Research Library

Selected publications · March – April 2026

White PaperMarch 2026 · ~5,000 words

The Intelligence Takeoff

Why AI infrastructure spend is not a bubble

A structural analysis of AI capital expenditure trajectories, arguing that current infrastructure spending reflects a rational response to demonstrated capability scaling, not speculative excess. Examines the compute-to-revenue conversion pipeline across hyperscalers, the emerging inference-vs-training cost inversion, and why traditional bubble metrics misread the current cycle.

NVIDIAAlphabetMicrosoftSK Hynix Inc.Celestial AI

Company AssessmentMarch 2026 · ~3,300 words

SK Hynix: Memory as AI Infrastructure

Company assessment — the HBM monopoly position

SK Hynix controls approximately 50% of the High Bandwidth Memory market, the single most constrained component in AI accelerator supply chains. This assessment evaluates the durability of that position through the HBM4 transition, ADR valuation implications, and the structural dependency every major GPU vendor has on Hynix's roadmap.

SK Hynix Inc.NVIDIABroadcom

Executive ExplainerMarch 2026 · ~1,200 words

Zero-Day Depreciation

How AI reshapes capital equipment economics

When the next generation of AI hardware offers 2–4× performance at similar cost, the economic life of installed infrastructure compresses dramatically. This memo examines how zero-day depreciation dynamics affect datacenter ROI calculations, create perverse incentives for delayed deployment, and open arbitrage windows for operators willing to optimize for inference throughput rather than training peaks.

Monolithic Power SystemsGroq

Research AnalysisMarch 2026 · ~2,300 words

Karpathy/BLS AI Job Exposure Analysis

Which occupations face the highest AI displacement risk

Cross-referencing Andrej Karpathy's task-level AI capability assessments with Bureau of Labor Statistics occupational data to produce a ranked exposure model. Identifies the occupations with the highest near-term displacement risk and the structural factors that make some roles more resistant than capability benchmarks suggest.

OpenAIDeepMind

Client DeliverableMarch 2026 · ~2,500 words

Hiawatha National Bank Peer Analysis

Community bank competitive positioning — v3 Final

A comprehensive peer analysis for a Midwestern community bank, benchmarking financial performance, efficiency ratios, and digital readiness against comparable institutions. Delivered as a presentation-ready document for executive review.

Industry MapMarch 2026 · ~5,700 words

The Intelligence Layer of U.S. Reindustrialization

AI, automation, and software companies that make reshored manufacturing viable

A company-by-company analysis of the AI and automation stack enabling U.S. reindustrialization, spanning industrial AI, robotics, digital twins, semiconductor equipment, additive manufacturing, IoT, and computer vision. Each company is reality-checked against deployment maturity, competitive position, and reindustrialization relevance. Includes policy tailwind analysis and tiered investment framework.

NVIDIAASML

Scenario AnalysisFebruary 2026 · ~4,200 words

AI 2027 — Investment Thesis Synthesis

Extracting investment signal from the AI 2027 scenario

A rigorous analysis of the AI 2027 scenario — the most detailed publicly available AI timeline projection — evaluating which predictions have been validated, where the authors themselves have revised their timelines, and what the investment implications are for compute infrastructure, energy, and the broader AI supply chain. Includes the strongest critiques and why the directional thesis holds even when specific dates don't.

NVIDIAAlphabetMicrosoft

Policy AnalysisApril 2026 · ~1,900 words

Nuclear Fuel DPA Synthesis

Defense Production Act mobilization of the nuclear fuel supply chain

The U.S. government invoked wartime mobilization authority for nuclear fuel — classifying domestic enrichment as essential to national defense and granting antitrust immunity to a 90+ company consortium. This analysis maps the policy chain from AI compute demand to uranium mines, identifies the investable entities at each stage of the fuel cycle, and explains why this is the strongest evidence yet that reindustrialization is being operationalized with real authority and real money.

FrameworkFebruary 2026 · ~2,100 words

Agent Economy Game Theory

The buyer, seller, attacker triad reshaping digital markets

Traditional markets model buyers and sellers. Agent economies introduce a third native actor: the adversarial agent, operating at the same marginal cost as legitimate participants. This framework examines how the attacker changes equilibrium economics for every interaction, why trust becomes the most valuable primitive, and what infrastructure is being built to mediate the triad — from Coinbase agent wallets to Cloudflare's dual role as gateway and gatekeeper.

Research AnalysisFebruary 2026 · ~1,900 words

Prediction Markets for Fund Strategy

Event contracts as institutional hedging instruments

An assessment of whether prediction markets — Polymarket, Kalshi, CME ForecastEx — are mature enough for institutional fund hedging. Examines settlement mechanics and counterparty risk in detail (including Polymarket's proven oracle manipulation vulnerability), compares platform architectures, and identifies specific use cases from policy cliff hedging to tail risk capture. The settlement analysis alone answers the question: will your bookie pick up the phone?

Executive NoteMarch 2026 · ~800 words

AP1000 Fleet Economics

Nuclear-for-AI baseload — the cost reality

Westinghouse's PwC-commissioned fleet study for 10 AP1000 reactors conspicuously omits CapEx and LCOE. This note fills those gaps with independent data from MIT, INL, and EIA — comparing first-of-a-kind Vogtle costs against fleet learning curve projections and explaining why standard LCOE comparisons mislead when the buyer needs 24/7 firm power at 99.9% availability.

Pipeline OutputApril 2026 · ~2,100 words

SSA Target Assessment: 64 High-Conviction Entities

Cross-vertical scoring and adversarial tribunal results

The Synthetic Strategy Apparatus scored 168 entities across 8 verticals — robotics, semiconductors, AI platforms, datacenter infrastructure, advanced manufacturing, defense, nuclear energy, and space nuclear — against custom rubrics. 64 entities reached High Conviction thresholds and underwent adversarial tribunal review. 47 received BUY verdicts. This document summarizes the top picks, key catalysts, and principal risks across the full portfolio.

Methodology Note

All research is produced using the Synthetic Strategy Apparatus (SSA), a multi-stage analytical pipeline combining structured entity scoring, adversarial tribunal review, and convergence mapping. The SSA operates across 8 industry verticals and maintains continuous monitoring of 168+ entities. Scoring rubrics, tribunal transcripts, and supporting data are available on request.