Type: WKAP Radar Feed

WKAP Radar Feed - 2026 - 10 - 14

Market date: 2026-10-14
Canonical URL: https://wkap.ai/radar/wkap-radar-feed-2026-10-14.html

WKAP Radar Feed

*October 14, 2026*

*Cheaper AI, Same Bottlenecks: Who Keeps the Savings?*

*3 Thesis Objects: MSFT, NVDA, ASML*

Preheader:

Make your AI track X alpha.

WKAP Feed compresses earnings inflections, KOL flow, supply-chain
bottlenecks, and short-duration attention into daily AI-ready context.

Set up your AI to read this Feed.

Ask:

“What did I miss?”

## HUMAN USER — START HERE

Investigate who keeps cheaper inference's savings, not which chip
wins. MSFT → ASML → NVDA is the research queue, not a return ranking:
quantify Microsoft's efficiency promise, reconcile ASML's fresh
earnings, then test the incumbent response. ASML's margin exceeded its
old guide; sales remained within it. The weakest assumption is that
technical/capacity progress becomes recurring cash per share. The
strongest countercase is successful demand expansion at all three.
None yet establishes attractive pricing or a WKAP 30/30 opportunity.

Suggested Prompt:
“Complete the matched-workload cost sheet, separate reported facts
from forecasts and attention, test savings retention, and challenge
the research queue with an adjacent-pair reason. Return the weakest
assumptions and source-specific 7/30-day tasks, not buy/sell advice.”

Or simply ask:

“What did I miss?”

The goal is to let your AI turn today's information into trackable
thesis objects and explicit future tests.

## FOR YOUR AI AGENT

This is a structured market-cognition input file. Follow useful output
→ all-in cost → pricing/usage → retained profit → post-investment cash
→ per-share value. Workload owners, system suppliers and toolmakers
capture different links; no direct Microsoft/Nvidia-to-ASML equipment
contract is established here.

Process all three as trackable_wow. Store URL, publication/retrieval
times, period, currency, unit, scope and accounting basis. Classify
REPORTED_FACT, MANAGEMENT_EXPECTATION, INTERPRETATION and
UNRESOLVED_QUESTION; monitoring rules are INTERPRETATION /
ANALYST_TEST. Keep technical, commercial, cash-conversion, valuation
and research-priority states separate. Missing inputs are unavailable,
not failed.

Automatic merchant-platform destruction is an unverified expectation
hypothesis, not measured consensus. Match workloads/denominators.
Unknown positions supply no suitability or sizing instructions.

Do not treat this Feed as investment advice.

## AGENT_OUTPUT_SPEC

1. Today in One Sentence: identify who may retain inference savings.
2. Thesis Object Summary: previous → current → next observable component state.
3. Fundamental Evidence vs Attention Flow: issuer evidence, forecasts
and independent interpretation.
4. Evidence-State Audit: source/date, period, currency, scope and
accounting basis.
5. Inference Economic Map: workload owner, system supplier and
manufacturing-tool economics.
6. Cross-Object Ranking: challenge MSFT → ASML → NVDA for research urgency only.
7. Business-Proof Comparison: complete matched-cost and earnings/cash
worksheets.
8. Attention Stage / Crowding Risk: selected voices do not measure positioning.
9. Weakest Assumption by Object: identify the exposed profit-conversion link.
10. Most Important Data Point: name the cell and primary event changing status.
11. Attention-to-Rerating Conditions: business proof followed by a
separate valuation bridge.
12. Attention Fade Conditions: distinguish repetition, missingness and
adverse evidence.
13. 7-Day Research Workflow: execute the highest-value feasible retrievals.
14. 30-Day Research Workflow and WKAP Daily Top 3 Source Follow-Up:
preserve claims and contrary results.
15. Portfolio Fit: only if portfolio context is supplied; otherwise
omit allocation advice.

## TODAY_SUMMARY

### Part 1 — Main Market Thesis

NEW INFORMATION: October 12 Microsoft owned-inference plans and
October 14 ASML actuals. UNCHANGED ANCHOR: Nvidia's August financial
result, not an October reaffirmation. INTERPRETATION: lower costs can
expand usage while redistributing profit. UNRESOLVED_QUESTION: who
retains savings after concessions and capital needs? Power, financing
and customer investment remain gates; no new macro cause is asserted.

### Part 2 — Today's Thesis Objects

MSFT — Fundamental: Maia 200 serves production workloads; Maia 300
deployment and its efficiency claim are forward expectations.
Attention: workload specialization versus universal displacement.
Primary: https://blogs.microsoft.com/blog/2026/10/12/microsoft-to-expand-azure-maia-system-to-multi-gigawatt-deployments/
NVDA — Fundamental: reported Q2 earnings and dated Q3 revenue/margin
ranges; no competitive financial miss is established by another
company's announcement. Attention: a limited inference-performance
preview. Primary:
https://investor.nvidia.com/news/press-release-details/2026/NVIDIA-Announces-Financial-Results-for-Second-Quarter-Fiscal-2027/default.aspx
ASML — Fundamental: Q3 sales within July guidance, margin above it,
stronger installed-base sales and more new systems sold. Attention:
lithography as a bottleneck across competing chip designs. Primary:
https://investor.asml.com/news-releases/news-release-details/asml-reports-eu11.7-billion-total-net-sales-and-eu5.0-billion-net-income-q3-2026
. Company-issued accessible release:
https://www.globenewswire.com/news-release/2026/10/14/3380319/0/en/asml-reports-11-7-billion-total-net-sales-and-5-0-billion-net-income-in-q3-2026.html

### Part 3 — Attention Flow Today

The voices connect specialization, incumbent response and fabrication
constraints: a qualitative debate map, not measured marketwide
rotation. Neither architectural estimates nor a preview proves profit.
Priority source: https://x.com/AryaTschand/status/2109747529059557533
. Ricky Ho discloses Long ASML; no holding statement was visible in
the other posts, which does not establish absence of interests.

### Part 4 — The Better Question

Instead of “does Microsoft's chip kill Nvidia?”, ask: for the same
useful task, what changes in the customer's total bill, the provider's
retained contribution and the manufacturing investment needed to
deliver it?

## MARKET_REGIME

RISK_TONE: Company-specific evidence scrutiny; broad positioning unmeasured.
MAIN_DRIVER: A prospective inference-cost change meets an actual
equipment-results checkpoint.
MARKET_CONTEXT: U.S.-listed price observations must retain their
currency, session and displayed trade time; no quote explains a move
by itself.
ATTENTION_ENVIRONMENT: Chip-war and bottleneck narratives move faster
than matched-workload cost or cash evidence.
WKAP_VIEW: Support demonstrated progress, preserve the unresolved
profit bridge, and assess the stock's purchase price separately.

## STATE_CHANGE_LEDGER

CANONICAL DECISION DASHBOARD — one append-only update record, not
competing rankings in each card.
Object / urgency | Previous → current | Missing cell / next
observation | Analyst update condition
MSFT / 1 | Maia 200 operating experience → October 12 Maia 300
scale/efficiency plan | Matched useful-output all-in cost and retained
savings; October 28 context, later actual deployment | Support
technical improvement only for demonstrated comparable workloads;
profit capture remains pending until the cost/pricing bridge is
sourced.
NVDA / 3 | August financial strength → October 14 competitive
benchmark preview | Reproducible system comparison and next financial
revenue/margin/cash bridge | A benchmark updates its tested
configuration, not earnings; apply original Q3 ranges to actuals when
reported.
ASML / 2 | July service strength and Q3 guide → actual sales WITHIN /
margin ABOVE | Service/system profit and timing-adjusted cash; future
accepted capacity | Support historical engines separately; future
demand/supply language does not establish delivery or recurring cash.

UPDATE PROTOCOL: Preserve original claim/vintage; record evidence,
affected/unchanged components, alternative explanation, missing cell
and next source/date. Nvidia: classify revenue BELOW/WITHIN/ABOVE
$105.84–110.16bn and each margin basis against 73.5–74.5%; mixed stays
mixed. Microsoft: October 28 group actuals cannot attribute future
Maia 300 profits. ASML: Q3 is observed, not imminent; Q4's new guide
does not replace July's old hurdle.

Urgency changes require an adjacent-pair reason: new evidence,
consequential unresolved task and displaced task. Resolution can lower
urgency while confidence rises. No automatic return upgrade or 30/30
admission.

## CROSS_OBJECT_EVIDENCE_MATRIX

Object | Evidence / classification | Comparison required | Separate
attention input
MSFT | REPORTED_FACT: existing Maia 200 production;
MANAGEMENT_EXPECTATION: Maia 300's 2027 deployment and 40% more
tokens/MW | Match workload/output quality, translate energy into
all-in cost, then retained contribution | Arya's architectural
interpretation
NVDA | REPORTED_FACT: FY27 Q2 revenue $96.221bn, GAAP margin 75.0%,
adjusted diluted EPS $2.22; MANAGEMENT_EXPECTATION: Q3 ranges |
Revenue and margin separately, then cash/mix and reproducible
system-cost comparison | SemiAnalysis preview, not universal result
ASML | REPORTED_FACT: Q3 €11.662bn sales, 57.5% margin, €3.289bn
installed-base business, 113 new systems sold | Original guide versus
actual; service/system mix; gains and cash timing separately | Ricky
Ho's bottleneck interpretation, with long position

SAVINGS-RETENTION SHEET — analytical calculation, not a measured
Microsoft result:
If matched useful output rises 40% at equal power, power per output
falls by 1−1/1.4 = 28.6%. Let E be baseline energy cost per useful
output in currency, ΔO the change in all other costs per output
including allocated development/deployment, and ΔP the revenue
concession per output. Conditional cost saving G=(2/7)×E−ΔO; retained
contribution change ΔM=G−ΔP. Positive ΔO means additional costs;
positive ΔP means a lower price. All amounts must use the same
unit/currency and comparison period. Neither 40% nor 28.6% is an
established all-in cost reduction. No energy share or actual
retained-savings amount is supplied here.
Model volume, fixed costs, investment and shares separately; no
invented usage forecast. Historical financial context:
https://www.microsoft.com/en-us/investor/events/fy-2026/earnings-fy-2026-q4

ASML CASH/PROFIT SHEET — Q3 statements in EUR, not nine-month totals:
€4.9604bn net income includes an investment gain; its
cash-reconciliation adjustment is −€0.5088bn. CFO €11.7976bn includes
€7.009bn changes in assets/liabilities. Reconcile reported components
and investment separately; no double gain adjustment or annualized
working-capital benefit. Genuine cash counters reflexive skepticism,
but timing does not prove recurrence. Statements:
https://ml-eu.globenewswire.com/Resource/Download/5bcb0f5c-1729-4046-baad-849ca7feb9a3

## RANKED_UNANSWERED_QUESTIONS

1. MSFT: who retains lower useful-output cost? Retrieve service
pricing, energy and other costs; October 28 supplies group context,
not Maia 300 proof.
2. ASML: what repeats beyond gains/timing, and what verifies delivery?
Use quarterly statements and issuer delivery/upgrade detail.
3. NVDA: does system value preserve paid demand/pricing? Retrieve
reproducible methods and actual financial conversion, not a universal
conclusion from one model.

## SCENARIO_BRANCHES

Confirm: matched evidence gives G>0 and ΔM>0. Support that workload's
contribution, not all Azure margins or valuation. Nvidia demand and
profitable ASML deliveries can coexist. Apply the canonical urgency
rule.
Mixed/customer capture: G>0 but ΔM≤0. Technical success remains;
retained profit is unconfirmed. Lower prices may expand usage, but
elasticity is unknown. Investigate volume/investment.
Break: comparable deployment fails G>0; persistent substitution
accompanies worsening Nvidia conversion; or ASML pushouts/weaker
commitments accompany deteriorating conversion. Weaken only the
evidenced component, not missing disclosure or an unmatched model.
Countercase: market expansion, incumbent improvement and productive
fabrication investment benefit all three—and may already be priced.
Require normalized earnings/cash, reinvestment, dilution and downside
before allocation.

## ATTENTION_TRADE_BOARD

WKAP Attention View: strongest fresh reported financial change—ASML;
cleanest analytical asymmetry—MSFT savings retention; largest
technical-to-commercial extrapolation risk—NVDA's preview. Most
crowded is unavailable without positioning evidence. Highest
selected-narrative fade risk is a universal NVDA benchmark claim. Best
durable-rerating candidate remains unresolved pending conversion and
price-implied expectations.

## RADAR_OBJECT_INDEX

THESIS_OBJECT_1: MSFT
THEME: Owned inference / savings retention / cloud investment
STATUS: active_trackable — existing production supported; future
retained profit pending
POSITION_CONTEXT: [not provided]
PRICE_AT_PUBLISH: USD 543.45; U.S.-listed MSFT; market-data latest
trade 2026-10-14 12:29:40 UTC (19:29:40 ICT), retrieved about 19:44
ICT. Before U.S. regular open; session classification is by clock, not
provider label.
DATE_FIRST_ADDED_TO_RADAR: 2026-08-25 — preserved historical core-card date
SETUP_TYPE: Technical efficiency→retained economics
ATTENTION_STAGE: Owned-silicon interpretation
ATTENTION_WINDOW: October 28 financial context; later actual 2027 deployment
KEY_QUESTION: What part of lower all-in cost remains after concessions
and investment?

THESIS_OBJECT_2: NVDA
THEME: Full-system inference / demand and pricing durability
STATUS: active_trackable — historical financial strength supported;
competitive extension pending
POSITION_CONTEXT: [not provided]
PRICE_AT_PUBLISH: USD 228.52; U.S.-listed NVDA; market-data latest
trade 2026-10-14 12:29:24 UTC (19:29:24 ICT), retrieved about 19:44
ICT. Before U.S. regular open; session classification is by clock, not
provider label.
DATE_FIRST_ADDED_TO_RADAR: 2026-08-25 — preserved historical core-card date
SETUP_TYPE: Incumbent response→paid economic proof
ATTENTION_STAGE: Limited benchmark preview
ATTENTION_WINDOW: Public benchmark methods and next actual FY27 Q3
result; reporting date not asserted
KEY_QUESTION: Does system value preserve paid demand and margins
despite selective substitution?

THESIS_OBJECT_3: ASML
THEME: Lithography / installed-base productivity / delivered capacity
STATUS: active_trackable — Q3 historical engines supported; future
conversion pending
POSITION_CONTEXT: [not provided]
PRICE_AT_PUBLISH: USD 1,717.76; U.S.-listed ASML; market-data latest
trade 2026-10-14 12:27:37 UTC (19:27:37 ICT), retrieved about 19:44
ICT. Before U.S. regular open; session classification is by clock, not
provider label.
DATE_FIRST_ADDED_TO_RADAR: 2026-10-13 — preserved prior issue's core-card date
SETUP_TYPE: Results and capacity-to-cash reconciliation
ATTENTION_STAGE: Reported earnings versus forward bottleneck framing
ATTENTION_WINDOW: Current Q3 follow-through; Q4 results and actual
delivery updates
KEY_QUESTION: Which earnings/cash benefits repeat as planned capacity
becomes accepted output?

## THESIS OBJECTS

First-added dates preserve prior durable Radar core cards, not
background mentions:
https://wkap.ai/radar/wkap-radar-feed-2026-08-25.html and
https://wkap.ai/radar/wkap-radar-feed-2026-10-13.html . A new theme
does not reset an object's introduction date. Prices are reference
observations, not executable offers or this edition's prospective
closing-price baselines. Retrieval time is later than the displayed
latest-trade time; no newer price is implied.

### THESIS_OBJECT_1 — MSFT

CARD_ID: WKAP-RADAR-2026-10-14-MSFT
CARD_TITLE: Find the Owner of the Savings
TYPE: trackable_wow
THEME: Workload-specific inference and retained unit contribution
STATUS: active_trackable — production supported; Maia 300 profit
conversion pending
POSITION_CONTEXT: [not provided]
PRICE_AT_PUBLISH: USD 543.45; U.S.-listed MSFT; market-data latest
trade 2026-10-14 12:29:40 UTC (19:29:40 ICT), retrieved about 19:44
ICT. Before U.S. regular open; session classification is by clock, not
provider label.
DATE_FIRST_ADDED_TO_RADAR: 2026-08-25
ATTENTION_STAGE: Architecture announcement→economic test
ATTENTION_WINDOW: October 28 context; actual 2027 deployment thereafter

WHY_A_READER_CARES_NOW: Efficiency headlines can change margin
assumptions before costs/pricing are measured.
PREVIOUS_STATE: Maia 200 operating experience and July group financial context.
CURRENT_STATE: Larger future inference role announced; Maia 200 actual
and Maia 300 future remain distinct.
DECISION_OR_MONITORING_PAYOFF: Fill the cost/pricing sheet before
attributing profits.

THESIS_SUMMARY
Workload co-design can reduce purchased compute expense, not eliminate
other costs. Shareholders need retained cash, not throughput alone.

WKAP_ANGLE
The surface-level frame: substitution automatically creates profit.
The alternative frame: service-specific costs, concessions and
investment control retention.
The key research question: what savings survive the commercial bridge?

CORE_THESIS
REPORTED_FACT: Maia 200 production. MANAGEMENT_EXPECTATION: Maia 300
deployment in 2027, 40% more tokens/MW than leading GPU systems—not
independently established all-in economics. July FY26 Q4 context: CFO
$55.4bn / cash PP&E $35.8bn / FCF $19.6bn; cloud margin 65%. Headline
capex $41bn includes leases. FY27 lease-classification changes require
a comparable bridge. No disclosed Maia margin.

ATTENTION_TRADE_FRAME
Attention Source: @AryaTschand —
https://x.com/AryaTschand/status/2109747529059557533
Why Today: October 12 workload/data-movement interpretation beside the
issuer plan.
Attention Stage: Architectural analysis, not earnings.
Attention vs Evidence
Hard evidence: production status and historical group context; future
efficiency is expectation.
Primary sources:
https://blogs.microsoft.com/blog/2026/10/12/microsoft-to-expand-azure-maia-system-to-multi-gigawatt-deployments/
; https://www.microsoft.com/en-us/investor/events/fy-2026/earnings-fy-2026-q4
Attention / interpretation: specialization can favor interactive
workloads; GPUs compete elsewhere. Estimated ceilings/demand shares
excluded.
Attention Path: Architecture→matched useful output→all-in
cost→pricing→retained contribution→cash.
Attention Asymmetry: visible chip story, unmeasured savings ownership.
Crowding Risk: unknown positioning; internal services≠all Azure compute.
What Could Sustain Attention: measured costs, deployment and retention.
What Could Make Attention Fade: headlines without commercial evidence.
Attention-to-Thesis Conversion: cost sheet, attribution gate, then
productive-investment/share bridge.

WEAKEST_ASSUMPTION: cost savings remain after development, deployment
and customer concessions.
MOST_IMPORTANT_DATA_POINT: sourced G and ΔM for the same useful task,
not token throughput alone.
NEXT_DATA_POINT: October 28 FY27 Q1 after U.S. close, then actual
deployment. Calendar:
https://news.microsoft.com/source/2026/10/07/microsoft-announces-quarterly-earnings-release-date-69
. Slippage/no comparable improvement weakens; persistent uneconomic
deployment contradicts the profit route.

### THESIS_OBJECT_2 — NVDA

CARD_ID: WKAP-RADAR-2026-10-14-NVDA
CARD_TITLE: Test the Whole System, Then Its Price
TYPE: trackable_wow
THEME: Merchant inference systems and earnings durability
STATUS: active_trackable — Q2 earnings supported; future competitive
economics pending
POSITION_CONTEXT: [not provided]
PRICE_AT_PUBLISH: USD 228.52; U.S.-listed NVDA; market-data latest
trade 2026-10-14 12:29:24 UTC (19:29:24 ICT), retrieved about 19:44
ICT. Before U.S. regular open; session classification is by clock, not
provider label.
DATE_FIRST_ADDED_TO_RADAR: 2026-08-25
ATTENTION_STAGE: Benchmark preview versus substitution narrative
ATTENTION_WINDOW: Reproducible disclosure and next financial result

WHY_A_READER_CARES_NOW: Selective substitution need not reduce total
purchases or earnings.
PREVIOUS_STATE: August result establishes demand and margin at a
historical date.
CURRENT_STATE: Debate/preview supplies questions, not a financial miss.
DECISION_OR_MONITORING_PAYOFF: Test system cost and issuer
revenue/margin before revising the premium.

THESIS_SUMMARY
Software, networking and deployment improvements can coexist with
rivals taking workloads. Market expansion and paid pricing need
commercial evidence.

WKAP_ANGLE
The surface-level frame: proprietary chips destroy Nvidia—or a ratio
proves superiority.
The alternative frame: scope, system costs and market growth permit
mixed outcomes.
The key research question: does reproducible customer value sustain
paid economics?

CORE_THESIS
REPORTED_FACT: July 26-ended FY27 Q2 revenue $96.221bn; both
gross-margin bases 75%; GAAP/non-GAAP diluted EPS $2.46/$2.22. FY27
non-GAAP includes stock compensation, excludes specified gains.
MANAGEMENT_EXPECTATION: Q3 $108bn±2%, both margin bases 74%±0.5 point.
Preserve canonical ranges.
INTERPRETATION: Article reference $228.52/($2.22×4)≈25.7 times an
annualized quarter, not consensus/target. Repeating earnings is an
assumption; normalized earnings, investment and downside remain
missing.

ATTENTION_TRADE_FRAME
Attention Source: @SemiAnalysis_ —
https://x.com/SemiAnalysis_/status/2110203976293466498
Why Today: October 14 Rubin preview supplies the response lead.
Attention Stage: Limited benchmark, not universal economics.
Attention vs Evidence
Hard evidence: financial result/guide; technical claims require methods.
Primary sources:
https://investor.nvidia.com/news/press-release-details/2026/NVIDIA-Announces-Financial-Results-for-Second-Quarter-Fiscal-2027/default.aspx
Attention / interpretation: specific Qwen-model preview; broader
submissions forthcoming:
https://x.com/SemiAnalysis_/status/2110203978369601646 . No ratio
adopted as independently reproduced universal/profit evidence.
Attention Path: Configuration→reproducible useful-output cost→customer
demand/pricing→margin and cash.
Attention Asymmetry: ratios travel faster than scope/cost constraints.
Crowding Risk: unknown positioning; preview generalization risk.
What Could Sustain Attention: transparent methods and paid economics.
What Could Make Attention Fade: ratios without scope or financial evidence.
Attention-to-Thesis Conversion: tested configuration first, issuer
financial/cash test separately.

WEAKEST_ASSUMPTION: full-system customer value preserves premium
pricing as alternatives improve.
MOST_IMPORTANT_DATA_POINT: actual revenue and matching-basis gross
margin versus the original Q3 ranges, with mix/cash explanation.
NEXT_DATA_POINT: methods and actual FY27 Q3; date unasserted. In-range
sales with sub-73.5% margin needs reconciliation. Persistent
substitution/worsening conversion weakens; benchmark success alone
cannot validate valuation.

### THESIS_OBJECT_3 — ASML

CARD_ID: WKAP-RADAR-2026-10-14-ASML
CARD_TITLE: Separate the Forge's Earnings From Its Future Queue
TYPE: trackable_wow
THEME: Lithography, productive upgrades and manufacturing constraints
STATUS: active_trackable — actual Q3 engines supported; forward
delivery/cash durability pending
POSITION_CONTEXT: [not provided]
PRICE_AT_PUBLISH: USD 1,717.76; U.S.-listed ASML; market-data latest
trade 2026-10-14 12:27:37 UTC (19:27:37 ICT), retrieved about 19:44
ICT. Before U.S. regular open; session classification is by clock, not
provider label.
DATE_FIRST_ADDED_TO_RADAR: 2026-10-13
ATTENTION_STAGE: Fresh results→bottleneck interpretation
ATTENTION_WINDOW: Q3 follow-through; Q4 actuals and disclosed delivery
milestones

WHY_A_READER_CARES_NOW: Competing designs still need fabrication;
supplier earnings have a different clock.
PREVIOUS_STATE: July Q3 sales €11–12bn / margin 55–57% awaited actuals.
CURRENT_STATE: Sales WITHIN/margin ABOVE; installed-base €3.289bn
versus €2.762bn; systems 113 versus 86. Future constraints are
expectations.
DECISION_OR_MONITORING_PAYOFF: Support actual engines; don't annualize
gains/timing or count interest as delivery.

THESIS_SUMMARY
Systems and upgrades monetize different stages. Designer competition
need not erase tool economics; customers still must accept productive
output.

WKAP_ANGLE
The surface-level frame: all metrics beat; the queue guarantees cash.
The alternative frame: guide classification, mix, gains and working
capital matter.
The key research question: what repeats as capacity becomes accepted business?

CORE_THESIS
REPORTED_FACT: Q3 €11.662bn/57.5%; productivity upgrades helped
margin. MANAGEMENT_EXPECTATION: 2027–2028 supply shortfall; 2029–2030
order interest—not delivery. Q4 €14.2–15.2bn/~57%; full-year €44–45bn.
Preserve old Q3 guide and the cash worksheet; no social bookings
figure.

ATTENTION_TRADE_FRAME
Attention Source: @rickyho_1989 —
https://x.com/rickyho_1989/status/2110252940640555439
Why Today: October 14 result meets design-agnostic bottleneck interpretation.
Attention Stage: Investor thesis, Long ASML, not issuer evidence.
Attention vs Evidence
Hard evidence: actual sales/mix/system count and quarterly cash statements.
Primary sources:
https://investor.asml.com/news-releases/news-release-details/asml-reports-eu11.7-billion-total-net-sales-and-eu5.0-billion-net-income-q3-2026
; https://www.globenewswire.com/news-release/2026/10/14/3380319/0/en/asml-reports-11-7-billion-total-net-sales-and-5-0-billion-net-income-in-q3-2026.html
; https://ml-eu.globenewswire.com/Resource/Download/5bcb0f5c-1729-4046-baad-849ca7feb9a3
Attention / interpretation: bottleneck across GPU/foundry/memory
designs; unsourced valuation/consensus comparisons excluded.
Attention Path: Customer investment→accepted systems/productivity
upgrades→profit→timing-adjusted cash.
Attention Asymmetry: broad bottleneck, specific measured engines.
Crowding Risk: unmeasured positioning; disclosed long interest.
What Could Sustain Attention: profitable deliveries/upgrades, repeatable cash.
What Could Make Attention Fade: interest without accepted business.
Attention-to-Thesis Conversion: historical engines and future capacity
separate; accepted output/cash required.

WEAKEST_ASSUMPTION: planned manufacturing capacity earns productive
returns and converts beyond advance/timing effects.
MOST_IMPORTANT_DATA_POINT: accepted system/upgrade economics alongside
reconciled quarterly cash components.
NEXT_DATA_POINT: issuer delivery detail and Q4 actuals; date
unasserted. Pushouts/weaker commitments plus deteriorating conversion
weaken expansion. Missing details remain pending.

## CROSS_OBJECT_ATTENTION_COMPARISON

Cleanest Attention Trade: MSFT's savings-ownership question, a
research setup rather than a recommended trade. Most Evidence-Backed
Attention Trade: ASML's fresh reported financial change. Most Crowded
Attention Trade: unavailable. Highest Fade Risk: a universal NVDA
benchmark extrapolation. Best Candidate to Become a Durable Thesis:
unresolved until economics and price-implied expectations are both
tested.

## 7_DAY_RESEARCH_WORKFLOW

Review by October 21, 2026 at existing checkpoints. Do not invent
quarterly results inside this horizon.

### MSFT — 7-Day Checks

- Populate matched-workload cells with source/date and actual versus
planned availability. Mark confidential or absent cost inputs
unavailable.
- Run G/ΔM only on common useful-output, currency and period bases.
Record technical success separately from savings retained.
- Prepare October 28 cloud-margin, cash PP&E, lease and utilization
context fields. No Maia 300 production attribution before its actual
deployment.

### NVDA — 7-Day Checks

- Retrieve the preview's public method, exact model, hardware,
latency/utilization and price boundary. Do not combine it with a
different company's power claim.
- Preserve original Q3 financial ranges and the two gross-margin
bases. Find actual new issuer releases through the live index, not
guessed URLs.
- Track customer-specific substitution and total paid demand
separately; absent customer data does not prove loss or immunity.

### ASML — 7-Day Checks

- Archive original July guide versus actual Q3 and separate
service/system evidence; avoid repeating today's event as future.
- Reconcile the quarterly cash components, equity gain and investment
on their own bases; request actual delivery/upgrade detail.
- Preserve Q4 forward guide. New capacity or interest updates change
only those forward cells unless accepted business is disclosed.

### Cross-Object — 7-Day Checks

Return original claim | source/date | period/unit/basis |
previous/current component states | strongest counterevidence |
missing cell | next retrieval/review | adjacent-pair urgency reason.
NO_NEW_DATA preserves supported facts and pending extensions; next
review seven days after actual review or an earlier confirmed future
event, using existing weekday wakes. Keep older failed, weakened and
unavailable cohorts; a new edition does not erase them.

Primary-source retrieval registry: MSFT:
https://www.microsoft.com/en-us/investor ; NVDA:
https://nvidianews.nvidia.com/ ; ASML:
https://investor.asml.com/quarterly-results . Follow actual dated
issuer links; financial anchors and comparison bases appear above.
Attention-source retrieval registry: MSFT:
https://x.com/AryaTschand/status/2109747529059557533 ; NVDA:
https://x.com/SemiAnalysis_/status/2110203976293466498 ; ASML:
https://x.com/rickyho_1989/status/2110252940640555439
Evidence-state snapshot: MSFT=existing production supported, future
retention pending; NVDA=historical earnings supported, competitive
durability pending; ASML=Q3 engines supported, future delivery/cash
repeatability pending; classification=trackable_wow;
rule=source-matched component updates, not expected returns;
source-rule=primary facts, KOL interpretation separately.

## 30_DAY_RESEARCH_WORKFLOW

Review by November 13, 2026. This is a research horizon, not an
invented issuer reporting date.

### MSFT — 30-Day Checks

Validate the October 28 group financial bridge and any sourced
matched-cost evidence. Upgrade the specific retained-profit route only
when comparable G and ΔM are positive with productive investment;
aggregate margin alone cannot attribute it to future Maia 300.
Downgrade rollout or cost disappointment on actual comparable
evidence. Invalidate persistent uneconomic deployment or failure to
retain net savings; keep technical production successes independently
supported.

### NVDA — 30-Day Checks

Validate reproducible configuration evidence and any real intervening
financial result. Upgrade demand/pricing durability with paid
commercial conversion, not a ratio headline. Downgrade actual
margin/demand deterioration after mix and timing reconciliation.
Invalidate the proposed premium-profit route on persistent competitive
substitution with worsening conversion; missing results or a loss on
one unmatched benchmark is insufficient.

### ASML — 30-Day Checks

Validate the Q3 engines, gain and cash-timing bridge and any actual
delivery updates. Upgrade expansion confidence on accepted productive
output and credible repeatable cash. Downgrade pushouts/weaker
commitments alongside deteriorating conversion. Invalidate the
proposed capacity-to-value extension on persistent unproductive
capacity or lost demand; distant interest cannot rescue an adverse
observed mechanism. A future Q4 result remains pending until reported.

### Cross-Object — 30-Day Checks

Return separate SUPPORTED / WEAKENED / INVALIDATED / PENDING /
UNAVAILABLE business components, preserving original conditions.
Require as-of prices, normalized earnings/cash, reinvestment,
dilution, base/downside scenarios and price-implied assumptions before
an allocation or mispricing conclusion. None is admitted to the
established 30/30 list: potential +30% within 30 calendar days
requires a prospective price hurdle/deadline, catalyst, downside and
touch/close/hold probabilities, not research urgency.

Prospective return protocol: preselected QQQ for this mixed
cloud/semiconductor cohort, not factor-neutral or company-matched. Use
the first U.S. regular close after actual Article publication;
intended October 14 if publication precedes that close. Endpoint
sessions October 21 and November 13. Retrieve baseline and endpoint
adjusted prices from the same provider and adjustment vintage,
preserving raw initial observations and corporate actions.
Return=endpoint adjusted close / baseline adjusted close −1; excess in
percentage points versus QQQ. Unavailable comparable data stays
unavailable. Never substitute premarket references, switch benchmarks
after outcomes or treat small-sample outperformance as proof of alpha.
Earlier cohorts retain their own benchmarks.

## WKAP DAILY TOP 3

Three market sources worth feeding into today’s market chat. Not
required reading — WKAP has already extracted the signal.

### 1. @AryaTschand — Specialization Changes the Cost Question

URL: https://x.com/AryaTschand/status/2109747529059557533
WKAP signal: October 12 architectural interpretation emphasizes
workload specialization and data movement; quantitative ceilings and
broad demand-share claims are not adopted as facts.
Why it matters today: The post makes Microsoft's company claim a
matched-workload investigation. Arya's inspected bio identifies a
Modal researcher with prior industry experience; this is an
independent interpretive voice, not issuer earnings proof. No personal
stock position was visible in the selected post.
Themes/tickers: Owned inference, dataflow, all-in cost and retained
savings / MSFT.
Question to ask: Which sourced useful-output, energy, other-cost and
customer-pricing inputs make G and ΔM positive for the same service,
and which inputs remain unavailable?

### 2. @SemiAnalysis_ — A Preview Needs Its Denominator

URL: https://x.com/SemiAnalysis_/status/2110203976293466498
WKAP signal: October 14 inference-performance preview supplies an
incumbent-response lead; its follow-up narrows the model scope and
says broader public submissions are forthcoming.
Why it matters today: Independent research can guide reproduction, but
a selected configuration cannot prove universal superiority, customer
profit or issuer earnings. Keep hardware, latency, utilization,
software and system-price boundaries explicit. No holding statement
was visible in the inspected text.
Themes/tickers: Full-system inference, benchmark scope, demand and
pricing / NVDA.
Question to ask: What reproducible matched-workload cost evidence
changes the customer-value comparison, and does separately reported
revenue/margin/cash show that value remaining with Nvidia?

### 3. @rickyho_1989 — The Bottleneck Survives the Chip Duel

URL: https://x.com/rickyho_1989/status/2110252940640555439
WKAP signal: October 14 investor interpretation connects lithography
demand across competing chip designs; the author explicitly discloses
Long ASML.
Why it matters today: The causal map is useful beside fresh results,
but unverified forward multiples/consensus estimates are excluded. His
CIO/PM background is provenance, not a substitute for the issuer's
statements or independent positioning data.
Themes/tickers: Lithography, productive upgrades, accepted capacity
and cash / ASML.
Question to ask: Which accepted-system/upgrade and cash-component
disclosures support recurring supplier value beyond today's
gain/timing benefits and distant order interest, and what would weaken
that extension?

WKAP Radar Feed - 2026 - 10 - 14

Agent-readable facts