AI Giants Borrowed Billions, Then Lobbied for the Rules They Needed
A documented trail of bond issuances, sovereign wealth investments, and federal subsidy flows reveals how the companies financing the AI boom are simultaneously shaping the regulations governing it.
The single most consequential undisclosed financial arrangement in American technology policy today is the profit-share agreement at the heart of the Microsoft-OpenAI relationship. Microsoft's January 2023 press release confirmed a $10 billion investment in OpenAI — a figure verified in Microsoft's FY2023 10-K filed with the SEC in July 2023. But the reported terms, first published by The Information and Bloomberg in January 2023, describe a structure in which Microsoft receives approximately 75 percent of OpenAI profits until it recoups its investment, after which its stake converts to roughly 49 percent. Neither party has confirmed these terms in a public filing. That gap between what is reported and what is disclosed is where federal AI policy is being made.
The capital architecture underlying AI's expansion is now large enough to move bond markets. Microsoft filed a Form 424B5 with the SEC in November 2023 disclosing a $10 billion investment-grade bond offering spanning 2-year, 5-year, 10-year, 30-year, and 40-year tranches. The company did not designate proceeds for any specific purpose, but its stated AI investment priorities and the concurrent OpenAI relationship create a direct financial linkage. Microsoft's total long-term debt stood at approximately $43.9 billion as of its 10-Q filed with the SEC in October 2024. When a single company issues debt at this scale in service of a technology platform it co-owns, the line between private capital markets and effective infrastructure policy dissolves.
OpenAI itself secured a $4 billion revolving credit facility led by JPMorgan Chase in 2024, with reported participation from Citigroup, Goldman Sachs, Morgan Stanley, and Wells Fargo, according to Bloomberg reporting from July 2024. No official OpenAI filing exists confirming the terms, interest rates, or covenant structure. That last omission is material: if the facility's covenants include conditions tied to OpenAI's conversion from its capped-profit structure to a full for-profit entity — a conversion disclosed as a condition of the company's October 2024 equity raise per OpenAI's own announcement and Reuters reporting — then a consortium of the largest banks in the United States has a direct financial interest in the outcome of a California Attorney General regulatory review. The public record does not confirm or deny this linkage. That is the problem.
On the equity side, OpenAI's October 2024 funding round raised $6.6 billion at a $157 billion valuation, as confirmed by OpenAI's October 2, 2024 press release. Thrive Capital led with approximately $1.25 billion, per its own public statement and Wall Street Journal reporting. SoftBank contributed $500 million, per SoftBank's own quarterly earnings disclosure. MGX, a vehicle adjacent to Abu Dhabi's sovereign wealth apparatus, reportedly invested approximately $200 million, per Bloomberg reporting, though MGX has not independently confirmed the precise figure. The full scope of Gulf sovereign participation in OpenAI remains undisclosed. This matters because OpenAI removed its prohibition on military use from its terms of service in January 2024 — a change documented via archived versions of the policy and reported by The Intercept — and subsequently announced a partnership with Anduril Industries for defense AI applications, confirmed by both companies' press statements. A company with undisclosed sovereign wealth investment from Gulf states and an active defense AI partnership is not a purely commercial entity. It is an actor in national security. The dollar value of any Department of Defense contracts with OpenAI is not in the public record.
Anthropol presents a parallel and equally incomplete picture. Amazon's investment of $4 billion in Anthropic, completed in March 2024 and confirmed in Amazon's Q1 2024 10-Q, was structured not merely as equity but as a compute-for-equity arrangement requiring Anthropic to use AWS as its primary cloud provider. Google's investment of up to $2 billion, announced in October 2023 per Alphabet's own press statement, carried a parallel commitment to Google Cloud as a secondary provider. The result is that Anthropic's two largest investors are also its two largest vendors, and the equity stakes those vendors hold are at least partially financed by the operating revenue Anthropic pays them for cloud services. The percentage equity stakes, the compute credit components, and whether either investor holds board representation are not disclosed in any public filing. Jeff Bezos, founder and largest individual shareholder of Amazon, has also made a reported personal investment in Anthropic of between $500 million and $1 billion, per Bloomberg and Wall Street Journal reporting. No SEC filing is required for a personal investment in a private company, so no public document governs the conflict between his personal stake and Amazon's institutional stake.
The federal subsidy layer compounds these structural questions. The CHIPS and Science Act authorized $52.7 billion for semiconductor infrastructure — infrastructure that is the physical prerequisite for AI compute capacity. That figure is verified in the text of Public Law 117-167, signed August 9, 2022. The lobbying-to-award pipeline for those funds runs through the same companies whose bond issuances and equity rounds are described above. The precise lobbying expenditures by OpenAI, Anthropic, Microsoft, Google, and Amazon in the periods preceding and following specific CHIPS Act award announcements are available in quarterly filings with the Senate Office of Public Records, and cross-referencing those expenditures against Commerce Department award timing is a documented analytical method this investigation has not yet fully executed. That cross-reference is the next necessary step.
What remains hidden is substantial. The covenant structure of OpenAI's $4 billion credit facility is not public. The equity percentages held by Amazon and Google in Anthropic are not public. The terms of Microsoft's profit-share arrangement with OpenAI are not confirmed in any SEC filing. The CFIUS review status — if any — of Saudi sovereign wealth participation in Anthropic's Series E round is not in the public record, despite Anthropic's documented access to defense and intelligence-adjacent federal contracts. The instrument that would reveal most of this is straightforward: a Senate Commerce Committee subpoena for the full credit agreement between OpenAI and its bank syndicate, combined with a mandatory CFIUS disclosure request covering all foreign government-adjacent investments in AI companies holding federal contracts. Until those records are compelled, the financial architecture governing American AI policy remains a structure the public is financing but not permitted to read.