AI Infrastructure: Commitments, Delivery and Access

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

This work was produced through an AI-assisted workflow directed by the author. Selected dated AI infrastructure commitments, delivery stages and access obligations, with executable reproduction of named calculations and tables. No industry-wide causal relay or matched project-finance advantage is claimed. The research design, method, sourcing decisions and analytical judgements are the author's. OpenAI GPT-6 prepared the integrated manuscript, source catalogue and reproduction package through an AI-assisted process; the text was artificially generated. Claude and Gemini contributed earlier source work, drafting and challenges. A separate Gemini review examined the complete dated manuscript and its held source evidence; that review is distinct from the author's release decision. Gemini's earlier contributions limit its independence from related work. A separate Gemini run reviewed the reproduction addition without independently executing its tests. The earlier manuscript review alone does not constitute acceptance of that addition. Separate review does not establish complete validation. The assisting models are not always neutral parties to the subject matter. OpenAI is a named infrastructure counterparty and appears in the political discussion. Anthropic's leadership is discussed, and Google competes with several subjects. The source catalogue and Verification note identify the documents, inspected locations and evidence limits. They do not certify every underlying company statement or a complete causal account. No matched project-finance baseline, complete historical replay, bank reconciliation, site inspection or customer acceptance audit is claimed. Guarantee: within the scope stated in the Verification note, reproduce.py recalculates the named balances and regenerates selected numerical expressions and the follow-up table from frozen inputs. It checks output parity, not source truth or causal validity. A reader can check this without trusting either party. What the author cannot guarantee: errors or omissions can survive model review, including confident but incorrect review findings. The author does not claim the review is exhaustive. Corrections are logged against the DOI when surfaced. No warranty is offered beyond the terms of the CC BY 4.0 licence.