Loading current data…
DT-SYS // HYD-09 — AI SUPPLY CHAINSEC FILINGS · RESEARCH

Hydra // AI supply chain

From physical inputs to AI applications. Filing evidence, relative value and a research allocation.
The intelligence infrastructure / filings research

From electricity and silicon to the applications built on top. Explore who supplies what, what their filings reveal, and where strong businesses trade at lower multiples.

View the research portfolio ↗

Loading the dated filings snapshot…

What the evidence says

Screening the current filings universe…

01 / Follow the inputs upward

The AI supply chain

Click a role to explore its suppliers

Read from the bottom up: physical inputs feed chips and infrastructure, which support cloud platforms and applications. Connections show how categories depend on one another; they do not assert contracts between individual companies.

Supply role Contains a quality + value candidateCompanies can appear in more than one role. On narrower screens the map stacks into tiers, inputs at the bottom.
02 / Understand the business and the price

Supplier intelligence

A strong business is not automatically a cheap stock. Quality uses filing metrics; relative value compares positive earnings and free-cash-flow multiples within a comparable business peer group.

CompanySuppliesFindingQualitySales growthOp. marginFCF marginP/EP/FCFPeriod ended
03 / Know what is measured

Evidence before conviction

A

Business quality

Five equally weighted filing metrics: sales growth, operating margin, free-cash-flow margin, net debt/assets and share dilution. Missing inputs mean no score.

See the exact scoring rules

Each component runs from 0–100, clipped at the bounds. Growth: 0% → 30%. Operating margin: 0% → 30%. FCF margin: 0% → 20%. Net debt/assets: 60% → 0%. Dilution: 10% → 0%. Average all five.

“Quality” requires a score ≥65, positive growth and margins, net debt/assets ≤50%, dilution ≤5%, and no known interest shortfall or profit-without-cash flag. These are analyst heuristics, not return-tested signals.

B

Relative valuation

Compare price/earnings and price/free-cash-flow against at least five other profitable suppliers in the same business peer group. Both must be below the median, with at least one 15% cheaper.

Inputs, dates and limitations

Capitalization is a completed raw daily close × the last disclosed outstanding share count, adjusted for effective splits. Four consecutive quarters form trailing earnings and cash flow. FCF = operating cash flow − capital expenditure. Price inputs expire after seven calendar days; financial periods and share counts after 200 days. Valuation must match the current filing period and issuer.

Peer groups distinguish, for example, chipmakers from connector manufacturers and utilities from solar equipment. Each comparison excludes the company itself. Foreign/ADR and known multiple-class share bases are withheld for review. Capitalization is an estimate; subsequent issuance or repurchases can change it. P/E can be affected by one-offs and P/FCF by capital-spending cycles. Peer similarity is imperfect. No intrinsic price target or future return is inferred.

C

Coverage & confidence

The catalogue spans the chain and is joined against every company in the live filings universe. Industry-code discovery surfaces additional candidates for review; it does not establish AI exposure.

Inspect missing names and discovery limits

AI revenue shares and supplier/customer contracts are not inferred from total company sales. Figures in the supplied image are not used as financial inputs.

04 / Turn the findings into a research allocation

A portfolio built from the evidence

Balanced research allocation

Target holdings & position sizes

12% maximum per company25% maximum per supply role55% maximum per broad sector20% combined for highly correlated pairs18% annual volatility budget · 14% in market stress6–10 holdings · at least 4 primary roles
HoldingWeightSupply rolePrice multiplesWhy it earns a place

What to hold, add, reduce or sell

Company / actionPrevious targetNew targetChangeDecision evidence
Decision history
    How diversification, sizing and exits work

    Selection and conviction. Require complete current filing metrics, matched accounting checks, an active listing, positive earnings/cash-flow multiples and a complete adjusted-price history. Rank 60% business quality + 25% earnings/cash-flow yield score + 15% three-month momentum, minus 5 points per accounting flag (maximum 20). Yield scores run from 0 at 0% to 100 at 5%; momentum scores run from 0 at −20% to 100 at +20%. Existing holdings get a 5-point selection preference to reduce turnover. Prefer uncovered supply roles, then conviction relative to volatility and correlation. No peer discount is invented for small peer groups.

    Position sizes. Use 126 common trading sessions for volatility and correlations, adjusted for corporate actions. Start selected holdings at 2%, then allocate in 0.5 percentage-point increments according to conviction per unit of marginal covariance risk. Off-diagonal correlations are shrunk 25% toward zero for the risk estimate; the raw correlations govern the pair cap. Every increment must respect 12% per company, 25% per supply role, 55% per broad sector and 20% combined for pairs with correlation ≥0.75. The normal equity ceiling is 90% and estimated annual volatility ceiling 18%. If SPY is below its 200-session average and its three-month return is negative, these ceilings fall to 70% and 14%. Unused capacity stays cash. These are transparent research heuristics, not an optimized or validated trading strategy.

    Cut rules. A warning caps a holding at 6%; two warnings, or failing the entry-quality rules, cap it at 3%. Warnings include a 10-point quality decline, a 10 percentage-point slowdown in sales growth, a 5-point drop in operating or cash-flow margins, a verified comparable guidance cut of at least 5%, trading below the 200-session average, a drawdown of at least 25%, or both valuation multiples more than 25% above peer medians. Quarterly deterioration compares the prior quarterly reading. Guidance uses the latest available comparable extracted disclosure within 90 days and links its filing; coverage is partial.

    Exit rules. Propose a zero target for confirmed delisting, an interest-coverage or profit-without-cash failure, quality below 50, non-positive operating and cash-flow margins together, or net debt above 65% of assets. At a routine rebalance a holding may also leave when better candidates satisfy the risk and diversification rules. Missing or stale evidence triggers REVIEW and pauses additions; it does not, by itself, trigger a sell.

    When it updates. The daily filings workflow refreshes completed market prices, source filings and available guidance, checks every position and saves decisions. Full selection and sizing run at the first fresh publication each week, after a held company reports a new financial period, or after a rule-version change. Evidence-backed cuts and exits apply at every daily review. The browser displays the saved targets, not a second allocation from its reduced data view. It rechecks evidence expiry between publications and pauses changes if the saved review is more than two days old or required evidence expires. At least six holdings across four primary roles are needed to establish a new allocation. Model targets are compared with the previous published targets, not actual account weights; execution, drift, taxes and trading costs are not modeled.

    Information limits. The system uses SEC financial data, accounting flags, verified numeric guidance where available and market prices. It is not a comprehensive news monitor. Unknown customer contracts, unstructured news and absent guidance remain research gaps. Inspect filing dates and linked evidence before acting. Real-estate suppliers require a REIT-specific valuation model and remain excluded from new allocations.