Frameworks
The named analytical frameworks that structure the work on adikumar.co. Each framework has a canonical definition page covering what it decomposes, why it exists, how it maps, and where it has been applied. Follow the link on each card for the framework's full definition and the corpus of essays that apply it.
Grid-to-chip
The end-to-end AI power pathway from the utility interconnection to the accelerator die. Five physical layers (electrons, grid access, campus power, hall power, rack to gate) and a software envelope that monitors and controls them. Grid-to-chip is the operational name; the AI Power Chain is the formal six-layer decomposition that applies the same scope with vendor concentration and standards state per layer.
Definition page · five layers + software architectureThe AI Power Chain framework
The engineering stack that runs from grid interconnection through utility interface, medium-voltage distribution, protection and switchgear, rack-level power, cooling and commissioning to the accelerator die. Vendor concentration, standards state and lead-time exposure at each stage. Composed with the Investment Layer framework above and the operational Due Diligence framework alongside.
Framework page · applied across the corpusThe Investment Layer framework
Six pools of capital, five sponsor playbooks, four M&A through-lines, three downside cases. Maps how AI infrastructure gets financed, consolidated and stress-tested. Sits above the AI Power Chain framework as the capital-allocation view.
Framework page · applied across the corpusThe Due Diligence framework
The 14-workstream commercial and technical DD framework for AI infrastructure acquisitions, organized across four stages (scoping and setup, substantive analysis, structural and paper, post-close validation), with two companion reference assets (Question Banks and Seller-side workbook). Composes with the AI Power Chain and Investment Layer frameworks as the operational view that applies them to live transactions.
Framework page · applied across the corpus-
01Bottom-up TAM, growth vintage decomposition, executable-demand testing.
-
02Concentration decomposition, pricing power diagnostic, competitive moat audit.
-
03Product-vs-spec test, roadmap credibility by evidence tier, differentiation audit.
-
04Capacity vs demand, yield analysis, four-category IP audit, tier 1/tier 2 supply audit.
-
05Marginal unit economics, three downside scenarios, compound stress, refinancing risk.
-
06Execution-capacity testing beyond the org chart. Where post-close friction concentrates.
-
07Material contracts, change-of-control, non-compete enforceability, IP + FTO, litigation.
-
08Debt sizing, MAC clauses, refinancing wall, structural subordination.
-
09How DD findings translate into deal paper. R&W with insurance, escrow carve-outs, earnout mechanics.
-
10100-day validation testing thesis against operating reality. Six-month retrospective into the next transaction.
-
11QoE normalisation, working-capital bridge, EBITDA to cash conversion for industrial-tech targets.
-
12CFIUS, EU FDI, environmental compliance, cyber posture, export controls — the regulatory-approval workstream.
-
13DD staffing, phase-by-phase timeline, and the data-room structure that makes commercial DD executable.
-
SCompanion supplement: the self-diligence workbook sellers should have ready before buyers show up.
Live-deal calibration
These are not additional frameworks; they are case studies applying the three frameworks above to 2024-2026 transactions. Each includes stage-by-stage scoring, the 2×2 verdict visual, and the assessment basis.
-
AThe clean-deal calibration anchor. Physical ~22% of 162 (LOW-MEDIUM), thesis ~28% of 105 (LOW), veto GREEN.
-
BParent-guarantee absorbs an open tail that would fire the veto in a thin-equity SPV. Verdict: RESTRUCTURE with AMBER, not RED.
-
CFramework flagged single-tenant concentration pre-close; the failure mode materialised in Mar 2026 when Oracle + OpenAI dropped the expansion. Microsoft + Meta stepped in.
Investment Layer · applied essays
Essays applying the Investment Layer framework to specific capital-structure questions.
-
AHow hyperscalers moved from building their own data centres to leasing from developer platforms. 40 to 65-70 percent lease share, 2019 to 2026.
-
BWhy service revenue, not equipment, becomes the largest profit pool in AI-infrastructure industrials by 2030.
-
CWhy specialty vendors have less equity-research coverage than their strategic importance, and where mispricing concentrates.
-
DDemand-side shock, supply-side reversal, and the 2026-2028 refinancing wave. Compound scenarios and asset-class exposure.
Framework originals (archive): /ai-power-chain/ · /investment-layer/ · /due-diligence-for-the-ai-buildout/
The Grid-to-Chip framework
Follows the electron from the grid interconnect to the accelerator die across the full AI power chain, with per-stage vendor concentration, standards state, capex intensity, and the specific inflection years each stage is likely to pivot on. Series completed Aug 2026 across five stages: P1 · Electrons, P2 · Grid access, P3 · Campus power, P4 · Hall power, P5 · Rack to gate. Companion deep-dives extend the stages: Mount Diablo rack architecture (P4 companion), optical interconnects vs copper (P5 companion), and the battery role migration masterframe (R1 backup to R6 grid-forming, the connective frame across the P4.2 stack). Inside P4 sits the P4.2 investment stack with companions: P4.2b lithium UPS, P4.2c hall BESS, EU EED regulatory overlay, load-side flexibility and the role-migration masterframe. Composes with the AI Power Chain, Investment Layer, and Due Diligence frameworks as the cross-layer synthesis view.
Live-deal calibration · additions
Continues the case-studies list above. Each applies the frameworks to a specific 2026 disclosure or company.
-
DReference underwriting for the Interport multi-port SST architecture. Nvidia MGX qualification chain, single-point-of-failure, customer-adoption, and the falsification triggers that would break the thesis.
-
EThe three US SST pure-plays applied to the 800V DC + MV transformer supply gap. Head-to-head architectures, socket map, adoption scenarios, and the specific milestones that will decide the winner.