Q001 — Self-Audit¶


Research	R0005 — AI Company Profitability Before 2030
Run	2026-03-17
Query	Q001

ROBIS 4-Domain Audit¶

Domain 1: Eligibility Criteria¶

Rating: Low risk

Criterion	Assessment
Query clearly defined	Yes — profitability of named AI companies before 2030
Scope appropriate	Yes — 8 companies across 3 categories (infrastructure, diversified, pure-play)
Hypotheses comprehensive	Yes — H1 (yes), H2 (no), H3 (depends on definition) cover all positions
Inclusion criteria consistent	Yes — sources selected for financial data quality, independence, and recency

Notes: All commonly identified major AI companies covered. Scope deliberately excludes Chinese AI companies (Western-market focus of the question).

Domain 2: Search Comprehensiveness¶

Rating: Low risk (minor concern)

Criterion	Assessment
Multiple search strategies used	Yes — 19 web searches + 4 page fetches across different angles
Searches designed to test each hypothesis	Yes — searched for profitability evidence, loss/capex evidence, and skeptic analysis
All results dispositioned	Some concerns — experimental run, not all results fully accounted for
Source diversity achieved	Yes — financial news, SEC filings, research orgs, investment analysis, tech press

Notes: 23 searches executed, ~200 results reviewed, ~30 selected. Minor concern: paywalled sources (Seeking Alpha, CNBC) not fully accessible. One WebFetch failed (auth wall).

Domain 3: Evaluation Consistency¶

Rating: Low risk

Criterion	Assessment
All sources scored using same framework	Yes — 8-dimension bias assessment applied to all 8 sources
Evidence typed consistently	Yes — Reported/Analytical/Statistical distinction applied
ACH matrix applied	Yes — all 12 evidence items scored against all 3 hypotheses
Diagnosticity analysis performed	Yes — most and least diagnostic evidence identified

Notes: Same evaluation criteria applied to each company: current profitability, AI-specific impact, forward projections. No source received preferential treatment.

Domain 4: Synthesis Fairness¶

Rating: Some concerns

Criterion	Assessment
All hypotheses given fair hearing	Yes — H2 tested despite seeming unlikely given Nvidia
Contradictory evidence surfaced	Yes — capex pressure, FCF destruction, $115B losses all documented
Confidence calibrated to evidence	Yes — different confidence levels per company reflect evidence strength
Gaps acknowledged	Yes — 5 specific gaps identified

Notes: The evidence landscape is structurally tilted toward optimism — companies publishing projections have fundraising incentives. Bear-case analyst coverage is underrepresented. Rating elevated to "Some concerns" because this structural bias in available data cannot be fully mitigated.

Overall Assessment¶

Overall risk of bias: Low (with synthesis fairness caveat)

The research process was thorough relative to the complexity of the query. The main limitation is not process quality but data availability — private company financials are unaudited and inherently optimistic. The methodology surfaced this limitation explicitly rather than ignoring it.

Researcher Bias Check¶

Survivorship bias — focus on largest companies may miss smaller AI companies with different profitability profiles.
Optimism bias in source material — company projections are inherently optimistic; investor-facing documents tend to present best-case scenarios.
Recency bias — Anthropic's 2-month revenue doubling trend may not sustain at scale and could overweight the assessment.
Category confusion — grouping hardware companies (Nvidia) and model companies (OpenAI, Anthropic) under "AI companies" conflates fundamentally different business models.