AI CapEx investment sentiment
Scoring Model
Tracked AI CapEx universe translated into investment actions for infrastructure exposure.
Transparent scoring engine
How AI filings become an investment action
Signal Engine starts every company at a neutral score of 50, extracts material AI CapEx signals from SEC filings, applies deterministic category weights, then translates the final score and trend into a plain-English action. The model is intentionally explainable: every point comes from a named signal, a fixed rule, and a visible source.
Ingest
Pull SEC filings from official EDGAR company filing feeds.
Classify
Map filing language into the fixed AI CapEx signal ontology.
Weight
Apply base category weight, materiality, and confidence adjustments.
Interpret
Convert final score, delta, and macro backdrop into an investor action.
Current AI Calculation
The exact model math for the current score: neutral base, weighted AI CapEx factor score, inactive external overlay, and final clamped score.Base score
50
Every company starts from neutral.
Company signal score
53
Unique company signal categories from filings are added once each.
External overlay
-10
No external feed impacts are active in the current source set.
Final score
43
Caution: Do not add yet.
Formula
finalScore = clamp(0, 100, 50 + weightedAiCapexFactorImpacts + externalOverlay)
factorImpact = factorWeight / 2 × signal polarity × directness × confidence. Factor impact is capped so no factor can exceed its allocation.
AI Action Bands
The final score is translated into an investor action band so the output is easier to use than a raw number.0-40
Reduce / hedge
Negative signals dominate. Preserve capital first.
41-55
Do not add yet
Mixed or fragile readout. Wait for confirmation.
56-70
Hold / selective add
Constructive, but monitor execution pressure.
71-100
Increase exposure
Strong AI CapEx opportunity when trend confirms.
CapEx Factor Weights
The factor allocation behind the score: 30% hyperscaler CapEx revisions, 20% HBM/DRAM, 15% advanced packaging, 15% data center/power, 10% AI revenue monetization, and 10% financial stress/free cash flow.Hyperscaler CapEx revision trend
AI CapEx factor
Largest allocation because hyperscaler CapEx revisions are the primary demand signal for the AI infrastructure cycle.
HBM / DRAM pricing and allocation
AI CapEx factor
Captures memory pricing, HBM availability, and allocation pressure that can either confirm demand or expose bottlenecks.
CoWoS / advanced packaging
AI CapEx factor
Tracks advanced packaging constraints because accelerator supply depends on CoWoS and similar capacity.
Data center / power
AI CapEx factor
Reflects data center deployment, power availability, and physical buildout constraints.
AI revenue monetization
AI CapEx factor
Measures whether AI infrastructure spend is translating into revenue, backlog, or durable demand.
Financial stress / free cash flow
AI CapEx factor
Acts as the risk control for free cash flow, liquidity, refinancing, and margin pressure.
Refinement weights
Second-order model adjustments that keep category extraction realistic: directness, confidence scaling, and inactive external overlays.Materiality multiplier
0.75x-1.20x
Raises or lowers a base category weight based on how directly it affects AI CapEx economics.
Confidence multiplier
0.88x-1.00x
Lower-confidence extracted signals still count, but with a smaller impact.
AI external overlay
Inactive
Generic macro and availability-only sources are excluded until direct AI CapEx leading indicators are connected.
Source-to-Signal Mapping
Shows which incoming sources map into which model categories. This is the ontology layer between raw data and score impact.SEC EDGAR
8-K, 10-Q, 10-K, and Form 4 filings produce company-specific demand, capacity, execution, financial stress, management, and insider signals.
Deterministic Keyword Extraction
Rule-based extraction maps SEC filing language into fixed AI CapEx categories before deterministic scoring.
FRED Transformer PPI
Transformer producer-price changes act as a numeric proxy for power equipment cost pressure affecting AI data center buildouts.
GDELT DOC 2.0
Recent news volume around AI data center power, HBM allocation, and CoWoS packaging adds a small early-warning pressure readout.
Excluded Source Rules
Documents which candidate sources are excluded because they do not provide direct, material AI CapEx early-warning evidence.Generic macro feeds
BLS CPI, CFTC COT, Federal Reserve Z.1, and Treasury TIC are excluded because they are broad context, not direct AI CapEx early-warning evidence.
Availability-only feeds
SEC Form ADV and FCA short-position pages are excluded because source reachability is not a meaningful investment signal.
Broad energy feeds
General electricity demand or CPI feeds remain excluded unless they can be tied to AI data center power constraints. Transformer PPI is included because it is directly connected to grid equipment cost pressure.
Macro rules influence the backdrop, not the core company thesis. That is why the overlay is capped at ten points in either direction.
Live inputs feeding this score
The actual company signals and macro observations currently feeding the displayed score.Company signals
Capacity Expansion
DLR: Capacity expansion supports AI infrastructure demand.
Fallback category mapping · Directness 1.15x · Factor impact +19 · Confidence 0.95x
Guidance Lowered
ORCL: Company language indicates lowered guidance or weaker outlook.
Fallback category mapping · Directness 1.20x · Factor impact -18 · Confidence 0.93x
Guidance Raised
MSFT: Company language indicates raised guidance or stronger outlook.
Fallback category mapping · Directness 1.15x · Factor impact +18 · Confidence 0.93x
Liquidity Stress
MSFT: Liquidity or cash pressure appears in the filing.
Fallback category mapping · Directness 1.10x · Factor impact -5 · Confidence 0.95x
AI external observations
CFTC COT
Leveraged futures positioning is risk-off across AI-adjacent macro markets.
BLS CPI
Inflation pressure is in the warning zone at 3.8% year over year.
GDELT DOC 2.0
CoWoS packaging news pressure is unavailable.
GDELT DOC 2.0
HBM allocation news pressure is unavailable.
GDELT DOC 2.0
AI data center power news pressure is unavailable.
Treasury TIC
Treasury TIC source is reachable
Federal Reserve Z.1
Federal Reserve Z.1 source is reachable
FCA SPR
FCA short-position disclosure source is reachable
SEC Form ADV
Form ADV data source is reachable
FRED
Transformer PPI moved +0.23% over three months.