data sovereignty for learning in the long run.

• The bottleneck in physical AI is not the algorithm but rather “power, cooling, real estate, and permits.”

• In partnerships aiming to fill America’s manufacturing gap, Korea has a clear competitive edge in mass production compared to Europe and Japan.

• The expansion capacity for HBM and packaging is slower than that for GPUs.

Case Study: OpenAI, and What Follows.

• According to discussions, OpenAI faced variables in its transition to a profit-oriented structure and fundraising.

• Big tech accelerates technological absorption through headhunting.

• Ultimately, the structure is expected to converge to one of “a few top-tier players + many vertical players”.

2025~2026 Scenario Map.

• Base Scenario: Continued expansion of data center CAPEX, beneficiaries of power and HBM, and visible vertical revenue.
• Upside Scenario: Large-scale orders for physical AI pilot projects and entry into the commercial phase for robotics and autonomous driving.
• Downside Scenario: A rebound in interest rates, regulatory shocks, performance shocks from mega-tech companies, and delays in facility investments.

Checkpoint 5.

• NVIDIA’s new product cycles and guidance on HBM expansion.

• Power companies’ facility investment plans and rate structures.

• The approval speed for large cloud providers’ data center expansions.

• The scale of pilot projects and delivery calendars for automakers and robotics.

• ARPU and churn rates in the public and financial verticals.

Keyword Memo.

Global Economic Outlook.

Interest Rates.

Inflation.

Stock Market.

Semiconductors.

< Summary >

• It is indeed a bubble, but it will not collapse like the dot-com bubble.

• The competition among mega-models has shifted to Vertical and Physical AI.

• Money flows into the data center value chain as well as power and HBM.

• Korea is a card that can fill America’s manufacturing gap through semiconductors, manufacturing, and data sovereignty.

• Investment focus is on HBM, power, cooling, optical modules, and Korean-style Vertical AI.

[Related Articles…]

Signals of TSMC’s Supercycle and the HBM Investment Map

Physical AI Era: Reset Strategies for Korean Manufacturing

*Source: [ Jun’s economy lab ]

– AI 버블 절대 아닙니다(ft.최재붕 교수 1부)

---

● Jensen’s Bombshell Sparks Samsung Shock, HBM4 Packaging War – TSMC 2nm Crisis, 2025 AI Capex Frenzy

What Changes After Jensen Huang’s ‘Samsung Bombshell’ Statement: TSMC Inflection Point, HBM4/Packaging Battle, and 2025 AI Investment CapEx Landscape Overview

Jensen Huang hinted at the possibility of an “unexpected supply” arrangement with Samsung, covered the shift in the U.S. stance and TSMC’s Arizona loss issues, Samsung Foundry’s turnaround cards in GAA and packaging, the truth behind the HBM4 yield competition, and even the 2025 AI investment CapEx roadmap that connects directly to the stock market.It particularly focuses on the critical impact of the “packaging transformation” and “ABF/hybrid bonding” bottlenecks—aspects that other news seldom address—and the repercussions of exchange rate, interest rate, and inflation scenarios on the semiconductor cycle.It examines how the global economic policies and supply chain reconfigurations create opportunities and risks for Korean semiconductors, linking these to an actual investment checklist.

Breaking News One-Line Briefing

The U.S. is shifting its approach as manufacturing gaps become evident, leading to increased reliance on Korean memory and packaging.TSMC is focusing its early adoption of 2nm (N2) within Taiwan, while its U.S. production line faces high costs and low yields, resulting in estimated losses.Jensen Huang’s remarks are interpreted as a signal that Samsung might leave open the possibility of multi-sourcing for HBM, packaging, and parts of its foundry business.Samsung Foundry is gaining turnaround momentum with the stabilization of mass production lines for 3nm GAA and 4/5nm, as well as packaging (I-Cube/X-Cube).HBM4 is likely to be fully leveraged between 2025 and 2026, and the rumored 90% yield is still unofficial, requiring sober verification.

The U.S. Changing Its Stance: Acknowledging the Manufacturing Gap and Reinforcing the ‘Korean Card’

Amid the expansion of big tech AI investments, the limitations of domestic manufacturing capabilities are becoming apparent, prompting a pragmatic shift in policy.Despite subsidies under the CHIPS Act, accumulating cost increases from high wages, a shortage of skilled labor, and delays in equipment approvals have weakened competitiveness in the global economic environment.Consequently, there is a trend towards re-relying on Korea and Taiwan for memory (HBM), packaging (CoWoS/alternatives), and components (ABF/CPO).Although a prolonged period of high interest rates is making data center ROI more challenging, with inflation easing, there is a greater possibility of a stock market revaluation from the second half of 2025.As a policy risk, remarks about tariffs and procurement penalties may arise if foreign companies fail to be attracted, but ensuring a stable supply is the top priority.

TSMC’s Inflection Point: 2nm Favored in Taiwan, While the U.S. Faces the ‘Cost-Yield’ Challenge

TSMC is expected to concentrate its initial mass production of N2 (2nm) in Taiwan, deploying a strategy where the U.S. and Japan adopt lines with lower process difficulties.The Arizona line is burdened by a lack of skilled workers and an incomplete supply chain, resulting in high costs and recurring estimated losses in some quarters. Estimates of “trillions in negatives” should be viewed with caution.Although there are unexpected incentive proposals such as expanded overseas subsidies and linkage with defense procurement, the transfer of TSMC’s core technology is expected to proceed only in a limited manner.Issues related to Japan (such as allegations of equipment engineer technology leakage) remain unconfirmed, so it is necessary to distinguish between fact and rumor.Intel’s challenge with 1.8nm is not easy. It must simultaneously address process transitions, EUV/High-NA, and hybrid bonding, and when combined with packaging, the timeline risks are compounded.

Samsung Electronics’ Recovery Scenario: GAA, Packaging, and Multi-Sourcing

Samsung Foundry’s key turning point lies in stabilizing yields for 3nm GAA (mBCFET) and securing mass production volumes for 4/5nm.Claims such as universal application of Exynos and performance “six times that of Apple” may be exaggerated; however, improvements can be felt through AI NPU and on-device AI optimizations.Packaging capabilities (I-Cube/X-Cube/HBM-PIM) are rapidly emerging. By bypassing or supplementing CoWoS bottlenecks, Samsung is favorably positioned to receive multi-sourcing orders from its customers.Jensen Huang’s “unexpected supply” point is interpreted as a complex possibility involving HBM4 samples, advanced packaging, and some outsourcing of logic.Internalization and expanded cooperation in ABF substrates, as well as preparations for copper-to-copper hybrid bonding, are critical for long-term competitiveness.

HBM4 and the Yield Battle: It’s the Roadmap, Not Just the Numbers

HBM3E is expected to be a mainstay between 2024 and 2025, while HBM4’s mass production is anticipated to commence between late 2025 and 2026 as the base scenario.The notion of “90% HBM4 yield” remains at the level of industry rumor; customer validation (thermal, power, reliability, and performance) and overall packaging yield are the key factors.SK Hynix is maintaining its leadership in HBM3E, while Samsung is catching up with improvements in packaging and thermal management.Micron is raising its profile with its product portfolio, but it will need more time for customer validation and packaging integration.Ultimately, it’s not just the “die yield” of memory but the “overall yield including packaging,” delivery timelines, and customer-specific optimizations that will determine the winner.

The Often Overlooked Key Point: The Bottleneck Lies in ‘Packaging’, Not the Foundry

The real bottleneck in the supply of AI accelerators is advanced packaging capacity—specifically CoWoS/I-Cube-level facilities, ABF substrates, and hybrid bonding technology.Even if HBM die performance is excellent, shipment will be halted if packaging becomes an obstacle. The stock market momentum in 2025 will be driven even more by news of packaging capacity expansions.An expanded supply of ABF (large-area, multi-layer, low-loss) directly translates to improved AI server performance and energy efficiency.The supply chains of Korea, Taiwan, and Japan in ABF substrates are poised to be the decisive factors.When copper-to-copper hybrid bonding is adopted, improvements in thermal performance, delay, and energy efficiency will gradually enhance on-device AI and accelerate the transition to CPO (co-packaged optics).High-NA EUV is expected to be deployed in some nodes between 2025 and 2026, with full-scale implementation likely only after 2026. The focus is on long-term innovation rather than immediate yields.

U.S., Taiwan, Japan, and Korea: Winners in the Reconfigured Supply Chain

The U.S. is shifting its strategy to a “domestic + Asia multi” approach by leveraging subsidies and procurement, with actual volumes expected to come largely from Korea and Taiwan.Taiwan will maintain its position as the logic powerhouse, while also treating production in Japan (Raphidens) and the U.S. as “political options” to diversify risk.Japan holds a significant presence in equipment and materials (photoresist, etching, resist), but large-scale logic mass production will require more time.Korea is the only market that can prove itself through actual shipments via multi-sourcing in HBM, packaging, substrates, and foundry, making it a preferred category even amidst a global economic slowdown.

Macro Variables: Signals from Interest Rates, Exchange Rates, and Inflation in the Semiconductor Cycle

Interest rates are expected to gradually decline after remaining high. A decrease in the discount rate for data center CapEx will likely trigger a re-rating of the stock market from the second half of 2025.While a stronger won due to a weakening dollar may pose a short-term headwind for Korean semiconductor export profitability, a recovery in global demand is a larger positive.Easing inflation will lead to stabilized equipment and material costs, and improved margins. The semiconductor cycle is expected to enter a “second act of rising” by mid-2025.Keep a close eye on the global economy, interest rates, inflation, the stock market, and AI investments as key timing indicators.

Fact-Checking and Risk Management

Until official customer validation news emerges, the “90% HBM4 yield” should only be considered a reference value.Since the “TSMC U.S. massive loss” can vary significantly depending on quarterly/accounting treatments, caution is advised to avoid excessive exaggeration or understatement until divisional profit and loss details are disclosed.Claims like “universal use of Exynos” and “six times the performance” are largely marketing in nature, so actual benchmarks, power efficiency, and thermal performance must be verified.Policy risks (tariffs/procurement penalties) are factors of event-driven volatility. They should be mitigated through multi-sourcing and dispersing inventory and delivery timelines.

Investment Checklist: The True Beneficiaries of 2025 AI CapEx

Memory: Companies mass-producing HBM3E and those entering HBM4 early will benefit initially. Look for concurrent winners in thermal, power, and cooling solutions.Packaging: Companies involved in alternatives or supplements to CoWoS (I-Cube/X-Cube), TSV/micro-bump, copper-to-copper bonding, as well as test handler and burner firms, are poised for structural demand.Substrates: Companies expanding production of large-area, multi-layer ABF substrates could see long-term multi-fold re-rating.Equipment/Materials: Those related to EUV (mask, photo, cleaning), bonding/annealing/CMP, thermal interface materials (TIM), and gas/chemical materials are positioned for steady benefits.Exchange Rate/Interest Rate Scenarios: Growth stock multiples are expected to expand amid a weakening dollar and falling interest rates. Valuation pressures will be alleviated by clearer earnings visibility.

Key Observation: Samsung’s ‘Revival’ Begins with Packaging and Delivery

Samsung’s real turnaround will commence when its customer-customized packaging, delivery, and overall yield prove their worth, rather than merely capturing a specific node’s market share.While TSMC continues to reign as the king of cutting-edge logic, its pace of overseas expansion may be tempered by cost and political risks.Jensen Huang’s multi-sourcing signal could materialize in early 2025 through sample shipments and in the latter half of the year through mass production.In the stock market, keywords such as “packaging expansion,” “ABF investment,” and “HBM4 validation” will be the first to be priced in.

< Summary >The U.S. is expanding its reliance on Korea and Taiwan as it acknowledges manufacturing gaps, while TSMC is prioritizing 2nm production in Taiwan.Samsung is building turnaround momentum in GAA, packaging, and HBM, with the key battle for HBM4 hinging on “overall yield including packaging.”2025 AI investments in CapEx will represent a second act characterized by breakthroughs in packaging, ABF, and hybrid bonding.A relaxation in interest rates and inflation will help expand growth stock multiples; the key factors remain the global economy, the stock market, and AI investments.

[Related Articles…]Samsung and TSMC Quarterly Earnings Reveal a Shift in the Semiconductor CycleAI Server Investment: A Comprehensive Overview of the 2025 CapEx Roadmap

*Source: [ 달란트투자 ]

– 젠슨 황, 삼성전자 폭탄발언. 뜻밖의 납품에 겹호재 터졌다 | 박준영 대표 풀버전

---