The Sovereign Infrastructure Cycle: A Physical Gatekeeper Methodology for the 2026–2030 Innovation Regime

1. Introduction

Cross-sectional equity strategies that condition primarily on price and accounting panels assume, implicitly, that the mapping from fundamental shocks to returns is stable and that the binding state variable is well proxied by such panels. When the marginal constraint on value creation is physical capacity, permitting, or qualified industrial throughput, that mapping weakens: identical earnings surprises can imply different repricing depending on whether a regulator clears a queue in the next quarter.

This paper specifies a rule-based research and portfolio policy designed for that environment. The contribution is not a new asset-pricing factor in the classical sense, but an explicit pipeline: a sequence of identification filters that reduce the investable set to a small basket of structural gatekeepers, combined with a dynamic policy that maps simulated decision-maker behavior to position sizes and to exit stopping times, and with a surveillance loop that continuously tests whether the stated thesis still holds against incoming public information. We adopt mathematical notation throughout so that assumptions, inputs, and outputs are unambiguous. Implementation-specific parameters beyond those disclosed here remain proprietary.

Outline. Section 2 motivates the economic shift toward physical scarcity as the relevant unit of analysis. Section 3 defines formal objects. Sections 4.1–4.3 specify the identification stages for cycles, critical stages, and nodes. Section 5 formalizes the simulation layer, entry policy, and continuous thesis surveillance. Section 6 gives the weighting map and normalization. Section 7 defines exits as stopping times, including structural invalidation. Section 8 summarizes operational architecture. Sections 9–10 cover limitations and conclusions. Appendix A lists symbols; Appendix B defines operational terms.

2. Economic motivation

For an extended period, scalable software and advertising technology exhibited high gross margins and low marginal cost of replication. Competitive dynamics were dominated by distribution, ecosystem lock-in, and incremental feature velocity. The 2026–2030 regime adds a complementary state variable: the schedule of physical projects required to deploy compute, automation, and onshore capacity at scale. Those projects consume grid headroom, thermal rejection capacity, long-cycle manufacturing assets, and interconnection services.

When the binding constraint migrates from code to capacity, the cross-sectional elasticity of enterprise value with respect to an additional unit of software output can fall relative to the elasticity with respect to clearing a specific physical gate. Empirically, one observes this in filing language, appropriations schedules, and private-market capital formation concentrated on infrastructure-adjacent subproblems. Section 3 makes the notion of a gate precise enough to operationalize.

2.1 Declining marginal information ratio for price-only models

Let informational advantage be summarized, heuristically, by the information ratio of a strategy conditional on an information set. When the relevant shocks arrive first in primary documents and engineering queues rather than in prices, conditioning on lagged returns and consensus fundamentals alone yields a lower signal-to-noise ratio for the policy objective (clearance of a specific stage) than conditioning on a structured extraction of those documents. Section 5 embeds this idea in a filtration-based formulation.

3. Formal definitions and notation

Time is discrete, indexed by t (e.g. trading days). All random objects are defined on a measurable space (Ω, ℱ). Public information available through t is denoted ℱ_t; we take ℱ_t to be generated by admissible primary documents and licensed data releases through t, excluding any non-public information. A node refers to a candidate listed equity; in public materials we use the placeholder Node X rather than live tickers.

Definition 1 (physical scarcity). A technology or industrial program exhibits physical scarcity at date t if producing one additional unit of useful output at commercially relevant quality requires, with probability one conditional on the current technology stack, a discrete physical transformation that cannot be completed solely by additional software expenditure over a short horizon (operationalized internally as two weeks).

Definition 2 (gate). A gate is a directed edge in the dependency graph of a cycle such that the subgraph downstream cannot attain its next milestone unless the edge’s physical or technical condition is satisfied.

Definition 3 (innovation cycle). An innovation cycle is a physically grounded buildout theme that satisfies the intersection criteria in Section 4.1 and the physical-substrate requirement stated there.

Definition 4 (critical stage). Given a cycle k, a stage s is critical if the stall predicate σ_k,s defined in Section 4.2 equals unity.

Definition 5 (live structural thesis for a node). Fix a position in node i ∈ 𝒩^∗ admitted at some t₀. Define the binary structural thesis indicator Υ_i,t ∈ {0, 1} equal to unity if and only if all of the following hold at t: (i) cycle k(i) to which i belongs still satisfies Q_k(i),t = 1; (ii) the critical stage s(i) for that cycle still satisfies σ_k(i),s(i),t = 1; (iii) node i still satisfies the conjunction of admissibility predicates C_i,t^∗ := ∏_r C_i,t^r = 1 for r ranging over the screening family in Section 4.3; and (iv) no public document in ℱ_t establishes a qualified substitute or capacity path that voids the replacement-cost or non-substitutability claims that justified admission. If Υ_i,t = 0, the economic story under which the position was opened is no longer supported by the same identification logic, independent of short-run price action.

3.1 Measurement units

Empirical work maps corporate language to dimensional quantities (power, heat rejection, yield, queue time). A stage is not admissible for formal testing until such a mapping exists; unspecified bottlenecks are treated as failing admissibility until clarified.

4. Identification pipeline

Identification is the composition of three filters. Let 𝒦 denote the set of candidate cycles, 𝒮_k the set of stages for cycle k, and 𝒩 the set of listed nodes. The pipeline produces 𝒦^∗ ⊆ 𝒦, then for each k ∈ 𝒦^∗ a set of critical stages 𝒮_k^∗ ⊆ 𝒮_k, then a feasible node set 𝒩^∗ ⊆ 𝒩 with |𝒩^∗| = 12 in the canonical mandate.

4.1 Innovation cycle identification via multisource intersection

For each candidate theme k, define binary indicators on publicly observed data through the review date:

• I_k^VC = 1 if venture metrics for k rank in the top cohort by quarter-over-quarter growth in Series B and C financing over the trailing eighteen months and growth exceeds 100% over that window.
• I_k^FED = 1 if k is a named beneficiary of at least one enacted appropriation or obligation exceeding USD 5×10⁹ signed within twenty-four months.
• I_k^10K = 1 if at least five of the fifty largest U.S. technology and industrial issuers by market capitalization name k as a primary strategic capital priority in the most recent annual filing cycle.
• I_k^PAT = 1 if k lies in a top patent-velocity cluster over twelve months (USPTO-based clustering as internally defined).

Let P_k = 1 if cycle k requires a physical or architectural change to U.S. infrastructure, compute delivery, or energy systems (purely digital execution without such change implies P_k = 0). The cycle qualification rule is

Qualified cycles are ranked by a composite score (appropriations magnitude and immediacy, private-market acceleration, concentration of large-capitalization mentions, patent momentum). The output set 𝒦^∗ contains the top one to three themes by that score, subject to minimum evidence thresholds. Requiring three of four channels is a deliberate variance-bias tradeoff: it reduces single-source narrative error at the cost of excluding late convergers.

4.2 Critical stages and the stall predicate

For cycle k and stage s, let H_k,s denote the twenty-four-month horizon from the evaluation date. Let A_k,s ∈ {0, 1} indicate whether failure to resolve s within H_k,s would prevent the cycle from reaching its next commercial milestone regardless of progress on other stages (downstream stages that remain blocked by an unresolved upstream constraint are excluded unless upstream is already at industrial scale). Let D_k,s ∈ {0, 1} indicate active capital concentration toward resolving s (internal measurement from calls, private markets, and filings). The stall predicate is

Stages with σ_k,s = 1 are critical. Multiple critical stages per cycle are retained when they pass the test independently. Conceptual example (no securities named): if campus-scale load cannot be energized without a committed high-voltage delivery path, the interconnection-to-energization stage may satisfy σ_k,s while downstream software milestones do not.

4.3 Gatekeeper nodes: structural necessity screening (version 1.3)

For each critical stage, supply-chain mapping yields candidate nodes i ∈ 𝒩. Define the following predicates, each equal to 1 if satisfied:

• C_i^mkt: at least 60% of the relevant U.S. market for the gated input depends on at most three suppliers, and i is among them.
• C_i^share: firm i supplies at least 40% of the specific input the stage requires.
• C_i^repl: estimated greenfield replication cost exceeds USD 1×10⁹ and calendar time exceeds twenty-four months.
• C_i^sub: no viable technical substitute can be qualified into the production stack at scale within twenty-four months.
• C_i^pure: either revenue from the gated component is at least 50% of total revenue, or the gated segment’s margin exceeds twice the corporate average.
• C_i^margin: trailing-four-quarter gross margin is stable or improving and exceeds 30%, or exceeds 20% for designated heavy industrial OEMs where the moated asset is physical process capacity.
• C_i^list: U.S.-listed common equity on major exchanges (no OTC ADRs in the baseline mandate).

Node i is admissible if the product of the above predicates is 1. The canonical basket fixes |𝒩^∗| = 12, allocating slots across cycles in proportion to evidence strength. Tie-breaking favors higher replacement cost and higher verified input share when two issuers gate the same input.

5. Behavioral simulation layer and entry policy

5.1 Information structure

Let {ℱ_t} be the filtration generated by admissible public documents and data through t. All portfolio actions in this section are required to be ℱ_t-measurable at decision time. Let 𝒜 denote the finite index set of modeled decision makers (central banks, commodity producers, trade authorities, utility regulators, and related actors with reproducible public records). The full cardinality of atomic signals monitored internally is not published; we denote the signal index set by 𝒥 as a large finite set for scale reference only.

5.2 Simulation mapping

For each signal j ∈ 𝒥, the engine implements a measurable mapping

where the output is a triple (c_j,t, d_j,t, τ_j): a confidence score c_j,t, a discrete direction d_j,t aligned or opposed to a stated thesis (or neutral), and τ_j a next resolution time in a structured calendar set 𝒯_j. The maps Φ_j are proprietary; readers may interpret c_j,t as a monotone transformation of an internally estimated probability of a thesis-friendly resolution conditional on ℱ_t, or as a calibrated index with the same ordinal ranking property. We do not claim that c_j,t equals a market-implied probability under ℙ.

For each node i ∈ 𝒩^∗, associate a primary signal j(i). Define thesis alignment indicator θ_i,t := 𝟙{d_j(i),t is aligned with the critical-stage thesis for node i}.

5.3 Entry as an ℱ_t-measurable rule

Fix thresholds c⁺ = 70 and c⁻ = 60 (percent units). Let π_i,t ∈ {0, 1} indicate that no long position in i is held at t. The entry trigger is

If E_i,t transitions from 0 to 1, an order is staged to execute within forty-eight hours of the crossing, subject to session and liquidity constraints. If before execution c_j(i),t < c⁻, the staged entry is suspended until c_j(i),t ≥ c⁺ again. Entries are not triggered by price momentum, sell-side consensus, or earnings surprises in isolation.

5.4 Identification versus estimation

Sections 4.1–4.3 are identification: they define admissible cycles, stages, and nodes by predicates. Section 5.2 is estimation: Φ_j produces scores from ℱ_t. The split mirrors standard econometric practice: the investable set is rule-governed; the timing overlay is model-governed and subject to misspecification risk documented in Section 9.

5.5 Continuous world monitoring and thesis maintenance

Entry is not the terminal state of the research process. The engine maintains a surveillance map on the same filtration {ℱ_t}:

The family of surveillance states (Ψ_i) over active nodes is updated whenever the surveillance job runs. Intuitively, Ψ_i = THESIS_BREAK when Υ_i,t = 0; REVIEW when new documents trigger a mandatory re-simulation or predicate recompute before the next scheduled cadence but do not yet falsify Υ; OK otherwise. The map is evaluated on a dual trigger schedule: (a) calendar refresh at fixed intervals aligned to filing seasons, earnings density, and appropriations calendars; and (b) event-driven refresh when monitored feeds register material deltas (enacted statutory text, agency orders, docket decisions, executive actions, major supplier or utility disclosures, and analogous primary sources). Each refresh re-ingests admissible data, re-runs the relevant components of Φ_j for signals tied to node i, and recomputes the predicates in Definition 5 from updated panels.

Structural versus behavioral monitoring. Behavioral monitoring tracks whether decision makers are still moving in a direction consistent with clearance of the critical stage (the objects c_j,t, d_j,t, τ_j in Section 5.2). Structural monitoring tracks whether the identification stack that admitted the name is still true: venture and policy capital may diverge, the binding bottleneck may migrate on calls, margins may violate floors, or a credible substitute may enter the qualified supply chain. Both channels read from ℱ_t; neither is redundant. Price levels are not inputs to Ψ_i except where they coincide with reported corporate actions (e.g. delisting) that affect admissibility.

Logging and reproducibility. Each evaluation of Ψ_i is appended to an audit log with timestamp, document identifiers that moved ℱ_t, simulation model version, and the resulting predicate vector underlying Υ_i,t. This supports ex post review of whether exits were triggered by documented structural breaks versus behavioral resolution, and calibrates how often the world state invalidates a thesis before markets converge on the same conclusion.

Portfolio implication. While Ψ_i = REVIEW, capital allocation rules may freeze new adds, flag accelerated human review, or tighten confidence bands internally; while Ψ_i = THESIS_BREAK, policy mandates exit per Section 7 (structural invalidation). The exact REVIEW playbook is operational detail; the existence of a mandatory re-evaluation path is not.

6. Portfolio construction: confidence-to-weight map and normalization

For each node i ∈ 𝒩^∗ with primary confidence c_i,t := c_j(i),t, define the raw weight (fraction of capital) before normalization:

Thus c_i,t = 70 implies w_i,t^raw = 0.05; c_i,t = 100 implies w_i,t^raw = 0.15; the map is affine on [70, 100] before the cap. Let S_t := ∑_i w_i,t^raw.

6.1 Normalization operator

Define the feasible simplex with per-name cap 0.15:

The normalization operator Π : ℝⁿ₊ → Δ_cap maps the vector of raw weights to the closest point in Δ_cap under the procedure: (i) if S_t ≤ 1, allocate residual 1 − S_t per mandate (cash or secondary rule); (ii) if S_t > 1, scale by λ = 1/S_t; (iii) if any cap binds, redistribute surplus proportionally among uncapped names; iterate until all caps and the budget constraint hold. Π is deterministic given inputs, ensuring reproducible lots.

6.2 Rebalancing

If |c_i,t − c_i,t−Δ| > 10 percentage points for a node, recompute w_i,t and apply Π within forty-eight hours, subject to market hours and liquidity.

6.3 Correlation and concentration

With n = 12, idiosyncratic risk is material. Partial correlation among nodes exposed to the same physical bottleneck is monitored ex post; the intersection design in Section 4.1 and slot allocation across cycles are partial ex ante mitigations.

7. Exit rules as stopping times

For each open position in node i, closing is permitted only on two classes of events: resolution of the bottleneck (success) or invalidation of the thesis (failure). Invalidation subsumes both behavioral and structural channels defined below.

7.1 Resolution

Let T_i^res be the first date at which the critical stage for i is cleared and a thirty-calendar-day observation window has completed, so that residual uncertainty is treated as ordinary operating risk rather than structural scarcity.

7.2 Invalidation: behavioral and structural

Let T_i^bhv be the first date at which the monitored behavioral signal resolves against the critical-stage mechanism or timeline (documented decision path inconsistent with the position narrative under ℱ_t). Let T_i^str be the first date at which Υ_i,t = 0 per Definition 5—equivalently, Ψ_i(ℱ_t) = THESIS_BREAK—including failure of Q_k(i), failure of σ_k(i),s(i), failure of any admissibility predicate C_i^r, or verified emergence of a substitute or capacity path that voids the original gatekeeper logic.

Define invalidation time T_i^inv := min{T_i^bhv, T_i^str}. The mandate retains exactly two top-level exit labels—resolution and invalidation—with structural breaks classified under invalidation for reporting purposes.

7.3 Exit stopping time and execution

The exit stopping time is

Liquidation occurs at the next eligible exchange open after τ_i^exit. Invalidation triggers (behavioral or structural) target completion of instructions within forty-eight hours of the triggering event. Discretionary holds through invalidation are excluded by policy: the exit set preserves the mapping from maintained thesis to P&L.

8. Implementation architecture

Mandates are implemented as separately managed accounts (SMAs) with beneficial ownership at the client level. Discretionary order placement uses broker application programming interfaces (execution currently routed via Alpaca), shortening the control loop between signal resolution and execution relative to many commingled fund workflows and reducing flow-induced distortion in a concentrated book.

8.1 Reporting and liquidity alignment

Position-level transparency supports allocator reconciliation between disclosed methodology and observed holdings. SMAs avoid certain pooled-vehicle frictions (subscription lines, redemption gates) that can interfere with maintaining twelve-name structural exposure through volatile policy windows.

9. Limitations and model risk

Predicate-based identification does not eliminate judgment in mapping text to indicators. Patent velocity can be noisy; appropriations can be reprogrammed; venture growth can reverse. The stall predicate σ_k,s is not a physical law. The maps Φ_j can be misspecified under regime shifts in institutional language. Normalization Π dilutes relative conviction when many c_i,t are simultaneously high.

9.1 Pre-specified update triggers

The framework implies observable updates enforced by Ψ: divergence of private capital from appropriations should demote a cycle (failing Q_k); migration of binding language on calls should move the critical stage (failing σ); margin paths below stated floors should remove nodes (failing C_i^margin); persistent Φ_j calibration error should widen intervals or defer entry. Structural breaks detected by Section 5.5 must map to T_i^str when predicates fail, not to discretionary overrides. These are discipline devices, not ex post rationalizations.

9.2 Comparator strategies

The mandate is distinct from generic price momentum, broad infrastructure indices, long-volatility hedges, and private credit: different information sets, capital structure layers, and payoff profiles.

9.3 Information ethics

The process uses public and commercially licensed data only. Ambiguous filings trigger conservative confidence downgrades rather than informal inference.

10. Conclusion

We have specified a sovereign infrastructure cycle methodology as a composable pipeline: multisource cycle qualification Q_k, critical stages via σ_k,s, admissible nodes via conjunction of screening predicates, behavioral scores via Φ_j, continuous thesis surveillance via Ψ on ℱ_t, affine-capped weights w_i,t^raw, normalization Π, and binary exit stopping times with structural invalidation explicit. Published structure and proprietary implementation are deliberately separated: the functional forms and thresholds material to external governance appear here; signal enumeration and internal estimation of Φ_j do not. Performance attribution and allocator materials are maintained under separate documentation.

Appendix A. Notation reference

Appendix B. Glossary

Innovation cycle. A theme k with Q_k = 1, expressed as a concrete physical buildout object.

Critical stage. A stage s with σ_k,s = 1 under Section 4.2.

Gatekeeper. An admissible node controlling qualified capacity for a gated input.

Node X. Public placeholder for a specific listed issuer; live basket not disclosed herein.

Resolution / invalidation. The two top-level exit classes in Section 7; invalidation includes behavioral (T_i^bhv) and structural (T_i^str) realization times.

Thesis surveillance. Continuous re-evaluation of Υ_i,t and Ψ_i on ℱ_t per Section 5.5.

Where performance is shown for a calendar year or year-to-date period, figures are actual returns for that sleeve and period from internal investor reporting materials—not hypothetical or illustrative scenarios. These materials are for informational discussion only and do not constitute an offering. An offering may be made only by delivery of a confidential offering memorandum to appropriate investors. PAST PERFORMANCE IS NO GUARANTEE OF FUTURE RESULTS.

Mathematical notation in this document is descriptive of internal policy and may omit implementation detail. Vektor Securities retains proprietary signal specifications and estimation procedures not enumerated here.