Read the contract before the cap table.
Before asking who owns AI, ask who pays whom and who can switch. Power is a bundle of distinct rights, not a single ownership percentage.
Contents & sources
An ownership percentage is an incomplete map of AI. The shareholder, compute supplier, distribution channel and license holder occupy different positions. Cash and control can move in different directions within the same relationship.
When investment becomes a sales channel
The FTC’s January 2025 staff study examined major cloud–AI partnerships. Alongside equity and revenue-sharing rights, it described cloud spending commitments and consultation, control and exclusivity rights. Some investment was linked to purchases of the investor’s cloud services. The study did not determine that anyone had engaged in unlawful conduct.
Our inference: if a capital provider is also a major supplier, a cap table cannot explain the full economic relationship. Map the purchase obligations and cash-recovery routes that accompany the investment.
Different rights can change independently
Microsoft’s 27 April 2026 announcement retained its position as OpenAI’s primary cloud partner while allowing OpenAI to serve all products across clouds. Its IP license runs through 2032 and becomes non-exclusive. Microsoft’s revenue payments to OpenAI end; payments in the other direction continue through 2030 with a total cap.
Our interpretation is to separate four rights: equity, technology access, distribution and claims on cash flow. A weakening of one does not erase the others. A single word such as monopoly or liberation obscures changes inside the bundle.
Freedom includes the practical ability to leave
Permission to choose another supplier and the economics of switching are distinct. Our proposed investigation checks non-cancellable purchase obligations, migration costs, validated alternatives, portability and control of customer relationships. This is an analytical framework, not a claim about undisclosed terms at a particular firm.
Dependency is more than continuing to use a supplier. The relevant question is whether a buyer can respond when terms deteriorate. A firm that stays for competitive price, quality and reliability is not necessarily trapped.
What remains before a market conclusion
The argument proposes an order of investigation, not a stock pick. First decompose contractual rights. Then inspect workable alternatives. Finally ask which future cash flows the market price already assumes. Strong rights can still be a poor allocation at the wrong price.
The countertest is concrete: easier switching, stronger buyer bargaining power and wider access to technology and revenue can weaken concentrated ownership. A contract reveals power. It does not make that power permanent.
Who owns? Who pays? Who can switch? Put all three questions on the same map.
Primary sources checked
An analysis developed from Kamui’s 28 September 2026 post using public primary sources, checked on 4 October 2026. A company announcement summarises disclosed terms; it is not the full private agreement. Interpretations and investigation criteria are kamuitranslator’s editorial judgments.