SNOWSHAGALChina remote-access curbs: Nvidia risks losing guidance upside before revenueNO.41
SNOWSHAGALINVESTMENT INSIGHTNO.41
SYS//ALERTNVIDIA × U.S. CHIP RULES

CHINA REMOTE-ACCESS CURBSSALES MAY HOLD
BUT UPSIDE AT RISK

CHIP SUPPLY: FULLFULLWAITLISTQUEUEPermit?If a license is denied,→ the queue fills the slotQ-001NEXT IN LINEQ-002Q-003⋯ ×2.8~6.52.8~6.5vs. demand at riskSCENARIO · KIWOOMChips can go elsewhere.Buildings cannot.→ Overseas data centers bear the strain
760× H100License above this threshold
(12 million TPP)
16.3vs46.2–105.9Demand at risk vs. unmet demandUSD bn · Kiwoom estimates
90DAYSExisting facilities:
license deadline
CHIPS MOVE // BUILDINGS STAYSNOWSHAGAL.COM
INVESTMENT INSIGHT NO.41 ·

The U.S. now requires licenses for chips supplied to overseas data centers that provide AI computing to Chinese companies. With chips still scarce, however, other customers could take the freed capacity. Kiwoom Securities therefore sees a bigger loss to Nvidia's ability to raise its revenue guidance than to near-term sales.

THREE QUESTIONS
  1. Q1What exactly does the new rule restrict?01
  2. Q2Will shutting out Chinese customers cut Nvidia's revenue?02⁠·03
  3. Q3Who bears the cost—and what should investors watch?04⁠·05
+12%Q4 revenue guidance above Q3 despite assuming zero China data-center compute revenue
77%Potential demand loss ÷ November increase in FY28 outlook (Kiwoom estimate)
0.9GWAdditional data-center commissioning delay that would put guidance at risk under the base case (Kiwoom)

Sources: U.S. Federal Register, Nvidia, Kiwoom Securities, DayOne Form F-1, IBK Securities, Yuanta Securities and others

Published Nov. 21, 2026 · Data through the Nov. 20 U.S. close (morning of Nov. 21 in Korea) · About 9 min read

SUMMARYFive Key Points
  1. The U.S. has put licensing conditions on AI computing rented to Chinese companies. An overseas data center providing more than roughly 760 H100-equivalent GPUs of processing to a Chinese-headquartered company must secure a license before receiving covered U.S. chips. Existing providers have 90 days to obtain one.
  2. Nvidia has already assumed zero direct China data-center compute revenue in its outlook. Even so, Q4 guidance is 12% above Q3 revenue. The new exposure is Chinese demand served through overseas facilities, which Kiwoom estimates at $16.3bn a year.
  3. With chips in short supply, other customers could fill any openings. Kiwoom puts unmet demand at 2.8–6.5 times the potential loss. On that assumption, the first casualty is not current revenue but the headroom to raise future guidance.
  4. Jefferies was reported to see a potential $44bn–$62bn hit to next year's revenue. Its methodology differs, so the two estimates cannot be compared directly. Whether replacement demand materializes is decisive.
  5. The burden may fall harder on data centers than on chips. Chip orders can be redirected. A data center built around Chinese tenants cannot be relocated so easily. For Nvidia's outlook, the pace of U.S. data-center completion may matter more than China itself.

Hover over dotted-underlined terms (tap on mobile) for definitions. Magenta underlines mark the key concepts in this report.

On Nov. 16, U.S. time, the Commerce Department's Bureau of Industry and Security (BIS) announced new export-control rules. The licensing requirement now reaches U.S. AI chips supplied to overseas data centers serving Chinese companies. Nvidia fell 4.1% that day and the Philadelphia Semiconductor Index lost 5.3%.1 But the selloff also followed CEO Jensen Huang's departure, CFO Colette Kress's appointment as interim CEO and a broader decline in AI stocks. The rule alone cannot explain the market move.2

Nvidia's Q3 results the following day beat expectations. We examined those results in Research No. 14. Anxiety about AI shares persisted: the Philadelphia Semiconductor Index fell 10.35% over the five sessions through Nov. 19,3 and South Korea's KOSPI plunged 5.21% to 4,869.96 on Nov. 20.4

Selling accelerated on Nov. 20, U.S. time. Nvidia was already down more than 6% when Reuters reported during the session that the Securities and Exchange Commission (SEC) was investigating its accounting. The stock ended 14.0% lower; the Philadelphia Semiconductor Index closed down 10.7%.5, 6 The scope of the investigation has not been disclosed, and an investigation does not establish wrongdoing.

That raises the central question: will restrictions on Chinese customers reduce Nvidia's sales? The answer also matters to Korean chipmakers supplying Nvidia with high-bandwidth memory (HBM), the fast memory mounted beside AI processors. Kiwoom's thesis is that the bigger hit is to headroom, not revenue. That headroom is demand waiting for additional chip supply—potential sales that could support future guidance upgrades. The scenario assumes licenses are denied and chips remain supply-constrained.

01What the Rule Covers

Not just chip sales:
access to computing power

Unable to buy many U.S. chips directly, Chinese companies have been renting computing capacity in overseas data centers through the internet—known as remote access. The U.S. think tank Institute for Progress (IFP) estimates that 23% of AI chips used by China could be accessed this way by year-end.7 The new rule targets the route by which chips enter those data centers, rather than blocking remote computing services directly.8

Two separate license triggers:
either one is sufficient

Where U.S. chips enter data centers

U.S. AI CHIPSNVIDIAOVERSEAS DCOutside the U.S.CHINA HQGPU USERFIVE DESIGNATED COUNTRIESID · MY · SG · TH · BRABA · Serving Chinese-HQ firmsabove 12mn TPPAnywhere abroad · 90-day windowB · Five countries, any compute scaleU.S.-HQ recipient/user exceptionEITHER triggerrequires a licenseExisting: direct China chip exports also need licenses

Sources: Federal Register; Morgan Lewis8, 9

The Rule at a Glance

Key provisions governing remote access to advanced semiconductors
ProvisionRequirement
A · Overseas data centers serving Chinese-headquartered firms above the threshold
Controlled transactionA license to send covered U.S. advanced chips to those data centersThe restriction acts on chip transfers, not the remote service itself
Customer coveredEntities headquartered, or ultimately parented, in countries of concern such as China and MacauForeign subsidiaries are covered when the ultimate parent is Chinese-headquartered
Compute thresholdMore than 12 million TPP of aggregate processing capacityRoughly 760 H100s · Connected facilities are aggregated · Applies anywhere outside the U.S.
Existing providersMust obtain a license within 90 daysFiling an application is not enough; absent a license, subsequent chip exports face restrictions
License reviewCase-by-case reviewApproval rates remain unknown
ExceptionComputing used solely to recommend products or contentGenerative AI systems that create new content do not qualify
B · Chip shipments to five countries
CountriesIndonesia, Malaysia, Singapore, Thailand and BrazilLicense required regardless of the compute provided
ExceptionThe chip recipient or end user (or its ultimate parent) is U.S.-headquartered and complies with monthly reportingHaving a U.S. company among the data center's customers does not itself qualify
Provisions common to both
TimingEffective Nov. 19, 2026 · Comments due Jan. 19, 2027An interim final rule takes effect before comments close · Eligible shipments already in transit have until Dec. 21, 2026

Sources: Federal Register; Morgan Lewis; Greenberg Traurig8, 9, 10 · For detailed conditions, expand “More on the Rule” below

How Large Is 12 Million TPP?

02Demand at Risk

Direct China sales: already zero.
New exposure: China renting abroad

There are two distinct China exposures. Nvidia's forecast already assumes zero revenue from direct China data-center chip sales. What the new rule adds is Chinese demand for AI compute rented at overseas facilities. Mixing the two would double-count the regulatory impact.

Already in guidance
Direct sales to China
$2.4bn → $0USD

Nvidia sold $2.4bn of licensed, older-generation Hopper chips in fiscal Q3 (Aug.–Oct.). Shipments are now suspended, and those sales are excluded from Q4 guidance.11, 12

New rule
Chinese demand served offshore
$16.3bn/yrUSD

Annual demand for Nvidia equipment that could be lost if licenses are denied. The calculation follows.

Nvidia Raised Guidance Despite Excluding China

Nvidia results and outlook
MetricValueWhat it means
Q4 revenue guidance (Nov.–Jan.)$130bn ±3%Assumes zero data-center compute revenue from China13Nvidia did not specify whether the China-related share of overseas data centers was also excluded
Q3 revenue (Aug.–Oct.)$115.8bnQ4 guidance is 12% above Q3 revenueQ4 has 14 weeks, one more than Q3; the per-week increase is 4%
Next fiscal year's (FY28) revenue growth outlook~70% → ~75%Raised in November from the initial estimate issued in August12

How Kiwoom Gets to $16.3bn a Year

Why $16.3bn Could Be Too High or Too Low

Factors that could raise or lower the $16.3bn estimate
DirectionBasisImplication
Could be higherBroader scenario: annualizing fiscal Q3 revenue from customers directly billed by Nvidia and headquartered in China or Hong Kong ($11.05bn), after subtracting mainland China sales (Kiwoom)1Geographic revenue follows the headquarters of direct customers, which may differ from the actual exposure under this rule15$34.6bn/yr
Could be higherCountry-list exposure (B): The $16.3bn estimate covers Chinese customers renting compute (A). In the five named countries, chip imports by data centers unrelated to China can also require licenses.Delays in licensing could widen the impactAmount unknown
Could be lowerDayOne's largest customer: The Information cited sources saying the customer's work did not involve AI training or inference, and that its share of 2027 contracted capacity would fall below 50%.16Not confirmed by the companyThe share actually covered by trigger A could be smaller
03Who Fills the Gap

Potential loss: $16.3bn.
Unmet demand: 2.8–6.5× as large

When chips are plentiful, losing one customer means losing its revenue. With supply constrained, another customer may take the capacity that opens up, according to Kiwoom's model.1

Evidence That Chips Remain Scarce

Two pieces of evidence underpin the argument. First, Nvidia's FY28 outlook appears calibrated to how much it can produce. It has committed to buy $203bn of components and production capacity for FY28 through purchase commitments. If cost of revenue is 27.5% of sales, that spending equates to around $738.2bn of potential revenue—almost identical to the $741.4bn implied by a 75% growth forecast.1, 11

Kiwoom's revenue conversion from supply commitments
MeasureCalculationResult
Company outlook, converted to dollarsEstimated FY27 sales of $423.7bn × 1.75 (75% growth)$741.4bn
Implied by purchase commitments$203bn ÷ 27.5% cost-of-revenue ratio$738.2bn
DifferenceGap between the two estimates0.4%

Sources: Nvidia; Kiwoom Securities calculations11 · 27.5% cost ratio = 100% minus the midpoint of next year's projected gross margin (72%–73%). Treating commitments as costs is a simplifying assumption, not proof of a hard revenue ceiling.

Second, signs of constrained supply appear across the industry.

Other indicators of tight AI-chip supply
IndicatorEvidenceSource
Company commentaryDemand substantially exceeds supply; Nvidia still expects strong results despite new export controls (Nov. 17 earnings call)Nvidia12
Rental pricesThe cost of renting Blackwell B300 chips has risen more than 60% since launchKiwoom Securities1
Contracted backlogAt major cloud providers (hyperscalers), contracted backlog rose 142% year over year in Q3IBK Securities17
Manufacturing bottleneckTSMC's advanced packaging process (CoWoS) remains constrained despite capacity expansionYuanta Securities18

Demand waiting for chips
exceeds demand at risk

Annualized · Dashed segments show the range · All figures are Kiwoom estimates

LICENSE DENIEDDEMAND AT RISK16.334.634.6Broader caseCHIP WAITLISTUNMET DEMAND46.2105.9105.9LowHighEven $46.2bn exceeds the $34.6bn broader risk caseUSD billions per year · Kiwoom estimates

Source: Kiwoom Securities calculations1 · See “Methodology” below for the formulas

What Shrinks Is the Upside

If another buyer takes the freed slot, Nvidia's revenue holds—but the waiting list gets shorter. That unmet demand is the potential for further guidance upgrades when supply expands. Restrictions eat into that reserve.1

Put in growth terms, current guidance calls for 75% sales growth next year. If all unmet demand converted into revenue, the rate could theoretically reach 86%–100%. After deducting the modeled exposure, that range falls to 82%–96%. The chart measures the percentage-point upside above 75%.

Additional growth headroom:
+11–25pp → +7–21pp

Unmet demand ÷ estimated FY27 sales of $423.7bn · Bar endpoints show low and high cases

0102030%pBEFORE RULE10.925.0AFTER RULE7.121.2−3.8%p−3.8%p

Source: Kiwoom Securities calculations1

For the stock, that distinction matters. Nvidia's valuation has benefited from earnings beats followed by higher guidance. Even if near-term sales hold steady, a smaller reserve of potential upside could unsettle the expectations priced into the shares.

COUNTERPOINT

Jefferies: Revenue Could Fall

Jefferies was reported to estimate that the remote-access rule could reduce Nvidia's data-center revenue by $44bn–$62bn next year.19, 20 We could not obtain the original research note and therefore could not verify its assumptions. Some reporting suggests the figure includes both direct restrictions and knock-on effects.21

Comparing the Kiwoom and Jefferies estimates
QuestionKiwoom SecuritiesJefferies
What is measured?Potential demandPotential revenue loss
Scale$16.3bn/yr ($34.6bn broader case)$44bn–$62bn next year
CoverageChinese customers exposed to remote-access controlsEffects of the new rule (as reported)Components unverified
Replacement customersAssumes others take the freed capacityNot established by available reporting

The pivotal question separating the estimates is whether other buyers can actually absorb the displaced orders. Yet the scope of each estimate—including possible licensing delays in the five countries—and the way revenue is counted may also differ. Replacement demand is easier to find while chips are scarce; once supply catches up, an empty slot can translate directly into lost sales.

04Where Risk Lands

Chips can find a new buyer.
Buildings stay put.

Chip allocations can be redirected to customers with sufficient power and cooling. A data center developed for a particular tenant cannot readily change location or contract terms. That is why overseas facilities reliant on Chinese customers could feel the pressure first.1

On the day of the rule,
data-center exposure took a heavier hit

U.S. close, Nov. 16 · The day before Nvidia's earnings release

NASDAQ 100U.S. tech stocks−2.7%NVIDIAAI CHIPS−4.1%PHLX SEMICONDUCTORCHIP INDEX−5.3%EQUINIXU.S. colocation−1.0%GDSDAYONE INVESTOR−14.7%

Sources: Bloomberg; Kiwoom Securities1 · AI stocks sold off broadly that day, so the performance gap cannot be attributed entirely to the rule

Contracts Are Also at Risk in Southeast Asia

Malaysia is one of the five countries subject to trigger B. Data centers there need a license to receive covered chips regardless of compute size, unless they qualify for the U.S.-headquartered exception described in Section 01.8 There is already evidence of contract disruption: reports in early November said Oracle had canceled a roughly $7bn, five-year agreement to rent up to 100,000 AI chips in Southeast Asian data centers to China's Tencent after U.S. government concerns emerged.24 That preceded the new rule, so the cancellation cannot be ascribed to it.

05What Could Break

The outlook turns when
data centers cannot take delivery

Kiwoom argues the regulation alone is unlikely to force a cut in Nvidia's revenue forecast.1 Instead, it identifies three conditions to watch. Nvidia has also said its next-year outlook takes into account data-center readiness, end-customer demand and customer financing.12

Conditions that could undermine Nvidia's revenue outlook
ConditionWarning thresholdCurrent reading
① FacilitiesCompletion delaysMore than another 0.9GW of delayed data-center completions among waiting customers within one year (base case)Without operating facilities, customers cannot take delivery even when they want the chipsTexas grid-connection queue averaged across one year: roughly 1.3GWAn illustration of potential exposure—not measured delays · See chart below
② DemandNext earningsAt the late-February results, FY28 growth guidance drops below 75%, or the company does not reiterate that customer plans imply demand for twice this year's revenueGuidance was raised in November (Section 02)The “twice this year” statement was not repeated
③ FundingCustomer financesIf days sales outstanding (DSO) and customer-related guarantees both increase again next quarterDSO: 60 → 64 daysMaximum guarantees: $108.5bn → $118bn11

Sources: Kiwoom Securities; Nvidia CFO commentary1 · Kiwoom also flags the risk that Chinese export curbs on critical materials could reduce chip production itself, a supply shock that waiting buyers cannot offset

If all Texas projects in the connection queue slip,
can waiting demand absorb the delay?

Gigawatts · Worst-case stress test, not an observed delay

0123GWADDITIONAL DELAYTHAT CAN BE ABSORBED0.92.82.8TEXAS GRID QUEUEANNUALIZED~1.3~1.31.3GW delay exceeds 0.9GW cushion by 0.4GW (up to $13bn)

Source: Kiwoom Securities, based on ERCOT data1, 25 · Assumes an entire year's queue is delayed, all capacity represents Nvidia-equipment demand and none is already reflected in guidance · $13bn equals 1.8% of the FY28 revenue outlook

Look at the Structure of Customer Financing (Condition ③)

Implications for Korea's Chip Industry

If Kiwoom's supply-shortage assumption holds, the Korean semiconductor exposure needs to be split in two.

INDIRECT EXPOSURE
Memory · HBM
U.S. build-out pace

As long as chips are scarce, Nvidia can shift capacity from Chinese customers to other buyers. For HBM suppliers, the more immediate variable is whether U.S. data centers are completed on schedule—not the China rule alone.

DIRECT EXPOSURE
Foundry · Packaging
Check China end users

The rule also tightens due diligence for foundries and packaging firms making advanced chips for Chinese customers.9 Taiwanese industry publications say TSMC, Samsung Electronics and others could defer orders from Chinese fabless chip designers until U.S. licenses are granted.27

We covered when new memory fabs will begin production in NO.34, and Korea's AI-export exposure in NO.36.

Dates to Watch

  1. DEC. 10 · EARLY KST
    Korea time
    FOMC decisionMeeting Dec. 8–9, U.S. time
  2. DEC. 10 · U.S.
    ERCOT audit report (target)Next step for 17 large-load projects awaiting grid connection
  3. JAN. 19, 2027 · U.S.
    Rule comment deadline · New Texas administration takes officeThe incoming administration has pledged a pause in data-center approvals
  4. AROUND FEB. 17, 2027 · U.S.
    90-day license deadlineCounting from the Nov. 19 effective date
  5. LATE FEB. 2027
    Nvidia fiscal Q4 earnings (expected)Next opportunity to update FY28 guidance

Sources: Federal Register; Utility Dive; FOX 7 Austin8, 25, 26 · Dates may change · Daol Investment & Securities suggests waiting for the early-December Texas audit update and FOMC decision before becoming more constructive28

CONCLUSIONConclusion

Watch the buildings,
not just China

The new rule imposes real constraints. But if chips remain scarce and other customers' data centers start on schedule, Kiwoom's model suggests the immediate cost falls more on Nvidia's room to raise guidance and on overseas facilities built around Chinese tenants than on Nvidia's sales. The critical question is not simply whether Chinese demand disappears, but whether replacement buyers are ready to take delivery. Three indicators matter:

  1. Are licenses issued within 90 days (around Feb. 17)? Watch whether existing overseas providers receive approvals and whether more Chinese customer contracts are canceled.
  2. Does the 75% growth outlook hold at the late-February earnings release? Check whether FY28 revenue guidance and purchase commitments remain intact or rise.
  3. Will data centers be ready to receive the chips? Watch the Dec. 10 Texas audit report for movement on grid connections and signs of slippage in U.S. construction schedules.

One separate risk has emerged: the SEC accounting investigation reported on Nov. 20. This analysis rests on Nvidia's disclosed numbers; if the scope of the inquiry becomes known, those figures will need fresh scrutiny.

This report is not a recommendation to buy or sell securities. Reference dates and calculation details appear in the expandable sections below.

Related ReadingResearch No. 14: Why Are Korean Stocks So Volatile?No. 27: A Blue Wave Across the Data-Center MapNo. 34: New Memory Fabs Open in 2027, Capacity Clusters in 2028No. 36: Korea Is Less Exposed to AI Than Taiwan—but More to Memory PricesResearch No. 12: Mapping Circular Financing in AI
Glossary
Remote access
Using computing power from chips housed in another data center over the internet, without buying the chips. Training an AI model in the cloud is a common example.
TPP
Total Processing Performance (TPP) is a metric used in U.S. export controls to assess computing capability. Assigned chip-level values are aggregated across devices. One H100 has about 15,800 TPP.
Unmet demand
Potential demand for chips that customers want but cannot yet obtain because of supply limits. It is not a firm order backlog; Kiwoom infers it from the gap between Nvidia's description of customer plans and its revenue outlook.
Purchase commitments
Amounts Nvidia has contractually committed to spend on components and manufacturing capacity. They indicate the scale of future supply obligations. At the end of fiscal Q3, commitments totaled $390bn, including $203bn for FY28.
Interim final rule
A rule that takes effect before the agency has finished receiving public comments. It can be revised afterward, but is enforceable in the interim.
Federal Register
The U.S. government's official journal for publishing regulations. This rule took effect on its publication date.
Countries of concern
Countries in U.S. Country Group D:5, associated with arms embargoes and related restrictions. China is included; the rule also specifically references Macau.
FY28
Nvidia's FY28 runs from February 2027 through January 2028; this report calls it “next year.” FY27 runs from February 2026 through January 2027.
Gross margin
The share of sales remaining after cost of revenue. A 72.5% gross margin implies a 27.5% cost-of-revenue ratio.
Hyperscaler
A major cloud provider operating large-scale data centers.
Backlog
Contracted work not yet recognized as revenue, including outstanding cloud-service commitments.
CoWoS
TSMC's advanced packaging technology that integrates GPUs with high-bandwidth memory (HBM). It is a major bottleneck before finished AI chips can ship.
ERCOT
The Electric Reliability Council of Texas, the state's grid operator. It manages the process of connecting large electricity consumers to the grid.
Days sales outstanding (DSO)
The average number of days between a sale and payment collection. A rising figure can indicate slower customer payments.
Guarantees
Commitments to assume specified obligations if customers cannot meet them. Nvidia has guaranteed portions of some customers' data-center leases and financing.
GW
Gigawatt, a unit of power capacity used to size data centers. One GW is on the order of the output of a large nuclear reactor. Kiwoom assumes about $32.5bn of Nvidia equipment sales per GW of capacity.
Case-by-case review vs. presumption of denial
Case-by-case review means applications are assessed on their individual merits. A presumption of denial is materially stricter: licenses are generally refused absent compelling circumstances.
Hopper · Blackwell · Vera Rubin
Nvidia's successive AI-chip generations. Hopper (H100, H200) is older; Blackwell (B200, B300, GB300) is the current flagship family; Vera Rubin is the next generation.
H100⁠·GB300
The H100 is an earlier-generation Nvidia AI GPU. The GB300 belongs to the Blackwell generation and is delivered in 72-GPU rack-scale systems.
HBM
High-bandwidth memory. A high-speed memory technology mounted close to AI processors and supplied by companies including SK Hynix and Samsung Electronics.
Foundry · Packaging
Foundries fabricate chips designed by customers. Packaging integrates finished chips with memory and other components into a working product.
DayOne
The largest data-center operator in Johor, Malaysia. Headquartered in Singapore, it operates across Southeast Asia, Japan and Europe. DayOne was carved out of Chinese operator GDS's overseas business; GDS remains a shareholder.
TCEQ
The Texas Commission on Environmental Quality, responsible for environmental permits including certain data-center generation facilities. It is distinct from grid operator ERCOT.
More on the Rule
  • Mechanism: This is an end-use control requiring licenses for exports, reexports and transfers (in-country) of covered U.S. chips to qualifying data centers—not a direct ban on remote access. It applies where the party supplying the chips knows the data center provides large-scale computing to Chinese-headquartered entities.8
  • Data-center ownership checks: Parties must also evaluate whether persons in countries of concern directly or indirectly own or control a data center.
  • TPP calculation: Add the processing performance of computing devices across the chips, aggregated for a single data center or interconnected facilities.
  • Review policy: Remote-compute applications receive case-by-case review. Exports of advanced chips to countries of concern and entities headquartered there face a presumption of denial.
  • Recommendation-system exception: Systems designed and used solely to rank, filter or suggest products, services or content may be excluded from trigger A if required details on company identity, aggregate TPP and due diligence are submitted to BIS. Content-generating systems do not qualify. The rule's explanatory text also says such operators must obtain licenses after 90 days, creating tension with the operative provision. Trigger B remains separate.
  • Conditions for the five-country exception: The ultimate consignee, end user or its ultimate parent in the export transaction must be U.S.-headquartered and comply with reporting obligations. The relevant party is the entity actually receiving or using the chips, not merely a customer renting capacity. Reports are due monthly, by the first business day of the following month.
  • After the 90-day period: The text says operators that fail to obtain a license face restrictions on “subsequent exports.” It does not clarify whether continuing to serve customers using chips already installed is prohibited. Using chips imported without authorization after Nov. 19 for a prohibited purpose is treated as a transfer (in-country).8
  • U.S.-based data centers: As an export-control measure, this rule does not directly reach data centers inside the U.S. A Remote Access Security Act bill is pending in Congress to address that gap.29
  • Foundries and packaging companies: The rule also narrows the route by which foundries and packaging firms producing advanced logic chips can demonstrate that an item is not a controlled chip.9
  • Rescission: The prior AI Diffusion Rule—along with its country tiers, model-weights controls and licensing exceptions—has been formally rescinded.10
Methodology
  • Calculations follow Kiwoom Securities research, using Nvidia's financial filings, DayOne's Form F-1 and ERCOT data as inputs.1Displayed numbers are rounded, while calculations use unrounded figures. Recalculating from the displayed values can therefore produce small discrepancies (for example, 423.7 × 1.75).
  • Estimated FY27 revenue of $423.7bn = $293.7bn recognized February–October + $130bn Q4 guidance (November–January).
  • Company outlook of $741.4bn = 423.7 × 1.75; purchase-commitment equivalent of $738.2bn = FY28 commitment of $203bn ÷ 0.275. The 0.275 cost ratio equals one minus the midpoint of projected FY28 gross margin (72%–73%). Difference: about 0.4%.
  • Potential demand at risk of $16.3bn = 0.5GW × $32.5bn per GW; $32.5bn is the midpoint of an assumed $25bn for Blackwell and $40bn for Vera Rubin. $16.3bn ÷ $741.4bn = 2.2%.
  • 0.5GW: DayOne operational capacity rose from 454MW (Dec. 31, 2025) to 962MW (Sept. 20, 2026), up 508MW in 263 days. That annualizes to about 0.7GW; multiply by 69.2% revenue concentration at its largest customer. The model applies a revenue share to capacity growth and converts that capacity into Nvidia-equipment revenue.
  • Broader $34.6bn case = (fiscal Q3 revenue of $11.05bn from customers headquartered in China/Hong Kong minus $2.4bn of mainland sales) × 4. Geography reflects direct customers' headquarters, not necessarily exposure under the new rule.
  • Unmet demand upper bound of $105.9bn = 423.7 × 2 − 741.4; lower bound of $46.2bn = 393.8 × 2 − 741.4. The $393.8bn base combines February–July actual sales, the then-projected $108bn for August–October and an equal amount assumed for November–January.
  • Multiples: 46.2 ÷ 16.3 = 2.8; 105.9 ÷ 16.3 = 6.5. In the broader case, 46.2 ÷ 34.6 = 1.3 and 105.9 ÷ 34.6 = 3.1.
  • Additional potential growth: 46.2 ÷ 423.7 = 10.9 percentage points (pp); 105.9 ÷ 423.7 = 25.0pp. Deducting 16.3 ÷ 423.7 = 3.8pp gives 7.1–21.2pp.
  • Guidance headroom: 5pp × $423.7bn = about $21.2bn; $16.3bn ÷ $21.2bn = about 77%.
  • Q4 growth: $130bn ÷ $115.8bn = +12.3%. Per-week growth: ($130bn ÷ 14) ÷ ($115.8bn ÷ 13) = +4.2%.
  • Additional delays that could be absorbed: ($46.2bn − $16.3bn) to ($105.9bn − $16.3bn) = about $30bn–$90bn; ÷ $32.5bn per GW = 0.9–2.8GW. Broader case: ($46.2bn − $34.6bn) = $11.6bn ÷ $32.5bn ≈ 0.36GW.
  • Converted into growth: 75% + 10.9–25.0pp ≈ 86%–100%; after the modeled restriction, 75% + 7.1–21.2pp ≈ 82%–96%.
  • Texas: 6.6GW ÷ roughly 5 years ≈ 1.3GW per year; (1.3 − 0.9) × $32.5bn ≈ $13bn; ÷ $741.4bn = 1.8%.
  • 12 million TPP ÷ approximately 15,800 TPP per H100 ≈ 760 H100s.
Source Discrepancies
License review standardKiwoom describes the remote-access licensing standard as a “presumption of denial.” The Federal Register text calls for case-by-case review of remote-compute applications; the presumption of denial applies to advanced-chip export applications involving countries of concern and entities headquartered there. This report follows the regulatory text.
Jefferies estimateTechCrunch reported that Jefferies projected a $44bn–$62bn decline in next year's data-center revenue due to the remote-access rule; Daishin Securities also cited it as a rule-related estimate. Other coverage says the figure includes both directly blocked sales and secondary effects. We could not obtain the original Jefferies note and cannot verify its composition.
China/Hong Kong revenueNvidia's geographic revenue is based on the headquarters of direct customers. The $11.05bn figure therefore does not measure remote compute used overseas by Chinese companies, and may omit chips accessed through firms headquartered elsewhere. That is why $34.6bn is a “broader scenario,” not an upper limit.
Nov. 16 stock performanceNvidia fell 4.1% on the rule's announcement day. Reuters also cited management changes after Jensen Huang's resignation and broader AI-sector risk aversion. The performance gap versus GDS and Equinix cannot be explained by rule exposure alone.
DayOneDayOne's Form F-1 does not name its largest customer, referring instead to a technology company with a globally popular short-form video platform. The 69.2% figure is its share of first-half 2026 revenue (69.4% in 2025), not its share of expanding capacity. The 1,328MW under construction as of Sept. 20, 2026 covers the entire company, not Johor alone. DayOne provides facilities, power and cooling—not the chips themselves.
H100-equivalent calculationKiwoom estimates roughly 758 H100s; CNAS and Morgan Lewis estimate about 760. The difference reflects slightly different TPP assumptions for each H100. This report uses “roughly 760.”
Purchase commitmentsAt fiscal Q3-end, Nvidia's supply and capacity commitments totaled $390bn, up from $279bn in one quarter. Most of the increase concerned multi-year memory procurement. Kiwoom's $203bn is the FY28 portion, matching the annual breakdown in Nvidia's CFO commentary.
Growth guidanceA summary passage in Kiwoom's report reads as though 75% growth was forecast in August. Its main text and the earnings call point to approximately 70% in August and approximately 75% in November.
Texas approvalsKiwoom wrote that Texas environmental regulators approved generation equipment for self-powered data centers on Nov. 16. FOX 7 Austin reported that TCEQ said late on Nov. 17 it was resuming normal permitting. Either way, TCEQ environmental permits are separate from ERCOT grid-connection approval.
Effective date and comment periodThe Federal Register version (91 FR 74970) sets Nov. 19, 2026 as the effective date, Jan. 19, 2027 as the comment deadline and Dec. 21, 2026 as the transit-shipment grace-period deadline. Some law-firm commentary instead gives Jan. 18 for comments.
IFP 23%The 23% figure is IFP's year-end forecast for the remote-access share of AI-chip use in China, as cited by CNAS. It is not an observed share.
Sources (29)
  1. Kiwoom Securities, “U.S. AI Chip Rules: No Empty Seats at a Packed Restaurant” (U.S. Market Today), Nov. 20, 2026. [In Korean]
  2. Reuters, 「Wall Street Falls as Nvidia Jitters Cloud Sentiment Ahead of Results」, 2026.11.16. gvwire.com
  3. IBK Securities, “Market Comment & Midnight News,” Nov. 20, 2026. [In Korean]
  4. Seoul Shinmun, “KOSPI Slumps 5% to 4,869.96 as Samsung Falls 7% on ‘Black Friday’,” Nov. 20, 2026. seoul.co.kr
  5. Reuters (via Euronext), “Nvidia's Accounting Is a Focus of US SEC Investigation, Sources Say,” Nov. 20, 2026. live.euronext.com
  6. Investing.com, 「Wall Street Plunges Into Correction Territory After Worst Week Since April 2025」, 2026.11.20. investing.com
  7. CNAS, 「Unpacking the Remote Access Rule」, 2026.11.18. cnas.org
  8. U.S. Federal Register, “Addressing Remote Access to Advanced Semiconductors” (interim final rule, 91 FR 74970), Nov. 19, 2026. govinfo.gov
  9. Morgan Lewis, 「Commerce Targets Remote Access to Advanced Semiconductors in New Interim Final Rule」, 2026.11.18. morganlewis.com
  10. Greenberg Traurig, 「BIS Interim Final Rule Targets Remote Access to Advanced Semiconductors, Rescinds AI Diffusion Rule」, 2026.11. gtlaw.com
  11. Nvidia, “CFO Commentary on Third Quarter Fiscal 2027 Results,” Nov. 17, 2026. q4cdn.com
  12. The Motley Fool (via Nasdaq), “NVIDIA (NVDA) Q3 2027 Earnings Call Transcript,” Nov. 17, 2026. nasdaq.com
  13. Nvidia, “NVIDIA Announces Financial Results for Third Quarter Fiscal 2027,” Nov. 17, 2026. investor.nvidia.com
  14. DayOne Data Centers, Form F-1 (U.S. IPO registration statement), Oct. 5, 2026. sec.gov
  15. Nvidia, Form 10-Q (quarter ended Oct. 26, 2025); explanation of geographic sales reporting by direct customer's headquarters, Nov. 2025. sec.gov
  16. The Information, “Singapore's DayOne Pushes Ahead With Data Center IPO Tied to TikTok,” Oct. 2026. Summary via Dealroom. dealroom.co
  17. IBK Securities, “2027 U.S. Equity Outlook: Approaching an Inflection Point,” Nov. 20, 2026. [In Korean]
  18. Yuanta Securities, “Beyond Memory, Taiwan: 2027 Taiwan Investment Strategy,” Nov. 20, 2026. [In Korean]
  19. TechCrunch, report on Nvidia's suspension of China H200 sales and the remote-access rule, Nov. 17, 2026. techcrunch.com
  20. Daishin Securities, “Contingency Plan,” Nov. 19, 2026. [In Korean]
  21. Railly News, 「Nvidia halts H200 chip sales to China, putting $62 billion in revenue at risk」, 2026.11.18. raillynews.com
  22. Bloomberg (via The Star), “Oracle Said to Advance Indonesia Cloud Services Plan,” July 14, 2025. thestar.com.my
  23. Bloomberg (via The Edge Malaysia), “Singapore-based DayOne Said to Pause US IPO Plans on Data Center, Market Concerns,” Nov. 20, 2026. theedgemalaysia.com
  24. Network World, 「Oracle cancels $7 billion Tencent AI chip deal」, 2026.11.10. networkworld.com
  25. Utility Dive, article on ERCOT's data-center audit, Aug. 21, 2026. utilitydive.com
  26. FOX 7 Austin, report on resumption of Texas data-center permitting, Nov. 18, 2026. fox7austin.com
  27. TrendForce, 「U.S. Tightens AI Chip Export Rules: What the Changes Mean for TSMC, ASE, and Taiwan’s AI Supply Chain」, 2026.11.18. trendforce.com
  28. Daol Investment & Securities, “Market Weekly Check ME Up!” (third week of November), Nov. 20, 2026. [In Korean]
  29. Ashurst Perkins Coie, 「BIS Issues Interim Final Rule Restricting Remote Access to Advanced US AI Chips」, 2026.11. ashurstperkinscoie.com