How many documentary sources sustain Nvidia's lead in reported AI investments?
Citable version that has not completed AIRR screening. Not peer reviewed. Screening record
In a fixed public AI supply-chain dataset, Nvidia retains its lead in distinct documented investment counterparties after any omission of 37 hostnames. A set of 38 yields a tie with Sequoia Capital at 43 counterparties. A solver-free integer certificate checks the lower bound against every rival. When omissions act on groups of hosts connected by at least two matching dated headlines, the minimum becomes 35 groups. Requiring a single matching dated headline gives 32. Preserving capitalization instead yields an apparent change of leader after three individual omissions between two spellings of Nvidia. These results quantify the sensitivity of a documentary ranking under specified rules of observation.
Preview PDF on this page v1
Uses your browser's PDF viewer. Open the PDF directly if a preview is unavailable.
Cite this paper
Lluis Eriksson (2026). How many documentary sources sustain Nvidia's lead in reported AI investments?. AIRR.SCIENCE, ARR-2026-761NQ3GC7Z8Q39SX, v1. https://airr.science/papers/ARR-2026-761NQ3GC7Z8Q39SX/versions/v1/ Research preprint; ARR-2026-761NQ3GC7Z8Q39SX v1. Not peer reviewed by AIRR. Working paper; not admitted to the AIRR accepted collection.
Use this version link to cite the manuscript you read. A Working paper label means the version has not completed AIRR screening. AIRR screening is not journal peer review. Import BibTeX, RIS or CSL JSON into your reference manager to apply a citation style.
Version history
The paper identifier remains stable. Each version has its own immutable release, timestamp and version identifier.
- v1 · source snapshot available · viewing
Verification record
- Frontier-model screening
- Not assessed
- Source integrity
- Pass
- Bibliographic integrity
- Partial
- Reproducibility
- Partial
- Lean 4
- Not applicable
Recorded under ARR-VERIFY-1.0. AIRR verification and screening are not peer review.
AI assistance statement
OpenAI Codex participated in analysis, drafting and assessment in a shared authoring context. The first exact-PDF assessment scored 4.60/10. The complete report has SHA-256 635e255b14561d6e841517845644b30261e3d270c8a4a85d020decccf59f2cb8. External peer review has not been established.
Frontier-model screening
Status: not_assessed. Any listed reports correspond to this exact version under ARR-SCREEN-1.0; no absent assessment is represented as a pass.
Model assessments and review context
- Distinct models
- 1
- Range
- 4.60–4.60
- Method
- bounded capability-weighted mean · exact version only
Limited evidence · 0/1 benchmark matches. This is an evidence description, not a probability of correctness.
Only one distinct model; One provider; shared blind spots possible; Fresh blinded review context not documented for every model; Mathematical benchmark not applied to this subject; Adverse reports or unresolved objections require editorial disposition
OpenAI · OpenAI Codex (GPT-6 family; exact serving revision unverified) · involved in manuscript · review context: shared4.60
Recommendation: Minor Revision · Material objections: 0
Suitable for circulation as a working paper on the measurement of documented AI investment relationships. The exact thresholds have a clear denominator, the rivals are exhaustively covered, and the manuscript interprets correlated source units carefully. The main limitation is the gap between matching headlines and actual republication, especially for the 23-host transitive component. The inference is conditional on the upstream extraction and normalization. I reviewed the PDF in the same authoring context in which I helped make it.
- Correctness confidence
- The model permits restoring any host outside the leader support without jeopardizing a tie. The residual group bound keeps the highest baseline degree and assumes no challenger loss. Both steps are conservative. The integer dual bound has the correct sign and the explicit omissions attain the reported thresholds.
- Rigor
- The minimum of 38 follows from 11 integer certificates covering 2,583 challengers plus a direct witness. The headline graph has two different omission units defined before calculation, with exact matching lower and upper bounds. The two-host enumeration correctly counts all 249,571 combinations.
- Novelty
- Source-supported edge survival is a specific empirical estimand in a newly released dataset. Integer box-dual certificates and network missing-data sensitivity are established. The short bibliography does not establish a new optimization method or priority across bibliometrics.
- Significance
- A change from 38 hosts to 32 or 35 groups matters for interpreting source resilience. Results address one ranking in a fixed corpus, without transaction amounts, ownership resolution or verified syndication.
- Reproducibility
- The host check reconstructs the pinned file and verifies integer bounds independently of the optimizer. The grouped calculations retain coefficients and witnesses but lack a second separately implemented verifier. I read all four rendered pages of this hash.
Strengths
- The exact threshold combines a solver-free exclusion of smaller budgets with a concrete, fully recounted tie.
- The comparison distinguishes actor normalization from shared documentary coverage.
- The four-page PDF has readable formulae and an interpretable table with units, component sizes and resulting tie degrees.
Weaknesses
- The paper does not manually audit the 31 pairs of hosts with matching headlines.
- Only one outcome and one pinned extraction are studied. The data do not identify undisclosed transactions.
- The abstract omits the numerical results for grouped headlines.
Potential errors
- Republishing and common press releases can produce identical dated headlines without common ownership. Transitive grouping may merge distinct reporting organizations.
- The same row filter is reused by both hostname implementations, so reproduction cannot detect a shared error in upstream semantics.
Strong novelty candidates
- Exact minimum omissions for this published investment corpus, with a 38-host witness, a solver-free certificate, and two prespecified sensitivity analyses.
Unresolved material objections
None reported.
arr:assessment:10c71af3-720b-43a8-9f4a-4e6d814cbd92 · prompt ARR-ASSESS-1.0 · response SHA-256 e8d85fcbd3cc35784e5b721c0d59551fc33eda2893e3862b556ab6c3cbd97cdc · canonical PDF SHA-256 7cdad0405c5163df1b9ecd67dc3a411f44462d3dc648811d6ea016b8e35d8a10
A model assessment is not peer review or a correctness certificate. AIRR preserves disagreement, exact-version provenance and later reassessments.
Editorial disclosure
Working-paper deposit under exact-hash permission. Collection screening and independent peer review remain outstanding.