The One Thing Nobody Gives Up: Why Billionaires Keep Their Companies After Divorce

One man paid 35.6 billion dollars and kept his company intact. Another paid 131 million, 0.4% of that, and lost everything. I spent a long time thinking I was reading about the marriages of the ultra-rich. I was actually reading a lesson in capital structure.

Two numbers that refuse to fit together

In April 2019 Jeff Bezos completed the most expensive divorce in history. He and MacKenzie lived in Washington, a community-property state that splits down the middle by default, and they had no prenuptial agreement. MacKenzie kept 25% of the couple's Amazon stock, about 4% of Amazon, worth 35.6 billion USD. She immediately became the company's third-largest individual shareholder.

By any conventional measure, that is a financial catastrophe.

Nine years earlier, Frank McCourt, owner of the Los Angeles Dodgers, settled his divorce from his wife Jamie. He paid 131 million USD. That is 0.4% of the Bezos figure. In 2011 the Dodgers filed for bankruptcy. In 2012 he had to sell the team to the Guggenheim group for 2.15 billion USD.

A 270-fold difference in money paid. Opposite outcomes.

If divorce risk were measured by the amount paid, these two results should be reversed. They are not. Which means we are measuring the wrong thing.

The thing none of them would give up

A common share bundles two completely different rights into one piece of paper. The first is the economic right: receive dividends, enjoy appreciation, sell for cash. The second is the voting right: decide who sits on the board, who runs the company, which road it takes.

Because the two usually travel together, most of us forget they can be separated. The people who kept their empires did not forget. They separated the two in advance, and when the event came, they gave up the first to keep the second.

MacKenzie Bezos did not merely keep 25% of the stock. She simultaneously transferred to Jeff all of the voting rights on those shares, along with all her interests in the Washington Post and Blue Origin. Bezos lost 35.6 billion dollars of economic value and did not lose a single vote.

When I lined up the biggest cases in exactly those two columns, the table below appeared. The right-hand column is the whole story.

Four ways to make that column zero

Four strategies, not four rungs on a ladder. Each trades away something different.

Buffett: never create a division event

Susan Buffett left Omaha for San Francisco in 1977. She never came back, and the two never divorced. It was she who introduced Astrid Menks to Warren in 1978. For nearly three decades the Buffett Christmas card was signed with three names: Warren, Susan and Astrid. He married Astrid in 2006, two years after Susan died.

The detail most accounts of this story omit: Susan was in no way financially on the outside. She was a director of Berkshire Hathaway and owned 2.2% of the company, worth nearly 3 billion USD when she died in 2004, enough to place her among the 153 richest people in the world.

So the mechanism here is not "he kept everything." The mechanism is that no legal event ever forced a valuation and division of the shareholding. Susan sat on the same side of the table, not the opposite side. And when she died, her will left 10 million USD to each child, 100,000 to each grandchild, 50 million to children's charities, and almost all of the Berkshire stock to the foundation bearing her name. Not a single lot was sold into the market to fund a division.

The cost of this model does not appear in any ledger. Buffett has called letting Susan go the biggest mistake of his life, and says 99% of the fault was his.

Ellison: turn shares into cash at the point of division

Six marriages, five divorces. As of July 2026, Larry Ellison still holds 40.6% of Oracle.

The most important detail is in the second divorce. In 1978 Nancy Wheeler sold her entire stake in the company back to Ellison for 500 dollars. That company is now Oracle.

Do not read the 500-dollar figure as an injustice. Read it as a mechanism. At the point of division, shares were converted into cash by a pre-agreed formula rather than divided in kind. Who keeps the shares and who takes the cash: that decision sat in a contract, not in a courtroom. Every successful case in the table above reuses that same principle.

And there is a notable detail running the other way: three months after marrying Melanie Craft in 2004, Ellison transferred 911,744 Oracle shares to her, about 11.9 million dollars. Separating property does not mean being stingy. It means every transfer is the owner's decision rather than someone else's ruling.

Ellison lives in California, a community-property state that splits down the middle by default. He still holds 40% of a trillion-dollar company after six marriages. The conclusion could hardly be clearer: the default law does not decide the outcome; the contract does.

Musk: remove the legal status itself

Musk married Justine Wilson in 2000. Six weeks after the wedding he handed her a postnuptial agreement: he would keep separately more than 14 million dollars of pre-marital assets plus all income earned during the marriage; the house would be hers only if they had children.

Justine sued. She argued the agreement was unconscionable, that he had undervalued his X.com stake by at least 43 million dollars, and that he had breached the fiduciary duty between spouses. The suit ran right up to Tesla's 2010 listing and was disclosed as a risk factor.

This is the most important technical lesson about timing, and it applies in Vietnam too: an agreement signed after the wedding is far weaker than one signed before it. Once married, a fiduciary duty arises between the two people, and every agreement after that is examined under a far stricter standard than an agreement between two people still bargaining as equals.

With Talulah Riley he applied the lesson. Married in 2010, divorced in 2012 for about 4 million dollars; remarried in 2013, divorced again in 2016 for 16 million in cash plus assets. Not a single Tesla or SpaceX share. And from 2016 to today he has not married again, though he has had children with at least two women. No marriage means no community property; what remains is child support, which is calculated on the child's needs rather than as a percentage of the father's wealth.

Murdoch and Arnault: freeze it before there is a dispute

Rupert Murdoch created the Murdoch Family Trust in 1999, and the immediate reason was to settle his divorce from his second wife, Anna Torv. She took 110 million dollars and gave up the fight for a share of the media empire. The trust divided eight votes: four for Rupert, one each for the four elder children.

The consequence is that the two later marriages, to Wendi Deng (divorced 2013) and Jerry Hall (divorced 2022), could never touch the voting shares, because those shares had left his personal ownership beforehand. His two daughters with Wendi Deng receive economic benefits equal to their siblings, but hold no votes.

But an irrevocable trust cuts both ways. In December 2024 a Nevada court rejected Rupert's own attempt to hand full control to his son Lachlan, finding that both father and son had acted in bad faith. Only in September 2025 did the family reach a settlement: a new trust running to 2050, with Lachlan holding all voting power; the three remaining children taking cash, funded in part by selling about 31 million class B shares, and barred from buying back shares in the two companies later.

Bernard Arnault went one step further. The parent company Financière Agache was converted into a partnership limited by shares. His five children hold equal stakes in the managing entity. The shares are locked for 30 years, with no sale or transfer without the unanimous agreement of the committee of five; after thirty years, only Arnault descendants or entities they wholly own may hold them. The result: the family holds 48% of LVMH's capital but 63.5% of the voting rights.

The crux of this structure fits in one sentence: the spouses of the children never become shareholders. At most they have rights to the economic value of a stake, never to the stake itself.

Four men who had structures, and still lost everything

By now the conclusion looks obvious: get an agreement. But stopping there would skip the most important part of the data.

Frank McCourt had a marital property agreement. Bernie Ecclestone had a trust. Steve Wynn had a shareholder agreement binding his ex-wife's shares. Michael Prest had a whole system of offshore companies holding title to assets. All four lost.

When the agreement exists but is not an agreement

In 2010 a court voided the McCourt agreement. The reason was not the law but the execution. Two different versions had been signed, one giving Frank sole ownership of the team, one without that clause. Jamie had no lawyer of her own when she signed. And both admitted they had never read either version. The court concluded that the two misunderstood each other and that in substance there had been no agreement at all as to content or legal effect.

With the agreement gone, California community-property law returned: the team was community property, split in half. Bankruptcy in 2011, a 131-million settlement, a forced sale in 2012.

Those three defects, a single version, separate counsel for each side, and both parties actually reading and understanding it, are precisely the three conditions for a prenuptial agreement to stand. Doing it properly would probably have cost under 100,000 dollars. The price of skipping it: a baseball team.

When the structure was built for tax, not for control

In December 1997 Slavica Ecclestone, not Bernie, created the Bambino Holdings trust in Liechtenstein. The principal beneficiaries were she and their two daughters. Through an intermediate entity in Jersey, that trust held the entire equity of the Formula One parent company, which later accumulated roughly 4 billion dollars.

The reasoning was sound at the time. Slavica was Croatian and not UK tax-resident, so she could set up a trust in a tax haven and avoid 40% inheritance tax if Bernie died. But the price was a fatal constraint: under UK law, Bernie was not permitted to control that trust. If he had any beneficial entitlement, the whole 4 billion would be treated as his and taxed at 40%.

In other words, to save tax he was obliged to stand outside the single largest asset of his life. They divorced in 2009; she received roughly 881 million dollars and later set up her own family office. In 2023 Bernie pleaded guilty to tax fraud relating to an undeclared trust and paid 652 million pounds to the UK tax authority.

Tax optimisation and retaining control are conflicting objectives. A structure designed so that an asset "is not yours" will succeed in exactly that literal sense, including on the day you need it to be yours. Arnault's Agache structure does the opposite: accept the tax, in exchange for every vote staying inside the bloodline. The same legal instrument, two different objectives, two different endings.

When you divide in shares, the shares come back

Divorced in 2010, Elaine Wynn received roughly 9% of Wynn Resorts, the company she co-founded, becoming the largest individual shareholder, a stake worth nearly 2 billion dollars by 2018. Steve Wynn was not naive: a shareholder agreement tied the voting rights of those shares to him.

But an agreement that binds another person is an agreement that can be litigated. Elaine sued to free her shares, and the case dragged on for years. Then in 2018, when the press published a series of allegations about Steve Wynn's conduct, that pending lawsuit suddenly carried a completely different weight. The two settled: he paid her 25 million dollars, resigned as chief executive, and sold his entire 12% stake. Her shares became free to vote.

Elaine immediately used those votes to launch a campaign to change the board, targeting a director she regarded as her ex-husband's man. All three major proxy advisory firms backed her.

Shares transferred to a spouse do not disappear. They sit on the register and wait. A clause binding voting rights is only a temporary fence: it postpones the conflict, it does not resolve it. Ellison converted shares into cash at the point of division. Bezos took a permanent voting proxy. Wynn chose the third way, and it is the only one of the three that can be litigated back.

When a company holds title but does not own

In Prest v Petrodel Resources (2013), the UK Supreme Court dealt with a very familiar structure: six properties and an interest in a seventh held in the names of two companies controlled by the husband. Total matrimonial assets were about 37.5 million pounds; Yasmin Prest was awarded 17.5 million.

The instructive part is the reasoning, not the number. The court did not pierce the corporate veil; it said expressly that owning and controlling a company is not by itself enough to do so. Instead the court used the doctrine of resulting trust: because the husband had provided the purchase money, he was the true beneficial owner, whatever the paperwork said.

The resulting principle applies to every offshore structure: a court does not need to break the company, it only needs to ask where the purchase money came from.

Two later cases show what happens when people try to fight that principle. Farkhad Akhmedov was ordered by an English court to pay 453.6 million pounds in 2016, 41.5% of the marital assets, then spent five years resisting the order; the court found he had made arrangements to put every penny beyond his ex-wife's reach, with his son participating. She recovered only about 50 million pounds before the two settled in 2021. Dmitry Rybolovlev was ordered by a Geneva court to pay 4.5 billion dollars in 2014, reduced on appeal to about 564 million francs, after six years of litigation spread across six jurisdictions.

Four men. Four structures. Not one of the four had the right kind of structure for the risk he actually faced. An agreement nobody read is not an agreement. A trust built to avoid tax does not retain control. A clause binding votes only postpones the conflict. A company holding title is not a company that owns.

The variable almost nobody talks about

One man in this whole study was in the worst legal position of all, and he still kept his company.

Harold Hamm, founder of Continental Resources, had no prenuptial agreement. Not a single clause. The Oklahoma court applied the doctrine distinguishing "active" appreciation, created by the spouses' efforts and therefore divisible, from "passive" appreciation driven by the market. In 2015 he was ordered to pay 974.8 million dollars.

He wrote the cheque. And kept the company intact.

Dang Le Nguyen Vu kept all of his Trung Nguyen shares because he could pay 1,318 billion dong in cash. The Murdoch family reached its 2025 settlement because it sold about 31 million class B shares to raise the money to buy out three children. McCourt had no source for his 131 million other than the team itself, so he lost the team.

A clause decides whether you have the right to buy the other side out. Liquidity decides whether you can exercise that right. These are two entirely different things, and the second is almost always forgotten when people discuss asset protection.

How this plays out in Vietnam

We have one very good precedent and one very large gap.

The precedent is the Trung Nguyen case. Total community property brought before the court was about 7,900 billion dong, divided 60/40. What matters is how it was divided: the court awarded all of the Trung Nguyen shares, worth about 5,655 billion dong, to Dang Le Nguyen Vu, and ordered him to pay Le Hoang Diep Thao 1,318 billion dong in cash to balance. She received seven properties worth about 376 billion and all bank deposits in her name, more than 1,551 billion.

That is exactly the "divide by value, not by shares" principle that Ellison achieved with a half-page contractual clause. A Vietnamese court is willing to do it. The only difference, and it is a large one, is the price of letting a court do a contract's work: four years of litigation, the brand exposed to the public throughout, and the requirement to have 1,318 billion dong in cash ready at a moment nobody gets to choose.

And here is what happens when there is nothing at all. In 2007 Le Quang Tien, vice chairman of FPT, transferred 1,854,815 FPT shares to his ex-wife under a ruling of the Ba Dinh district court in Hanoi. At the then price of 560,000 dong a share, that stake was worth about 1,040 billion dong, enough to place her among the twenty richest people on the Vietnamese stock market. It was the first time a major shareholder of a Vietnamese blue chip had to transfer a large block of shares because of a divorce, and the market at the time genuinely worried about selling pressure.

The 2011 Bao Son Group case shows a third kind of damage, the kind nobody books on a balance sheet: roughly 500 million dollars of disputed assets spread across seven related companies, running for years. When a dispute spreads across a group of companies, the biggest loss is not the share of assets divided. It is the years during which no entity in the group can make a major decision.

The tools that exist and their limits

Since 1 January 2015, Vietnam has for the first time allowed spouses to choose a contractual property regime instead of the statutory one. This is the equivalent of a prenuptial agreement, and most Vietnamese business owners still do not use it.

• Article 47, form. It must be in writing, notarised or authenticated, and must be made before the marriage is registered. It takes effect from the date of registration. There is no door open afterwards, which means the Musk lesson becomes absolute in Vietnam: a postnuptial agreement does not exist.
• Article 48, content. It identifies which property is community property, which is the wife's separate property and which the husband's; the rights and obligations of each party over each pool; and the conditions, procedures and principles for division when the regime ends.
• Article 50, invalidity. The agreement is void if it breaches the conditions for a valid civil transaction, breaches the principle of equality between spouses or the family's interests, or seriously breaches rights to support, rights of inheritance and the legitimate interests of other family members.
• Article 116 of the 2020 Enterprise Law. Voting preference shares are available only to founding shareholders, and convert automatically into ordinary shares three years after the business registration certificate is issued. Far too short to be the foundation of any long-term structure.

Practitioners have also pointed out two practical gaps: a marriage certificate does not record whether a couple has adopted the contractual regime, so banks and counterparties have no way to verify it; and the law is not explicit about whether notarisation is a condition of validity or merely a matter of form. For someone whose name sits on several legal entities, the first gap is a real problem: the agreement protects you against a spouse, but not necessarily against a creditor.

Because Article 116 all but closes the door on voting preference shares, the practical tool in Vietnam for separating votes from capital is an intermediate holding company plus a shareholder agreement and a voting proxy. In other words a miniature Arnault model, built with locally available materials.

Three things that are not in the charter

There is a local risk that research on the West does not capture, and a case in China exposes it very clearly.

Li Guoqing and Yu Yu co-founded Dangdang in 1999 while married. After their separation, Yu held 52.23% of the shares and all operating authority. In April 2020 Li went to the Beijing headquarters and took away roughly fifty company seals along with the finance seal, declaring that the board had elected him chairman. Police had to intervene.

In China as in Vietnam, the seal is the signature of the legal entity. Whoever holds the seal can sign contracts, open accounts and file documents. That is why someone holding a minority of the shares can still paralyse an entire company.

So alongside shares, three things decide who can actually operate a business tomorrow, and none of the three is in the charter: who is the legal representative, who holds the seal, and who is the bank account holder. A family dispute touches all three before it ever reaches the share register.

The QuietVC lesson

There is one historical detail that frames the whole subject, and it precedes everyone above by almost a century.

In 1867, before Hetty Green married Edward Green, her father insisted on a marriage contract providing that her inheritance would be free from the debts, control or interference of any husband. Eighteen years later Edward defaulted on 700,000 dollars. Hetty paid his debt, never forgave him, and kept her fortune intact. In 1909, when her daughter married, she required the groom to sign exactly such a contract.

Hetty Green understood what Buffett, Ellison, Murdoch and Arnault all understood, and what Hamm and Mr Vu had to learn with cash: an ownership structure is only trustworthy when it is built at a time when nobody needs it yet.

But the real lesson of this study is not "get an agreement." Four of the men who lost everything had perfectly real structures. What they lacked was not paperwork. It was a match between what the paperwork was for and the risk they actually faced.

And this is the most familiar mistake in investing, wearing different clothes.

An investor buys a stock for reason A, holds it through an event of type B, and is surprised that the thesis did not save them. A good business does not protect you from a bad balance sheet. A cheap price does not protect you from a dying business model. Insurance bought against fire does not pay for a flood. And a trust built to avoid inheritance tax does not retain voting control; it does precisely the job it was built to do.

The second point is even more practical. In all of the cases above, what decided the outcome was not a clause but liquidity. Whoever had cash on hand to buy out the other side kept the company, even with no contract at all. Whoever had to sell the core asset to pay lost it, whether the sum due was 131 million or 35.6 billion.

Margin of safety, once again, is not a number in a valuation model. It is the distance between the obligation you may have to carry and the money you can raise without selling something you do not want to sell. That holds for a stock portfolio, for a mortgage, and for ownership of a business.

One last thing, because it is the most honest part. A structure protects the company, not the people. Murdoch had to go to court against his own children. Buffett kept Berkshire whole through twenty-six years of separation, and calls that separation the biggest mistake of his life, taking 99% of the blame. The tables in this article measure a great many things. They do not measure that.

Silent discipline. Compound value.