The most important clause in a partnership agreement is the exit clause, the one you forgot about through management changes and won’t read until years after signing.
In August, Diageo’s new chief executive presented his turnaround plan for the world’s largest spirits maker: (1) a further billion dollars of cost savings, (2) a leaner operating model, and (3) a firm statement that there would be no further disposals. The 34% stake in Moët Hennessy — the champagne and cognac alliance inherited from another century — stays.
Commentators read that as priorities.
But I dug a little, and suggest to read the structure instead. Diageo holds an option to sell the stake back to Moët Hennessy, but at a 20% discount to fair value. The interests were worth roughly €4.3 billion in 2025, and spirits-sector valuations sit near multi-year lows. Exercising the option today means surrendering on the order of €860 million before the first negotiation begins — a penalty applied to a valuation already at the bottom of the cycle.
The stake stays – because leaving would be irrational. But that is not strategy. That is arithmetic written in 1997.
Why a pre-AI alliance is an AI-era warning
In the AI era, partnerships have become the primary mechanism for assembling scarce capabilities — data, models, compute, distribution, regulatory position — faster than rivals can build them. That is my argument in my book, The New Rules of Partnerships: these structures succeed only when commercial incentives, measurement, governance, and economic engineering are designed together, as one system, and they fail most often not at formation but in the years afterward, when nobody is watching the machinery.
The Diageo–Moët Hennessy alliance predates all of this, which is precisely the point. Under old-rules design, the bill for a static structure took twenty-five years to arrive. On the AI-era clock — where capability, distribution, and data positions reprice in quarters — the same design flaw presents its bill in eighteen months. This case is the slow-motion film of a failure mode that could now happen at speed.
The case, in three layers
The economics. The alliance traces to 1987, when Guinness and LVMH signed agreements creating seventeen joint ventures to distribute their brands across Asia-Pacific, the United States, and Europe. When the 1997 Guinness–Grand Metropolitan merger created Diageo, the new company inherited the 34% Moët Hennessy stake and the distribution ventures with it.
For decades the machine paid. And then it didn’t: Diageo’s share of profit from the stake and joint ventures fell from $455 million in fiscal 2023 to $219 million in fiscal 2025. The asset is now recovering — LVMH’s wines and spirits division grew 5% organically in the first half of 2026, with profit up 11% — but the recovery belongs to the partner who runs the business. Diageo rides along.
The control. Eight joint ventures remain. Diageo is the majority owner of all eight. Moët Hennessy manages six of them. Majority economics, minority control: exposure without levers. Meanwhile the two houses compete head-to-head in scotch whisky — a conflict the alliance’s architecture was never redesigned to govern as both portfolios evolved across three decades.
The stress test. In 2020, Moët Hennessy withheld a €181 million dividend. Diageo’s recourse was arbitration, and the relationship at the top went quiet for years afterward. Cash rights — the most basic economic protection in any partnership — failed under stress, precisely when they were needed. And in France, the joint distribution arrangement cost Diageo money, growth, and market share until it reclaimed direct distribution in January 2025: a market-by-market clawback, because wholesale redesign was never on the table.
The failure modes were predictable — because they were predicted
Each layer maps to a failure mode the new rules exist to prevent. Outcomes, not politics, should decide a venture’s future; here, cash rights depended on goodwill and collapsed with it. Economic exposure should never be decoupled from operational control; here, the majority owner manages a quarter of its own ventures. A partnership needs mid-life machinery — triggers, remediation paths, rebalancing rights — not just birth and death clauses; here, the only options were endure or exit at penalty. And scope must be re-governed as portfolios evolve; here, partners became competitors inside their own alliance and the documents never noticed.
Preventable — or simply past its useful life?
Both, because they are the same answer. Every partnership outlives its founding rationale; the founding rationale here was brilliant, and it carried the alliance for a generation. The failure is not that the world changed. The failure is that the structure assumed it wouldn’t. Prevention was never about stopping the aging — it was about designing for it: performance-triggered rebalancing of management rights, valuation collars that neutralize the cycle, scheduled re-attestation of what each partner contributes, and exit pricing symmetric enough that leaving remains a genuine option in bad years as well as good. A partnership that can only be exited in a boom is not a partnership with an exit.
Don’t leave partnerships to the lawyers
None of this is a drafting problem. Lawyers document what has been agreed; they do not design what should be agreed. The gap in this alliance was economic engineering — the waterfalls, collars, and triggers that keep incentives aligned as conditions move — plus neutral attestation of value, so that disputes are settled by measurement rather than arbitration. And there is now a tool the 1997 negotiators never had: AI-based scenario planning that can simulate a structure against demand shocks, valuation regimes, and trade disruptions before anyone signs. Run the 20% put option through that simulation and it fails instantly — it only works in a world where fair value never visits a trough. Structures deserve the same stress-testing we give balance sheets. This is work for multidisciplinary teams — commercial, economic, tax, legal, and analytical together — not for any one profession working alone.
The implication
Before signing any partnership, price your exit under the worst valuation regime you can imagine. If the number makes leaving irrational, you have not structured a partnership. You have entered a marriage in a jurisdiction with no divorce law.
Randy McGraw is the founder of M2 Ventures and author of The New Rules of Partnerships. He has structured over $2.3 billion in partnership and joint-venture value across Japan and Southeast Asia.