● Nvidia Ouster Rumor, KOSPI Crash Fear, AI Cashflow Risk
Assessing Claims of Sharp Declines in Nvidia and the KOSPI: AI Investment Risks Beyond Reports of Jensen Huang’s Dismissal
The first issue to verify is not the share-price outlook, but whether the premise that Jensen Huang was dismissed is accurate.
Without confirmation, the proposed chain of events linking a withdrawal of Nvidia funding, an AI ecosystem crisis, and a KOSPI sell-off remains unsubstantiated.
The analysis below examines the evidence required to assess the source’s principal claims, Nvidia’s transmission channels to Korean equities, and cash-flow risks in AI investment.
Scope: This assessment is based on the supplied source text.
Real-time filings, market prices, and government announcements were not reviewed. Reported management changes and market movements are therefore not presented as verified current facts.
1. Reviewing the Principal Claims: Significant Events Require Reliable Sources
The source links Nvidia management changes, U.S. government intervention, declines in major equity indices, and an IPO timetable.
However, it provides no event dates, official announcements, or corresponding market data.
References to “Friday,” “the weekend,” “today,” and “this week” cannot be applied to current markets without establishing the source’s publication date.
| Source Claim | Current Assessment | Required Evidence |
|---|---|---|
| Jensen Huang was dismissed as Nvidia CEO | Not verifiable from the supplied source alone | Official Nvidia announcements, board-related disclosures, and SEC filings concerning executive changes |
| Treasury Secretary Scott Bessent demanded the appointment of an outside CEO | The original statement and an authoritative source are required | U.S. Treasury statements, interview transcripts, and reliable reporting |
| Prices rose over the weekend on expectations of Huang’s return | The instrument and trading period are unclear | Identification of cash equities, futures, or separate trading platforms; price and volume data |
| Nvidia earnings and an Anthropic IPO are scheduled for this week | The reference date and official schedules require verification | Nvidia’s investor relations calendar, Anthropic announcements, securities registration statements, and exchange notices |
| The event caused declines in the Nasdaq and the KOSPI | Insufficient evidence to establish causation | Daily index and sector returns, foreign investor flows, exchange rates, interest rates, and concurrent news |
“Unverified” and “confirmed false” are distinct assessments.
The alleged dismissal and unsuccessful return should be treated as claims requiring official evidence, not established facts.
2. Claims About Huang’s Dismissal and Return: Separate Executive Changes From Government Pressure
① Corporate disclosures are central to verifying a CEO change.
Material executive changes at U.S.-listed companies such as Nvidia can generally be verified through official announcements and SEC filings.
A report that omits the successor’s identity, the effective date, and whether the appointment is interim provides an inadequate basis for investment decisions.
② Government preferences differ from board appointment decisions.
The U.S. government can influence corporate operations through export controls, national security reviews, and competition policy.
However, an alleged Treasury Secretary preference regarding an executive is separate from legal authority to appoint or remove a CEO.
CEO appointments and dismissals generally fall within the board’s responsibilities. Claims of government intervention require specific legal grounds and documented official action.
③ A management change would not automatically imply reduced investment.
Existing supply contracts, board approvals, customer purchase commitments, and financial strategies do not immediately disappear when a CEO changes.
The proposed sequence of “new CEO → withdrawal of funding → collapse of the AI ecosystem” therefore requires verification at multiple stages.
3. Assessing the Claim That Nvidia Distributed Funding Like a Central Bank
Nvidia is a core AI infrastructure supplier, not a central bank.
It neither issues legal tender nor provides system-wide liquidity as a lender of last resort.
The comparison may describe how investments in AI companies, technical partnerships, product supply, and software ecosystem development can support demand.
The relevant issue is transaction structure, not the characterization of funding as monetary distribution.
- Equity investments: Purchases of ownership interests in AI companies must be distinguished from product sales.
- GPU sales: Customer purchases of chips or systems require scrutiny of revenue recognition and payment collection.
- Supply and purchase commitments: Agreements securing future capacity do not necessarily involve immediate cash transfers equal to the announced value.
- Loans, guarantees, and credit support: Where present, these arrangements require assessment of borrower repayment capacity and ultimate loss exposure.
- Technology and service support: Software assistance and usage credits differ economically from cash investments.
Combining these categories without examining individual contracts can distort estimates of AI investment and risk.
The more relevant question is: Who ultimately pays for AI infrastructure, and where does that funding originate?
4. How Nvidia-Related Developments Affect the KOSPI
Adverse Nvidia-related developments can weigh on Korean equities.
However, assessing the KOSPI outlook requires greater attention to the following four transmission channels than to one executive’s position.
① AI semiconductor order expectations → Korean semiconductor earnings forecasts
Lower expected demand for AI accelerators can affect earnings estimates for suppliers of HBM, server memory, packaging, and selected equipment and materials.
Conversely, declines in U.S. AI-related equities may have a limited earnings impact on Korean suppliers if orders and supply contracts remain intact.
A share-price correction and an order decline should not be treated as the same event.
② U.S. equity risk aversion → Foreign capital flows
Global investors reducing technology exposure may also sell large Korean semiconductor stocks.
In such cases, portfolio-wide risk reduction may matter more than company-specific developments.
③ Exchange-rate movements → Different effects on capital flows and earnings
Won depreciation can reduce foreign investors’ dollar-denominated returns.
It can also support exporters’ reported revenue through currency translation.
The net effect depends on dollar-denominated costs and hedging arrangements. Currency depreciation is not uniformly positive or negative across companies.
④ Higher interest rates → Pressure on growth-stock valuations
Rising U.S. Treasury yields can pressure growth stocks whose valuations depend heavily on distant earnings.
Interest-rate expectations and changes in long-term yields affect AI-related equities and the KOSPI independently of Nvidia news.
Assessing a daily sell-off therefore requires a combined review of semiconductor sector declines, foreign investor activity in cash equities and futures, the won-dollar exchange rate, and U.S. Treasury yields.
5. Missing Information in the “Weekend Rebound Followed by Renewed Declines” Narrative
Regular stock trading on major U.S. exchanges generally does not operate on Saturdays or Sundays.
Any reported “weekend rebound” must first be linked to a specific market and instrument.
- Determine whether Friday after-hours trading was described as weekend activity because of time-zone differences.
- Determine whether the reference concerns equity index futures reopening on Sunday evening in the United States.
- Check whether prices came from separate products or limited trading facilities offered by certain platforms.
- Establish whether the statement describes actual executed trades or expectations following news.
Cash equities, index futures, and prices formed under limited liquidity are not interchangeable.
Low-volume price movements should not be generalized as representative of the broader investor base.
6. Nvidia Earnings and an Anthropic IPO: Separate Schedule Verification From Impact Assessment
① Nvidia’s next-quarter outlook matters alongside reported revenue.
Once the official release date is verified, analysis should cover revenue and earnings relative to consensus, forward guidance, and gross margin.
Product transition costs, supply constraints, customer concentration, and export restrictions also affect interpretation.
Strong results can still coincide with a share-price decline if valuations already reflect higher expectations.
② Reports of an Anthropic IPO process must be distinguished from a confirmed timetable.
Considering a listing, appointing underwriters, submitting listing documents, setting an offer price, and commencing trading are separate stages.
The supplied source does not establish that an Anthropic IPO is scheduled for “this week.”
Private fundraising or secondary sales by existing shareholders should not be confused with an IPO.
③ Selling ahead of major events is plausible but does not establish the cause of a decline.
Investors may reduce positions or hedge ahead of earnings announcements.
However, attributing a sell-off solely to uncertainty does not adequately explain its scale or drivers.
Trading volume, investor flows, options-market activity, and sector performance are needed to assess the actual market response.
7. Additional Analysis: AI Ecosystem Cash Flows Matter More Than Headlines
There is insufficient evidence to conclude that other media outlets have overlooked these issues.
Nevertheless, executive-related headlines can obscure the recoverability of AI investment and the structure of its financing.
① Is investment funding translating into revenue from end customers?
When funded AI companies purchase cloud services or GPUs, supplier revenue can increase.
This structure does not itself indicate financial distress or accounting irregularities.
However, if businesses depend more on external funding than on revenue from paying customers, tighter financing conditions could weaken infrastructure demand.
Key metrics include recurring revenue, customer retention, service delivery costs, cash burn, and additional financing requirements.
② Is the same funding being counted multiple times?
The same capital may flow through startup financing, cloud service contracts, data center capital expenditure, and semiconductor purchases.
Adding each announced amount as independent final demand risks overstating the size of the AI market.
Intercompany transaction values must be distinguished from actual end-customer expenditure.
③ Do long-term contracts represent noncancelable, committed revenue?
Even large contracts vary in economic value depending on purchase obligations, cancellation provisions, delivery schedules, and customer payment capacity.
Order backlogs, recognized revenue, and cash receipts must therefore be assessed separately.
④ Can GPU investment payback periods keep pace with technological change?
Rapid improvements in AI chip performance and energy efficiency can pressure rental rates and utilization of existing equipment.
Expected equipment service life may differ from its economically profitable operating life.
Investment returns should incorporate power costs, utilization, and maintenance expenses alongside depreciation assumptions.
⑤ Does growth in AI usage necessarily produce proportional growth in semiconductor revenue?
Model compression, inference optimization, and custom-designed chips can reduce computing costs per task.
Lower costs may also stimulate usage and increase aggregate infrastructure demand.
The AI industry outlook therefore depends partly on the relative pace of usage growth and unit-cost declines.
8. A Practical Verification Sequence for Investors
- Establish the source date: Translate “today” and “this week” into specific calendar dates.
- Review official evidence: Prioritize corporate filings for executive changes, official records for government statements, and company or exchange documents for listing schedules.
- Cross-check market data: Determine whether news release times align with the reported price movements.
- Separate flows from fundamentals: Distinguish risk-reduction selling from changes in customer orders and earnings forecasts.
- Assess embedded expectations: Evaluate results against what the market had already anticipated.
- Review concentration risk: Holdings across multiple companies may provide limited diversification if they depend on the same AI capital-expenditure cycle.
The source’s claim that it “anticipated the decline” is not a substitute for supporting evidence.
Forecast quality requires evaluation of the timing, conditions, projected range, and response criteria specified before the decline.
The appropriate conclusion is not that AI investment has ended, but that unverified executive-related reports must be separated from underlying industry cash flows.
< Summary >
- The supplied source does not verify Huang’s dismissal or U.S. government intervention in Nvidia’s CEO appointment.
- Claims of a “weekend rebound” and an “IPO this week” require identification of the market, reference date, and official timetable.
- Nvidia is not a central bank. Investments, sales, contractual commitments, and credit support must be distinguished.
- Assessing KOSPI exposure requires a combined review of AI semiconductor orders, foreign investor flows, exchange rates, and interest rates.
- Principal AI investment risks concern end-customer revenue, financing dependence, contract terms, and equipment investment payback.
[Related Articles…]
*Source: [ 내일은 투자왕 – 김단테 ]
– 젠슨황 컴백 실패? #젠슨황 #엔비디아 #코스피
● Nvidia Shock, CEO Exit Rumor, AI Chip Panic
Nvidia Sell-Off Reports and Claims of Jensen Huang’s Resignation: Key Variables for AI Semiconductor Investors
Nvidia’s investment outlook depends less on the magnitude of a share-price decline than on changes in expected earnings and market confidence.
This report separates unverified claims from established analytical considerations and examines management risk, AI infrastructure financing, semiconductor export restrictions on China, U.S. interest rates, and the dollar.
Particular attention is given to the relationship between GPU collateral values and lending terms, the distinction between customer billing locations and end-use locations, and the relationship between share-price declines and corporate value.
1. Separate Verified News from Unsubstantiated Claims
The source material combines corporate performance, management-related claims, political interpretations, and market forecasts.
Accepting its conclusions without verification could result in investment decisions based on unsupported assumptions.
| Principal Claim | Evidence Required | Interpretation Pending Verification |
|---|---|---|
| Nvidia shares have declined 33% from their peak | Reference date, peak date, intraday versus closing prices, and split-adjusted price data | Do not treat this as the current drawdown without a defined period and price basis |
| Jensen Huang has resigned and an interim CEO has been appointed | Official Nvidia announcement and relevant filings, including SEC Form 8-K | Do not assume a leadership transition without official confirmation |
| Retail investors recorded their largest-ever net purchases | Data provider, account coverage, measurement period, and estimation methodology | Do not interpret this as a market-wide record without confirming the source and scope |
| Investigations, hearings, and passage of semiconductor legislation are scheduled | Congressional committee notices, legislative text, and official schedules | Distinguish political statements from formal procedural developments |
| Anthropic will list that week | Company announcement, public securities registration statement, and exchange notice | Distinguish listing preparations, registration, the offering, and the start of trading |
Primary verification sources include Nvidia Investor Relations, SEC EDGAR, Congress.gov, and the U.S. Bureau of Industry and Security.
These links identify verification resources; they do not substantiate the individual claims in the source material.
2. Nvidia’s Share-Price Outlook: Does a 33% Decline Imply Undervaluation?
A substantial price decline does not necessarily make a stock inexpensive.
Valuation should reflect prospective earnings, cash flow, growth, and risk rather than the previous share-price peak.
① Separate earnings revisions from valuation-multiple changes.
In simplified terms, share price equals expected earnings per share × the price-to-earnings multiple.
If expected earnings per share decline by 10% and the valuation multiple contracts by 25%, the implied share price falls by approximately 32.5%.
This is an illustrative calculation, not a Nvidia forecast. It demonstrates that a decline exceeding 30% can result from simultaneous earnings deterioration and risk repricing.
② Larger losses require disproportionately larger recovery gains.
- A 20% decline requires a 25% gain to recover the original price.
- A 33% decline requires an approximately 49.3% gain.
- A 50% decline requires a 100% gain.
A short-term rebound therefore does not establish full loss recovery or a trend reversal.
③ Historical drawdowns inform risk management but do not establish a recurring cycle.
Nvidia has experienced substantial volatility, but there is insufficient evidence for a rule that its shares halve every four years.
Earlier gaming GPU and cryptocurrency-mining demand differed from today’s data center and AI semiconductor demand in customer composition and investment structure.
The more relevant question is whether the current correction reflects market-wide discount-rate changes or deterioration in the company’s competitive position.
3. Management Risk: Business Continuity Matters More Than the Founder Premium
Key-person risk arises when an individual founder has a substantial influence on corporate value.
The relevance of this concept does not establish that the leadership change alleged in the source material occurred.
① What supports a founder premium?
- Product strategy: The ability to allocate resources across technologies and markets.
- Supply-chain coordination: Execution across semiconductor manufacturing, advanced packaging, memory, and server supply.
- Customer confidence: Relationships supporting long-term investment coordination with major customers.
- Ecosystem development: The ability to expand software and developer adoption alongside hardware.
② A CEO transition does not automatically terminate contracts.
Major supply and financing agreements are generally entered into by the corporate legal entity.
The impact of management changes depends on termination provisions, key-person clauses, investment commitments, and customers’ actual purchasing intentions.
The importance of personal relationships alone does not justify concluding that a substantial share of business will disappear.
③ A rapid succession process is not sufficient on its own.
If a leadership transition is confirmed, investors should assess the successor’s appointment, retention of technical teams, product launch schedules, and capital allocation policies.
The board’s and management’s ability to address legal and compliance matters while preserving business continuity may matter more than an individual’s potential return.
④ The source material’s CEO-transition statistics are insufficient as investment evidence.
The company sample, event-selection criteria, return measurement period, benchmark, and calculation methodology are not disclosed.
In small samples, company size, economic conditions, and pre-existing operating weakness can materially affect results.
4. AI Infrastructure Risk May Depend More on Customer Financing Than Chip Sales
The AI industry connects semiconductor manufacturers, cloud providers, AI model developers, data center operators, and financial institutions.
These relationships are not inherently problematic, but excessive reliance on shared growth assumptions can transmit shocks across companies.
The following describes general risk channels; it does not verify the specific guarantees or loans mentioned in the source material.
① Examine structures linking investment and revenue.
A semiconductor company may invest in a customer that subsequently uses the proceeds to purchase its equipment.
Such arrangements do not, by themselves, establish financial distress or accounting misconduct.
Investors should assess end-user demand, transaction terms, and whether the customer can sustain purchases without external financial support.
② GPU-backed lending links technological change to financing conditions.
Companies borrowing against GPUs or servers are exposed to equipment resale values and rental income.
New product launches or increased supply can reduce existing equipment values, potentially triggering additional collateral requirements or less favorable refinancing terms.
Higher interest rates can further constrain investment capacity, even when equipment utilization remains sound.
③ Risk differs across AI businesses.
- Large cloud providers with strong cash generation: Operating cash flow provides greater flexibility to adjust investment.
- Neocloud providers dependent on external funding: Lending terms, customer concentration, and GPU utilization may have greater financial significance.
- Data center developers: Power availability, permits, construction schedules, and tenant creditworthiness are critical.
- Semiconductor suppliers: Order cancellation exposure, receivables collection, and inventory commitments require monitoring.
The central question is not only whether AI demand will grow, but also who can sustainably finance that demand.
5. Semiconductor Export Restrictions on China: Distinguish Billing Locations from End Use
U.S.-China technology competition and semiconductor export restrictions affect individual companies and the broader global economic outlook.
Potential regulatory tightening, enacted legislation, and established violations are distinct developments.
① Customer billing locations do not necessarily identify end-use locations.
A multinational company’s headquarters, purchasing entity, delivery destination, and operating data centers may be located in different jurisdictions.
Revenue concentration in a particular country does not, by itself, establish export circumvention or unlawful transactions.
Conversely, reported billing geography alone does not provide a precise breakdown of final demand by region.
② Claims of “no China revenue” require a defined scope.
Restrictions on specific advanced AI accelerators are not equivalent to the elimination of all China-related revenue.
Product categories, customer locations, and geographic reporting definitions must be examined before concluding that total China revenue is zero.
③ Legislative provisions and procedural status matter more than bill names.
- Determine whether a bill has been introduced, cleared a committee, or become law.
- Assess whether its scope covers finished products, semiconductor equipment, or cloud access.
- Implementation periods, licensing exceptions, and treatment of allied countries can produce different company-level effects.
The source material’s description of “three chip regulation bills” cannot be verified without bill numbers and legislative text.
Chip-location verification or usage-monitoring requirements should not automatically be interpreted as mandates to install physical GPS trackers.
④ Tighter restrictions may also affect competitors.
If restrictions apply according to product performance and end use rather than company identity, Nvidia’s competitors may face similar constraints.
Adverse developments for Nvidia therefore do not automatically benefit other semiconductor companies.
6. Implications for U.S. Equities: Separate Competitive Gains from Sector Rotation
① Opportunities differ across AMD, Broadcom, and proprietary AI chips.
AMD’s alternative GPUs, the custom semiconductor ecosystem involving Broadcom, and proprietary accelerators from Alphabet and Amazon may benefit from efforts to reduce dependence on Nvidia.
However, customers consider more than chip prices when switching suppliers.
Total cost of ownership includes software compatibility, model migration, development staffing, network configuration, and energy efficiency.
Short-term relative share-price gains must be distinguished from actual shifts in orders and revenue.
② Classifying Apple as unrelated to AI is an oversimplification.
Companies differ in AI investment strategies and monetization models; lower relative infrastructure spending does not imply an absence of AI exposure.
Share-price performance also reflects product sales, services revenue, repurchases, and differences in prior expectations.
③ Rotation into healthcare and consumer staples requires performance evidence.
Capital may shift toward sectors with stable cash flow during growth-stock corrections, but this pattern is not universal.
Obesity-treatment companies require assessment of clinical outcomes, regulatory approvals, insurance coverage, manufacturing capacity, and drug-pricing competition.
Costco’s special dividends are not guaranteed recurring distributions; announcements and ex-dividend effects should be evaluated separately.
The source material’s references to importing obesity treatments into South Korea raise separate questions concerning import rules, prescriptions, and safety.
④ Other major technology developments also require verification.
Claims concerning cloud outages, competing productivity tools, potential mergers, and entry into prediction markets require official announcements or service-status records.
Assessing earnings implications requires outage duration and customer compensation, paid-user conversion for new services, and counterparties and terms for potential mergers.
7. Interest Rates and the Dollar: External Risks to Nvidia
Strong corporate performance does not prevent share-price declines when market discount rates rise.
① Higher rates can pressure both growth valuations and infrastructure financing.
Companies whose valuations depend heavily on distant future earnings may be particularly sensitive to discount-rate changes.
Higher borrowing costs for data centers and servers can also affect customers’ investment plans.
Market responses nevertheless depend on whether rates rise because of stronger growth, inflation, or fiscal concerns.
② Dollar strength affects companies and investors through different channels.
U.S. companies may face foreign-revenue translation effects and reduced purchasing power among overseas customers.
For South Korean investors, changes in the dollar-won exchange rate affect returns measured in won.
Performance assessment should therefore incorporate currency movements and hedging arrangements rather than Nvidia’s share price alone.
③ Low index volatility does not imply low single-stock risk.
The VIX primarily measures expected volatility implied by S&P 500 options.
Broad index stability does not eliminate risks specific to semiconductor companies or AI infrastructure stocks.
The S&P 500 also has substantial exposure to major AI-related companies and should not be classified as a market unrelated to AI.
8. Five Underappreciated Linkages to Examine Before Reacting to Headlines
These considerations are not necessarily absent from other coverage, but they can be overshadowed by high-profile management headlines.
① Revenue collection matters as much as reported revenue growth.
If receivables grow faster than revenue, investors should examine payment terms and collection periods.
Timing differences may be temporary; assess operating cash flow and multiple reporting periods before drawing conclusions.
② Falling GPU prices can benefit buyers while weakening borrowers.
Lower equipment prices reduce entry costs for companies developing new AI services.
They may disadvantage operators that purchased equipment at higher prices and borrowed against it.
The same industry development can therefore produce opposing financial effects.
③ Regulatory tightening does not necessarily reduce industry-wide demand proportionately.
Demand elsewhere may absorb supply that can no longer be sold into a restricted region.
The earnings impact depends on supply availability, alternative customers, product mix, and pricing power.
④ Competitive gains during downturns may depend more on migration feasibility than supply capacity.
Even with available chips, competitors may struggle to capture revenue if customers face substantial software migration barriers.
Substitution can accelerate when switching costs are low and cost savings are material.
AI training and inference, and general-purpose GPUs and custom accelerators, should not be treated as a single homogeneous market.
⑤ Misinformation can move prices without establishing corporate value.
Higher trading volume or price volatility following a rumor does not validate the rumor.
Until official confirmation is available, distinguish market reactions from the truth of the underlying claim.
9. Earnings and Investment Decision Checklist
Relative dates such as “next week” or “next January” should not be entered into an investment calendar without confirming the source’s publication date and the official schedule.
Earnings releases and events should be checked against company or institutional announcements.
- Growth: Assess whether data center revenue and next-quarter guidance meet market expectations.
- Profitability: Examine how product transitions, sales mix, and supply costs affect gross margin.
- Cash flow: Determine whether net income growth translates into stronger operating cash flow.
- Financial exposure: Monitor changes in inventory, receivables, purchase commitments, and investment commitments.
- Customers: Review customer concentration, major customers’ capital expenditure plans, and dependence on external funding.
- Regulation: Confirm the products covered by sales restrictions and the company’s quantified impact estimates.
- Management: If leadership changes are officially confirmed, assess succession plans and product-roadmap continuity.
Recovery scenario: Demand and cash flow remain resilient while discount rates and company-specific uncertainty ease.
Range-bound scenario: Operating performance remains sound, but elevated expectations or regulatory uncertainty delay valuation-multiple recovery.
Further downside scenario: Customer investment reductions coincide with downward earnings revisions and higher financing costs.
None of these scenarios is an established forecast. Assessments should be updated as new filings and earnings data become available.
Staged purchases diversify entry timing; they do not eliminate the risk of declining fundamental value.
Retail net buying, online polls, prediction-market prices, and magazine covers are supplementary sentiment indicators, not sufficient standalone investment grounds.
A long investment horizon does not guarantee recovery. Investment decisions should reflect liquidity needs, portfolio concentration, time horizon, and loss tolerance.
< Summary >
- Claims concerning Jensen Huang’s resignation, a 33% Nvidia drawdown, hearings, and listing schedules should not be treated as established facts without official verification.
- Nvidia’s investment merits should be assessed through expected earnings, valuation multiples, and cash flow rather than its previous peak price.
- Customer financing, GPU collateral values, and receivables collection are key AI infrastructure risk variables.
- Export restrictions require analysis of legislative status and scope, with a clear distinction between billing locations and end-use locations.
- Potential competitive gains should be validated through software migration costs and actual order growth.
- A balanced assessment should incorporate U.S. interest rates, the dollar, earnings expectations, and portfolio concentration.
[Related Articles…]
*Source: [ 소수몽키 ]
– 엔비디아-33% 폭락, 젠슨 황 사임 나비효과가 증시 혼란 더 키울까


