NVIDIA GTC TAIPEI 2026 · Keynote & Market Verdict & Growth Map

Keynote and

Market Verdict

The day Jensen Huang declared the "GDP Generator",
the US stock market responded with an all-time high

GTC Taipei Keynote (2026/6/1) · Reflecting US Market Close on 6/1

Written on June 2, 2026 · Keynote Presentation + Reflecting US Market Close on 6/1
Three Pillars · 6/1 Market Winners and Losers · Korea & US Growth Company Map · 3, 6, 12-Month Outlook

Team Sweet Potato: Onion (Macro) · Acorn (Industry) · Peanut (Fundamentals) · Carrot (Charts) · Sweet Potato (Comprehensive)

Prologue The Declaration of "GDP Generator," and the Market's Response

Chapter 1 The Big Picture — Agentic Computing and the Three Pillars

Chapter 2 June 1, Market Verdict — Winners and Losers

Chapter 3 Vera Rubin — The Heart of AI Infrastructure and Korean Memory

Chapter 4 Vera CPU — New Growth Engine of the Agentic Era

Chapter 5 Enterprise AI Toolkit — Every Company Builds Agents

Chapter 6 Reinventing the PC — "An Event as Big as Phones Becoming Smartphones"

Chapter 7 Physical AI — Robots, Autonomous Driving, and Hyundai Motor

Chapter 8 AI Factory — DSX and the Gigawatt Era of Power & Cooling

Chapter 9 Where Korea Was Called Upon — NAVER, Bank of Korea, and Hyundai Motor

Chapter 10 Growth Company Map — Korea and the US (Reflecting 6/1 Closing Prices)

Chapter 11 What Lies Ahead — 3, 6, and 12-Month Scenarios

Chapter 12 What to Watch Out For

Epilogue Between Production Announcement and Revenue Recognition

Prologue

The "GDP Generator" Declaration, and the Market's Response

On the day Jensen Huang redefined the identity of AI on stage, the US stock market responded with an all-time high.

At 11:00 AM on June 1, 2026, at the Taipei Music Center in Taiwan, NVIDIA CEO Jensen Huang left two short sentences during his two-hour keynote. "AI is now a profit generator. AI is now a GDP generator." It was a declaration that AI now generates profits and drives national GDP. He added a figure: while the global payroll paid to AI is about $3 trillion, the productivity it generates is about $9 trillion.

That night (June 1, Eastern Time), the market responded. The S&P 500 closed at 7,599.96 (+0.26%), the Nasdaq at 27,086.81 (+0.42%), and the Dow at 51,078.88 (+0.09%), with all three major indices finishing at record highs. NVIDIA's stock price, which briefly weakened during intraday trading right after the keynote, +6.26%($224.36) strongly rebounded, lifting the entire tech sector. The new PC chip N1X announced by Jensen Huang moved the market. He said, "This reinvention of the computer is as big an event as the phone being reinvented into the smartphone we know today."

"AI is now a GDP generator, not a cost.
A $3 trillion payroll generates $9 trillion in productivity."
— Jensen Huang, GTC Taipei 2026

This report covers two key areas. First, it follows everything announced in the keynote—the three pillars (Vera Rubin, Vera CPU, Enterprise AI Toolkit), the massive PC overhaul, physical AI, and AI factories—to map out which companies will grow. Second, it captures how the US market responded to the announcement on June 1—who the winners and losers were—to deliver the actual verdict. By synthesizing these two, we project what will happen over the next three months to a year.

The focus of the keynote was not on a single chip. Jensen Huang structured his presentation into three pillars: first, the full production of Vera Rubin; second, the Vera CPU for the agent era; and third, the Enterprise AI Toolkit that enables all companies to build agents. What he called the "most important slide" was not a specific product, but rather how computing will operate over the next decade—the agent computing pattern. The growth map in this report is drawn upon those three pillars and the market's verdict on June 1.

Chapter 1

The Big Picture — Agent Computing and the Three Pillars

Reading this keynote simply as "an announcement of a faster GPU" is only capturing half the story. What Jensen Huang painted was an entire new way of computing where agents operate, supported by three pillars.

The Most Important Slide — Agent Computing Pattern

Midway through the keynote, Jensen Huang showed a slide and said, "This is the most important slide." It was a diagram of how agents work. An agent operates with one or multiple large language models housed within a framework called a "harness." The harness orchestrates them to perform productive tasks. When an input is received, the agent understands, observes, reasons, acts, and uses tools. That tool could be another language model or a C compiler. Each time this process occurs, an entire AI supercomputer rack is activated.

"This is the most important slide. This is the computing pattern of the next decade. Agents, harnesses orchestrating... agentic AI has arrived, useful AI has arrived."

The reason this message is critical for investors is clear. It means AI demand is no longer a one-time event of "training a model once," but has shifted to agents constantly performing inference and calling tools 24 hours a day, driving continuous computing demand. Training happens once, but inference is forever. This transition structurally expands AI infrastructure demand.

The Three Pillars Outlined by Jensen Huang Himself

Jensen Huang explicitly grouped his presentation into three parts. First is the full production of Vera Rubin; second is the Vera CPU for the agent generation; and third is NVIDIA's Enterprise AI Toolkit, which enables every enterprise and software company to build agents.

Pillar	Role	Growth Connection
1. Vera Rubin	The Brain of the Agent — AI Supercomputer (GPU+HBM4)	Korean Memory (SK hynix, Samsung Electronics) & Back-end (Hanmi Semiconductor)
2. Vera CPU	The Hands of the Agent — Tool Calling & Code Execution	Reshaping the Data Center CPU Market, NVIDIA's New Growth Engine
3. Enterprise Toolkit	The Framework for Building Agents — Open Shell Runtime	Enterprise SI & Cloud (Samsung SDS, LG CNS, NAVER)

These three pillars are not separate announcements, but three parts of a single picture. The GPU (Vera Rubin) performs inference, the CPU (Vera) calls tools, and the software (Open Shell) securely runs those agents within the enterprise. On top of this, PCs (RTX Spark, N1X), robotics (Isaac GR00T), autonomous driving (Alpamayo 2), and AI factories (DSX) are layered. To find growth companies, one must look at this entire map. Betting solely on memory is a narrow reading of the keynote's message.

Chapter 2

June 1, the Market's Verdict — Winners and Losers

A keynote doesn't end with the presentation. Which stocks the market bought and sold that day is the true interpretation of the keynote. On June 1, the U.S. stock market responded strongly to the keynote, with all three major indices closing at record highs, while simultaneously drawing a clear line between winners and losers.

Three Major Indices Close at Record Highs

Index	6/1 Close	Change	Notes
S&P 500	7,599.96	+0.26%	Record High
Nasdaq	27,086.81	+0.42%	Record High
Dow	51,078.88	+0.09%	Record High

What is noteworthy is NVIDIA's movement. During Taiwan trading hours when the keynote was underway, NVIDIA briefly showed weakness. This was due to the interpretation of it being a "sell-on-news" event and observations that momentum had shifted to memory. However, once the regular U.S. market opened, the sentiment reversed. NVIDIA closed strong at +6.26% ($224.36), lifting the entire tech sector. The market shrugged off the bearish interpretation of the keynote and turned bullish.

Winners — Beneficiaries of the PC Overhaul Powered by N1X

The clearest winner in the U.S. market on June 1 was the PC overhaul theme. The announcement of N1X, which Jensen Huang called "an event as big as when the phone became the smartphone," boosted PC-related stocks.

Winning Stocks	6/1 Close	Change	Keynote Connection
CoreWeave (CRWV)	$124.82	+13.96%	Vera Rubin NVL72 Deployment Partner
Dell (DELL)	$465.96	+10.70%	Vera Rubin Deployment + AI PC
CrowdStrike (CRWD)	$782.17	+7.00%	AI Security & Agent Beneficiary
Micron (MU)	$1,035.50	+6.64%	HBM4 Supply (3 companies)
NVIDIA (NVDA)	$224.36	+6.26%	Protagonist of the Keynote, N1X & Vera Rubin
Adobe (ADBE)	$274.03	+5.72%	Photoshop RTX Spark Redesign
TSMC (TSM)	$435.63	+4.11%	Vera Rubin & N1X 3nm Production
Broadcom (AVGO)	$459.97	+2.95%	AI Networking
Vertiv (VRT)	$323.39	+2.43%	Data Center Cooling & Power

This table is the core of the June 1 market verdict. Dell (+10.7%) and HP (approx. +8%) benefited from the AI PC theme, while CoreWeave, which actually deploys Vera Rubin, surged the most at +14%. Micron (HBM4), Adobe (RTX Spark software partner), and TSMC (production) were all strong. The three pillars of the keynote and the PC overhaul translated directly into individual stock performance.

Losers — Legacy PC Chip Order & Profit-Taking in the Power Theme

Losing Stocks	6/1 Close	Change	Reason
Constellation Energy (CEG)	$265.70	-7.66%	Profit-Taking in Nuclear & Power Theme
Intel (INTC)	$109.33	-4.67%	N1X & RTX Spark Threatening the PC Chip Market
Tesla (TSLA)	$415.88	-4.57%	Alpamayo 2 Highlighting Autonomous Driving Competition
Vistra (VST)	$154.76	-3.41%	Profit-Taking in Power Theme
Alphabet (GOOG)	$372.58	-1.02%	Mega-Cap Tech Sector Rotation
AMD (AMD)	$510.13	-1.16%	NVIDIA's Dominance Highlighted

The list of losers carries an even more interesting message. Intel's drop of -4.67% reflects the market's assessment that N1X and RTX Spark pose a direct threat to the PC chip market, which Intel has dominated for decades. The benefits of the "reinvention of the PC" will go to finished goods manufacturers like Dell and HP, as well as NVIDIA and MediaTek, leaving Intel—the center of the legacy order—as a loser. Meanwhile, the decline in nuclear and power stocks (CEG -7.66%, VST -3.41%) is largely a short-term profit-taking reaction, as the keynote emphasized power-"saving" technologies. This does not mean the long-term trend of AI data center power demand has weakened, but rather that profits were realized in the power theme, which had risen significantly.

One-line summary of the June 1 verdict. The market dismissed "sell-on-news" concerns and embraced the keynote bullishly (NVDA +6.26%, all three major indices at record highs). The biggest winners were the PC overhaul (Dell, CoreWeave) and HBM4 (Micron), while the biggest losers were the legacy PC chip maker (Intel) and power stocks undergoing profit-taking (CEG, VST). This winner-loser dynamic serves as a compass for reading the growth map ahead.

Chapter 3

Vera Rubin — The Heart of AI Infrastructure and Korean Memory

The first pillar, Vera Rubin, is not a single chip but a system composed of 7 chips and 18,000 components. Korean memory lies at its heart, and on June 1, Micron confirmed its value in the U.S. market by rising +6.64%.

The Scale of Vera Rubin

Jensen Huang emphasized twice that Vera Rubin is now in full production. The Vera Rubin he refers to is the entire system, "not just the GPU." It combines seven new chips manufactured on TSMC's 3nm process, and a single compute board houses 6 trillion transistors and over 18,000 components. The NVL72, which bundles these, and the 3rd generation MGX rack, which bundles those further, are made up of 1.3 million components. Jensen Huang stated that the supply chain built for Vera Rubin is twice the size of the previous-generation Grace Blackwell.

"It starts at TSMC, the seven new chips that make up Vera Rubin ... 3 nanometer process, CoWoS packaging, HBM4 memory from Micron, SK hynix, and Samsung ... 6 trillion transistors with over 18,000 components on one board."

Korean Memory — Explosion of the Absolute Market

This single sentence is the most important for Korean investors. NVIDIA will receive HBM4 memory, a core component of the Vera Rubin, from three companies: Micron, SK Hynix, and Samsung. It has been officially confirmed that the two Korean companies will enter the heart of the next-generation AI supercomputer. Micron's +6.64% rise in the US market on June 1 shows the market's high expectations for the entire HBM4 supplier base. The same logic applies to SK Hynix and Samsung Electronics.

Here, the key to growth is not the battle for market share, but the absolute size of the market. As AI inference becomes perpetual through agentic computing, HBM demand itself is skyrocketing. SK Hynix and Samsung Electronics are the two main suppliers in this exploding market. Although market share proportions may be slightly divided with Micron's entry, in a phase where the market pie itself grows several times over, the absolute revenue and profits of all three companies will expand together. This is why SK Hynix and Samsung Electronics are the most direct Korean beneficiaries of the keynote.

Benefits Spreading to Back-end Processing

The explosion of HBM4 spreads beyond memory companies to back-end equipment. HBM is made by vertically stacking multiple DRAM dies, and Hanmi Semiconductor virtually dominates the Korean market in TC bonders (Thermal Compression Bonders) used for that stack-bonding. As HBM4 volume increases, bonder orders follow suit, and demand also spreads to laser dicing, cleaning, and inspection equipment. Although Hanmi Semiconductor experienced a temporary earnings gap in Q1, it is structured so that orders will resume as the full-scale mass production of HBM4 approaches.

Companies poised to grow from this branch. First priority is SK Hynix and Samsung Electronics (explosion of absolute HBM4 revenue), second is Hanmi Semiconductor (TC bonder back-end processing), followed by EO Technics (laser dicing) and Soulbrain (CMP/cleaning). In substrates, Samsung Electro-Mechanics and LG Innotek, which manufacture FC-BGA, are candidate beneficiaries of AI chip packaging. In the US, Micron (6/1 +6.64%) is a direct beneficiary of the same line.

Chapter 4

Vera CPU — A New Growth Engine for the Age of Agents

The second pillar was surprisingly the CPU. The Vera CPU, which Jensen Huang called "the fastest and most successful product launch in our company's history," shows why CPUs are becoming important again in the age of agents. And Intel's -4.67% on June 1 tells the other side of the story.

Agents do not just perform inference. They call tools, execute Python code, and run code in sandboxes. These tasks are the job of the CPU, not the GPU. Every time an agent works, both the GPU and the CPU wake up together. That is why Jensen Huang called the Vera CPU "the CPU for the age of agents" and "our new major growth driver."

"Vera is the CPU for the age of agents. This is going to be our new major growth driver. ... It's going to make it the fastest and the most successful product launch in our company's history."

Olympus Core — What Has Become Faster

The heart of the Vera CPU is NVIDIA's Olympus core. Eighty-eight Olympus cores operate on a single monolithic mesh; because the cores are not split into chiplets, inter-core communication is 50% faster than traditional methods. The fabric connecting the cores runs at 3.6 terabytes per second—in Jensen Huang's words, "the speed of light"—with no latency across chip boundaries. Featuring a 10-wide decode engine and a neural branch predictor that evaluates two branches per cycle simultaneously, it is optimized for Python runtimes, tool calls, and sandbox code execution.

What Intel's Decline Tells Us

The rise of the Vera CPU and N1X shakes up the old order of the CPU market in both data centers and PCs. Intel's -4.67% drop on June 1 is proof of this. The market is pricing in concerns that Intel, which has dominated PC and server CPUs for decades, could lose market share to ARM-based chips designed by NVIDIA (Vera CPU for data centers, N1X for PCs). If NVIDIA bundles its own CPU design following GPUs, the value add of AI data centers will become even more concentrated within the NVIDIA ecosystem.

From the perspective of Korean investors, the Vera CPU is a double-edged sword. Since it is NVIDIA's own ARM chip and manufactured by TSMC, the direct benefit path for Korean memory and substrates is not as strong as that of Vera Rubin (HBM4). However, if the Vera CPU changes the standard for data center servers, the demand for memory and storage in those servers will increase, which is indirectly favorable for Korean memory.

Chapter 5

Enterprise AI Toolkit — Every Company Builds Agents

The third pillar was software, not hardware. It is an enterprise AI toolkit that enables every company to build and operate agents securely. This pillar is directly linked to the growth of Korean IT service companies.

Open Shell — A Secure Harness to Contain Agents

In the keynote, Jensen Huang introduced a harness called Open Shell. It is a runtime that keeps agents bound within security policies and protects privacy and permissions so they can operate securely inside enterprises. It has been released as open source, and Red Hat, Canonical, and Microsoft have announced plans to adopt it.

"You could run it inside this harness called Open Shell which will be highly secure ... The shell protects the agent, keeps it grounded in security policies. ... Open Shell is open source. Red Hat, Canonical, Microsoft, it's going to be adopted everywhere."

Cadence — Chip Design Becomes 40x Faster

An example showing the actual power of the enterprise toolkit is the collaboration with Cadence. It accelerates the verification cycle of chip design by 40x using AI agents—shortening "weeks to hours." This super-agent is orchestrated by OpenAI's Codex or Anthropic's Claude Code. This scene proved that agents are not just simple chatbots, but tools that actually increase industrial productivity by dozens of times. The +7.00% rise of CrowdStrike, a security and agent beneficiary, on June 1 is also related to this trend.

Korean Companies Poised to Grow from This Pillar

The message "Every company builds agents" is a direct growth opportunity for Korean IT service and cloud companies. For an enterprise to adopt agents, someone must integrate them into their IT environment. The ones in charge of this integration are SI (System Integration) providers. In Korea, Samsung SDS, LG CNS, and NAVER Cloud play this role. In particular, LG CNS is already running its own LLM, EXAONE, and Agentic AI businesses, so it has a wide range of touchpoints to combine with the NVIDIA toolkit.

Companies poised to grow from this branch. In Korea, LG CNS, Samsung SDS, and NAVER Cloud, which will handle enterprise agent integration, are candidates. In the US, direct beneficiaries include Red Hat (IBM) and Canonical, which adopted Open Shell; Cadence for the chip design collaboration; and CrowdStrike (6/1 +7.00%) for security.

Chapter 6

Reinvention of the PC — "An Event as Big as a Phone Becoming a Smartphone"

This is the main driver that moved the US market most strongly on June 1. Announcing N1X, Jensen Huang said, "The reinvention of the computer is as big an event as the telephone becoming a smartphone," and the market responded with Dell +10.7% and NVIDIA +6.26%.

RTX Spark — 33 Years of Condensation

RTX Spark is not just a GPU, but an AI superchip condensed into a single chip. Blackwell RTX GPU with 6,144 CUDA cores, 1 petaflop of AI performance, a 20-core Grace CPU co-designed with MediaTek, and 128GB of unified memory. It packs 70 billion transistors on TSMC's 3nm process and was built as an agent platform in close collaboration with Microsoft Windows. Adobe redesigned the core architectures of Photoshop and Premiere Pro to match this chip, making them twice as fast and agent-friendly.

RTX Spark	Specifications
GPU	Blackwell RTX, 6,144 CUDA cores, 1 petaflop AI
CPU	20-core Grace CPU (Co-designed with MediaTek)
Memory & Process	128GB Unified Memory, TSMC 3nm, 70 billion transistors
OS & Software	Microsoft Windows, Adobe Photoshop & Premiere redesigned (2x faster)

N1X, DGX Station, and Market Reaction

N1X is NVIDIA's laptop SoC co-developed with MediaTek, which Jensen Huang called "the most amazing chip the world has ever made." DGX Station for Windows is a deskside supercomputer designed to sit next to your desk, featuring 768GB of memory, 20 petaflops, and 8 terabytes per second of bandwidth to run 1-trillion-parameter models.

This major PC overhaul was the market's biggest topic on June 1. Along with Jensen Huang's "smartphone-class event" remark, OEMs **Dell by +10.70% and HP by about +8%**surged. Conversely, at the center of the old PC chip order, Intel fell -4.67%. The landscape clearly showed that the benefits of the new PC architecture go to NVIDIA, MediaTek, Microsoft, and OEMs, while Intel of the x86 camp is threatened.

Companies set to grow from this branch. Direct US beneficiaries are NVIDIA, Dell, Microsoft, Adobe, and MediaTek (listed in Taiwan). In Korea, SK hynix and Samsung Electronics, which provide high-capacity memory for RTX Spark, N1X, and DGX Station, and Samsung Electronics for NAND SSDs, are secondary beneficiaries. The loser is Intel of the old PC chip order.

Chapter 7

Physical AI — Robotics, Autonomous Driving, and Hyundai Motor

The latter half of the keynote turned to AI in the physical world: Cosmos 3, which enables robots to understand reality; a humanoid reference robot; and Alpamayo 2, the "world's first reasoning autonomous vehicle." The Korean connection to this branch is Hyundai. Tesla's -4.57% on June 1 is read as a sign of intensifying competition in autonomous driving.

Cosmos 3 — World Model for Robots

Jensen Huang introduced Cosmos 3, calling it the "frontier of physical AI." Cosmos 3 is a world foundation model that helps robots understand the physical world. It generates physically accurate synthetic videos from images, text, and video, closes the loop on robot policy learning as a simulator, and serves as the foundation for NVIDIA Omniverse. It is a tool that enables infinite virtual training for robots before they are actually deployed in reality.

Humanoid Reference Robot and Alpamayo 2

NVIDIA unveiled Isaac GR00T not just as a software stack, but as an actual reference humanoid robot. It features 25 degrees of freedom (DoF) per hand, 31 DoF overall, a height of 6 feet, and a weight of 150 pounds. Integrated with the new Thor chip and NVIDIA's entire software stack, it is an open reference platform for universities and research institutions. In autonomous driving, Alpamayo 2 was unveiled as the "world's first reasoning autonomous vehicle." Operating on the Halos operating system and Hyperion runtime, Jensen Huang noted that brands adopting Hyperion represent approximately 80% of global vehicles and about 97% of mobility services.

Hyundai — Connected in Two Ways

In this physical AI branch, South Korea's strongest connection point is Hyundai Motor. There are two pathways. First, humanoids. Boston Dynamics, a world-class robotics company, is a subsidiary of Hyundai Motor. As NVIDIA expands the humanoid ecosystem with Isaac GR00T, Hyundai Motor, which owns Boston Dynamics, will stand at the center of that momentum. Second, autonomous driving. Hyundai and Kia are highly likely to be included in the 80% of global vehicles represented by Hyperion. Tesla's -4.57% drop on June 1 reflects the market's perception that NVIDIA's autonomous driving full-stack (Alpamayo 2) threatens Tesla FSD's monopoly. Conversely, Hyundai Motor, which belongs to the NVIDIA camp in the same autonomous driving trend, can expect to benefit.

Companies poised to grow in this branch. In South Korea, Hyundai Motor (Boston Dynamics + autonomous driving) and Hyundai Mobis (autonomous driving components) are the most direct connection points. For cameras and sensors, LG Innotek and Samsung Electro-Mechanics are candidates. Since South Korean listed humanoid companies (Doosan Robotics, Rainbow Robotics) did not appear in the keynote, approaching them is more thematic than fundamental.

Chapter 8

AI Factory — DSX and the Gigawatt Era of Power & Cooling

In addition to RTX (chips) and DGX (systems), Jensen Huang unveiled a third product line called DSX (factory infrastructure). This is an era where AI data centers are being standardized as a single "factory." However, as shown by the decline in power stocks on June 1, this branch sees a divergence between short-term profit-taking and long-term demand.

DSX — The Blueprint for the AI Factory

DSX is a reference design for building and operating AI data centers with maximum efficiency and profitability. It consists of Omniverse-based simulation (DSX SIM), operating system (DSX OS), DSX Max LPS to increase GPUs within a power budget, and DSX Flex to interface with the power grid. With this, NVIDIA's product lineup now has three tiers: RTX (GPU chips), DGX (systems), and DSX (factory infrastructure). AI data centers are no longer defined as buildings, but re-defined as "factories that mint tokens."

The economics of that factory are overwhelming. Jensen Huang stated that the cost of building a single 1-gigawatt AI factory has risen from around $30 billion in the past to $50–60 billion today, and will soon reach $80–100 billion. The fact that this astronomical investment is not an "expense" but an "asset that generates revenue" connects directly to the GDP generator argument in the keynote.

5,000 Amperes of Power, 45-Degree Cooling

What runs this factory is power and cooling. The cooling for the Vera Rubin NVL72 uses a 45-degree hot-water method to save water and energy, while the liquid-cooled busbars carry currents of over 5,000 amperes. This current is equivalent to 20 electric vehicles accelerating at full speed simultaneously. Jensen Huang said, "More power goes to computing that generates revenue." This is combined with the world's first 200GB co-packaged optical Ethernet switch Spectrum-X, Bluefield 4, and NVLink Fusion. The +2.43% rise on June 1 of Vertiv, a global leader in data center cooling and power, is a prime example of a beneficiary in this branch.

Power Stocks: Short-Term Profit-Taking vs. Long-Term Demand

There is one point worth noting. On June 1, nuclear power stocks Constellation Energy fell -7.66% and Vistra fell -3.41%. The keynote's emphasis on power "saving" technologies (45-degree cooling, efficiency per watt) served as a pretext for short-term profit-taking. However, this does not mean the trend in AI power demand is faltering. The absolute surge to the gigawatt scale, costing $100 billion per gate, supports the demand for transmission, distribution, and transformers. Even if efficiency improves, if the number of factories built multiplies, the overall demand for power infrastructure will grow. This is a phase where short-term profit-taking and long-term demand diverge.

Companies that will grow from this branch. The top direct beneficiary is Vertiv (6/1 +2.43%), a leader in data center cooling and power. In Korea, LS Cable & System and Taihan Cable & System, which manufacture 5,000-ampere busbar high-current cables, as well as LS Electric, Hyosung Heavy Industries, and HD Hyundai Electric, whose outer power transmission and distribution systems are expanding along with the rise of AI factories, are mid-to-long-term beneficiary candidates. However, as shown by the profit-taking in nuclear power stocks on 6/1, the power theme has high short-term volatility.

Chapter 9

The Moment Korea Was Called Out — Naver, Bank of Korea, and Hyundai Motor

While listing global AI infrastructure partners, Jensen Huang called out three Korean entities: Naver Cloud, the Bank of Korea, and Hyundai Motor. Though it was just a brief sentence, this callout stamped NVIDIA's seal of approval on the growth stories of these Korean companies.

"Here's Naver Cloud in Korea, Bank of Korea, Hyundai. So many incredible companies."

Naver Cloud — Korea's Sovereign AI Gateway

The first Korean company called out by Jensen Huang was Naver Cloud. While Naver is Korea's No. 1 search and content company, the context of this callout was AI cloud infrastructure. Naver Cloud is Korea's leading cloud service provider, operating NVIDIA GPU clusters to train and service its proprietary large language model, HyperCLOVA. From NVIDIA's perspective, Naver Cloud is the AI infrastructure gateway to the Korean market. This signifies that Korea's Sovereign AI—AI infrastructure that protects national data sovereignty—is growing on top of the NVIDIA ecosystem, and Naver is at the center of it.

Bank of Korea — And the Path Leading to LG CNS

The most unexpected callout was the Bank of Korea. It is rare for a central bank to be included in NVIDIA's list of global AI infrastructure partners. Let's follow the growth story painted by this callout. The Bank of Korea is currently conducting "Project Hangang," a central bank digital currency (CBDC) pilot project. The company that solely built the core system for this project, in which seven commercial banks (Kookmin, Shinhan, Woori, Hana, IBK, Nonghyup, and Busan) are participating, is LG CNS.

Jensen Huang's callout of the Bank of Korea as an AI infrastructure partner can be interpreted as a signal that a path has opened for Korea's digital currency and financial infrastructure to connect with NVIDIA's AI ecosystem. The most natural beneficiary along that path is LG CNS. LG CNS is the second-largest cloud MSP in Korea and the sole builder of the Bank of Korea's CBDC infrastructure. It is also a business poised to expand into KRW-pegged stablecoins and security token infrastructure once the Digital Asset Basic Act is passed. At the center of the trend where AI agents are integrated into financial infrastructure—such as fraud detection, automated settlement, and AI financial services—stand the Bank of Korea project and LG CNS, the company that built it.

The growth picture painted by the Bank of Korea callout. The Bank of Korea called out as an NVIDIA AI partner → The AI transition of Korea's financial infrastructure merges with the NVIDIA ecosystem → LG CNS, which built that infrastructure (CBDC Project Hangang), emerges as a key beneficiary candidate. LG CNS has already reached 'emperor stock' status and is growing diversely across AI data centers, EXAONE, and stablecoin infrastructure. The Bank of Korea callout adds NVIDIA as a global catalyst to this growth narrative.

Hyundai Motor — Cloud Customer and Physical AI Partner

The third mention is Hyundai Motor. While Hyundai Motor was named as an AI cloud customer, its significance extends to physical AI. This includes its humanoid credentials as the parent company of Boston Dynamics, and its autonomous driving credentials, with Hyperion set to be integrated into 80% of its vehicles. Among Korean companies, Hyundai Motor is the most deeply intertwined with the NVIDIA ecosystem across multiple fronts. Its growth story is unfolding simultaneously across three domains: AI cloud, humanoids, and autonomous driving.

Chapter 10

Growth Company Map — Korea and US (As of June 1 Closing)

This corporate map captures both the growth pathways outlined in the keynote and the market's verdict as of June 1. It summarizes the June 1 closing prices and fluctuations for US stocks, and the growth thesis for Korean stocks.

Korea Growth Map

Pillar	Key Companies	Growth Thesis
HBM4 Memory	SK Hynix · Samsung Electronics	Vera Rubin HBM4 supply by the 3 major makers, and explosive growth in the HBM market driven by perpetual agent inference
HBM Packaging	Hanmi Semiconductor · EO Technics · Soulbrain	HBM4 volume increase → Expansion of orders for TC bonders, dicing, and cleaning
Substrates · Cameras	Samsung Electro-Mechanics · LG Innotek	AI chip FC-BGA, humanoid & autonomous driving cameras
Physical AI · Mobility	Hyundai Motor · Hyundai Mobis	Boston Dynamics (humanoid) + Hyperion autonomous driving
Finance · AI Infrastructure	LG CNS · Naver	Bank of Korea mentioned → CBDC · Sovereign AI · Enterprise Agents
Enterprise SI	Samsung SDS · LG CNS	Integration of Enterprise AI Toolkit and Open Shell by Korean Companies
Power & Cables	LS Cable & System · LS ELECTRIC · Hyosung Heavy Industries · HD Hyundai Electric	Transmission, Distribution, and High-Current Infrastructure in the Gigawatt AI Factory Era

US & Global Growth Map (Including 6/1 Market Verdict)

Segment	Company · 6/1 Change	Growth Thesis
AI Infrastructure Core	NVIDIA +6.26% · TSMC +4.11%	Three Pillars + DSX. Exclusive Production of TSMC 3nm & CoWoS
Operations & Cloud	CoreWeave +13.96% · Dell +10.70%	Vera Rubin NVL72 Deployment Partner
Memory	Micron +6.64%	First time all three HBM4 makers named, full-scale entry
PC & Software	Adobe +5.72% · MediaTek	RTX Spark · N1X CPU, Photoshop Redesign
Cooling & Power	Vertiv +2.43%	Global No. 1 in Data Center Cooling & Power
Security & Agent	CrowdStrike +7.00%	AI Security, Beneficiary of Enterprise Agents
Losers (Caution)	Intel -4.67% · CEG -7.66% · Tesla -4.57%	PC Chip Old Order Threatened, Power Arbitrage, Autonomous Driving Competition

A Word from Four Perspectives

🧅 Onion (Macro/Bullish)

"This is the inflection point where AI has become a GDP generator. The record highs of the three major US indices on June 1st and NVDA +6.26% are the answer. The Asia-US alliance, which holds the hardware value chain, is leading the market. In Korea, SK Hynix and Samsung Electronics for HBM4, Naver for Sovereign AI, and LG CNS for financial infrastructure are showing joint strength. I see a major upward trend rotating through memory in 3 months, AI PCs in 6 months, and physical AI in 12 months."

🌰 Acorn (Industry/Value Chain)

"The real bottlenecks are HBM4, CoWoS, power, and cooling. HBM4 from SK Hynix and Samsung is the top priority, Hanmi Semiconductor's TC bonder is the key to back-end packaging, and Vertiv is the global top priority for cooling and power. The surge in Dell and CoreWeave on June 1st shows the actual demand for deploying Vera Rubin. However, the Vera Rubin announcement is only a declaration of 'full production,' and revenue recognition happens after delivery and inspection, so we expect the real volume to come between the second half of 2026 and the first half of 2027. The direction is clear, but we must watch the timing of revenue recognition."

Chapter 11

What Lies Ahead — 3, 6, and 12-Month Scenarios

By combining the growth outlined in the keynote with the market's initial reaction on June 1st, we map out how things will unfold chronologically. This is a scenario of what comes first, what follows, and which companies take the stage at each phase.

3 Months (Summer 2026) — Memory and AI PCs Move First

The market on June 1st has already pointed the way. The first to move were the major PC overhaul (Dell · NVIDIA) and HBM4 (Micron). These two tracks will continue through the summer. As HBM4 deliveries become visible alongside the full production of Vera Rubin, expectations will concentrate on SK Hynix and Samsung Electronics, followed by signals of Hanmi Semiconductor resuming TC bonder orders. Pre-marketing of RTX Spark and N1X laptops will fuel expectations for AI PCs. However, one point of caution is that TSMC described advanced packaging (CoWoS) as "very tight," and since Vera Rubin's 7-chip structure puts additional pressure on packaging lines, that "full production" does not immediately translate into "revenue recognition". Whether actual HBM4 mass production and delivery are confirmed in the July quarterly earnings of TSMC, SK Hynix, and Samsung Electronics will be the first major watershed.

6 Months (Second Half of 2026) — AI PC Launch and Financial Infrastructure

Moving into autumn, the stage broadens. The AI PC cycle kicks into full gear with the launch of RTX Spark and N1X laptops alongside the DGX Station. When the RTX Spark-supported version of Adobe Photoshop is released, the content ecosystem will move with it. Memory benefits will spread beyond data centers into PCs and on-device AI. At the same time, another track will grow in Korea. As the Bank of Korea's CBDC advancement project progresses and discussions on the Digital Asset Basic Act move forward, LG CNS's financial infrastructure story will take shape. Naver Cloud will grow on the back of Sovereign AI demand. This is also around the time when Vera CPUs begin to penetrate data center servers.

12 Months (First Half of 2027) — Physical AI and Agents Become Reality

One year from now, the stage will belong to physical AI. Humanoids trained on Cosmos 3 and Isaac GR00T will begin entering industrial sites, and Alpamayo 2-based autonomous driving will move beyond the pilot phase. At the center of this trend in Korea will be Hyundai Motor (Boston Dynamics/autonomous driving) and Hyundai Mobis. At the same time, enterprise agents (Open Shell) will spread across corporate environments, with SI providers like Samsung SDS and LG CNS generating integrated revenue. AI factories will expand to gigawatt scales, driving demand for power, cooling, and cabling. At this stage, whether "AI is a GDP generator," as declared in the keynote, is proven through actual financial performance will be key to sustaining the cycle.

📅 Growth Stages and Key Companies

Timeline	Stage	Key Korean Companies	Key US Companies	Verification Events
3 Months	HBM4 & AI PC	SK Hynix, Samsung Electronics, Hanmi Semiconductor	Micron, Dell, NVDA	July Quarter Earnings (Mass Production Verification)
6 Months	AI PC Launch & Financial Infrastructure	Memory + LG CNS, Naver	Adobe, MediaTek	RTX Spark Launch, Digital Asset Act
12 Months	Physical AI & Agents	Hyundai Motor, Hyundai Mobis, Samsung SDS, LG CNS	NVDA, Vertiv, Cadence	Humanoid & Autonomous Driving Commercialization, Capex ROI

Chapter 12

What to Guard Against

Even if the growth story is solid and the US market responded with strength on June 1, pricing is a different matter. Here are the warning signals highlighted by Peanut (focusing on fundamentals) and Carrot (focusing on charts).

🥜 Peanut (Fundamentals/Bearish)

"Although the three major US indices hit record highs, the positive news from the keynote is already largely priced in. Korea's KOSPI also reached a record high on June 1, and foreign investors are taking profits with 16 consecutive trading days of net selling. Five risks to watch out for: priced-in positive news, dilution of the Korean premium due to intensifying HBM4 competition, deferred revenue recognition from CoWoS bottlenecks, skepticism over the ROI of hyperscaler capex, and profit-taking at record highs. While I advise against panic selling, it is reasonable to partially trim short-term gains while holding core positions. Since the HBM monopoly and thematic conviction remain valid, selling everything would be an overreaction."

🥕 Carrot (Charts/Sentiment)

"On June 1, the US market temporarily put sell-on-news concerns to rest, with NVDA rising +6.26%. However, the situation in Korea is different. Foreign investors recorded their largest-ever net selling in May, and a shift back to buying has not yet been confirmed. Since Korea's record highs have been supported primarily by retail and institutional investors, the upside remains capped until foreign investors return. Given the strong US close, the Korean market is likely to face upward gap pressure at the open in early June. However, it is safer to avoid chasing the rally, buy in installments during pullbacks that partially fill the gap, and use daily foreign net buying as a buy trigger."

How to Read Growth and Price Together

Synthesizing these four perspectives leads to a clear conclusion: The bullish camp (Onion & Acorn) is right about the direction of growth, while the bearish camp (Peanut & Carrot) is right about the timing of entry. The growth drivers outlined in the keynote — HBM4, AI PCs, financial infrastructure, and physical AI — are real, and the US market confirmed this direction on June 1 with strong performances from NVDA, Dell, Micron, and CoreWeave. However, these expectations are already largely reflected in the record highs of both the US and Korean markets, and foreign investors in Korea are still selling. Therefore, the practical takeaway from this keynote is to buy growth companies but avoid chasing them at record highs, waiting instead for pullbacks and a turnaround in foreign buying to build positions in installments.

Key Warning Signals — What to Watch

Warning Signal	Indicators to Watch	Implications
Deferred Revenue Recognition	July TSMC CoWoS guidance, quarterly memory earnings	The timing when "full production" translates into actual revenue
Capex ROI Skepticism	Meta · MS · Google capex guidance	Revenue validation relative to AI investment — the biggest variable for cycle sustainability
Foreign Capital Flows in South Korea	Shift to daily foreign net buying	Sustainability of South Korea's new highs, buying triggers
Power Theme Volatility	Nuclear and power stocks like CEG · VST	Distinguishing short-term profit-taking vs. long-term AI power demand

Epilogue

Between the Announcement of Production and Revenue Recognition

The growth roadmap drawn by the keynote is clear, and the market confirmed its direction on June 1. The remaining question is a matter of time: when will that growth translate into earnings?

The June 1, 2026 keynote redefined AI from a cost to a "GDP generator." On top of this, Jensen Huang established three pillars — Vera Rubin, Vera CPU, and Enterprise AI Toolkit — and added the PC overhaul, physical AI robots and autonomous driving, and gigawatt AI factories. And on that stage, he called out three South Korean entities: NAVER Cloud, the Bank of Korea, and Hyundai Motor. That night, the US stock market responded to the announcement with all three major indices hitting historic highs, and NVIDIA surging +6.26%.

The growth roadmap drawn by this announcement unfolds in several directions for Korean investors. The most direct connection is SK Hynix and Samsung Electronics for HBM4 memory, and Hanmi Semiconductor for its back-end packaging. For physical AI and autonomous driving, Hyundai Motor and Hyundai Mobis, which houses Boston Dynamics, are linked. Following the path opened by the mention of the Bank of Korea, LG CNS—which builds CBDC infrastructure and expands into financial AI and stablecoins—and NAVER with its Sovereign AI will grow. As enterprise agents proliferate, Samsung SDS and LG CNS will follow, and as gigawatt factories increase, power and cable companies will benefit. In the US, the market already gave its answer on June 1 — NVIDIA, Dell, CoreWeave, Micron, Adobe, and Vertiv were the winners, while Intel, Tesla, and power stocks were the short-term losers.

The remaining question is not about the stocks, but the timing. Jensen Huang declared that Vera Rubin is in "full production." However, full production means the manufacturing lines are running, not that revenue is being recognized. Considering TSMC's comment that advanced packaging is "extremely tight," full-scale revenue recognition could be deferred to the second half of 2026 or the first half of 2027. While the direction of growth is certain, the timing of when it translates into actual earnings must be verified on a quarterly basis.

Therefore, the practical conclusion from this keynote is simple: Buy growth companies, but do not chase them at new highs. Select growth companies in South Korea and the US along the branches drawn by the keynote — memory, AI PCs, financial infrastructure, and physical AI — but since expectations are already priced in at all-time highs, wait for pullbacks and a shift in foreign buying to enter in tranches. Price will be revealed over time, and the first slot on the timeline is the July quarterly earnings. Once it is confirmed whether "full production" is converting into "revenue" at that time, we will know if the first branch of this growth roadmap was real.

"Jensen Huang drew the roadmap for growth, and the market confirmed that direction on June 1.
All that remains is the time it takes for that roadmap to become actual earnings.
The direction is clear. What we must watch out for is timing."

NVIDIA GTC Taipei 2026 — Keynote & Market Verdict & Growth Roadmap · Written on June 2, 2026
Keynote Presentation (2026/6/1) + Reflecting US Market Close on 6/1 · Three Pillars and Growth Company Roadmap · 3, 6, 12-Month Outlook
Team Goguma: Onion (Macro) · Acorn (Industry) · Peanut (Fundamental) · Carrot (Chart) · Sweet Potato (Comprehensive)
This report is for informational purposes only and does not constitute a recommendation to buy or sell.