Cloudflare's Founders Wanted to Sell the Stock and Keep the Votes. Then a Judge Read the Texts.

On October 16 the Delaware Court of Chancery did something it does not often do to a $100 billion company that has hired a banker and a law firm, held eighteen committee meetings, kept hundreds of pages of minutes, and won a stockholder vote. It said stop. Cloudflare may not close the recapitalization that would let its two founders sell down their stock while keeping control of the company, and the three directors who approved it will sit in a witness chair and explain what they believed, and for whom, before the company issues a single nonvoting share.

That is the holding, and I want to state it with care, because the care is the point. Vice Chancellor J. Travis Laster granted a preliminary injunction because the stockholder plaintiffs showed a reasonable probability of proving at trial that the committee which approved the deal did not act in good faith, the one requirement of Delaware's new safe-harbor statute the court could not yet say was met. It is not a finding of bad faith; the court wrote that the committee members "may well testify credibly." What it could not do, yet, was believe them on paper, partly because of the evidence and "partly . . . due to the destruction of evidence."

The fraction

Matthew Prince and Michelle Zatlyn met at Harvard Business School in 2007 and started Cloudflare two years later. They took it public in 2019 at $15 a share with ten-vote stock for the insiders and one-vote stock for everyone else, and the stock hit $310.59 on August 10. The numerator shrinks every time the founders sell, which they have, about $1.18 billion for Prince and $629 million for Zatlyn since 2020. The denominator grows every time the company pays an engineer in Class A shares. Their share of the vote peaked at 60 percent in 2021, stood at about 51 percent in March and 50.3 in June, and at the historical pace would have dipped under a majority by December.

The cure has two steps. The founders trade each ten-vote Class B share for one Class A share plus a share of new preferred stock carrying nine votes and, in the court's words, "only nominal equity value." Then every Class A share splits into one Class A and one nonvoting Class C. The votes live in the preferred, which the founders keep; the money lives in the Class C, which they can sell. "Problem solved," the opinion says. By the court's arithmetic the founders could have raised about $3.7 billion before falling under the one-third voting floor the statute treats as the threshold of control; after the deal, about $5.8 billion. The committee never put a number on what the public holders gave up. The defendants' own expert, once the lawsuit forced the question, allowed that extending the founders' control could cost $3.2 billion, and argued they would earn it back.

This is corporate-governance work of the most familiar kind: a conflicted controller, and a committee of outsiders deputized to stand in for the people who cannot be in the room. Here they were Karim Lakhani, a Harvard Business School professor who once taught both founders, Scott Sandell, a venture capitalist sixteen years on the board, and Katrin Suder, a physicist and former German defense official. The board found all three independent, the court agreed, and that is where the defendants' good news ends.

What the junior bankers wrote down

Everyone understood a lawsuit was coming; Centerview Partners, the committee's banker, told it that 80 percent of the closest precedents had drawn one. So it built a record, and the irony of this opinion is that the record is what sank it. The most damaging documents are the notes of junior Centerview bankers, who on February 4 wrote down that the advisors needed "to figure out the narrative," and that at some point "the committee has to own the story." The defense witnesses swore, "vehemently and in lockstep," that the notes were unreliable. The court found them consistent with how the deal unfolded, and gave the reason that settles it: the junior team members "had every incentive to create accurate accounts for their superiors." The least powerful people in the process produced its most honest record.

On March 9 the committee met the founders to ask what they wanted. Centerview had put a "broad universe" of alternatives on the table; the committee showed the founders two, Prince "loved" them, and the rest were never presented again. The next day Prince created a Signal chat with Zatlyn and the three committee members, titled it with a bus emoji, and set the messages to vanish after a day. Prince testified he had no idea what the emoji meant, Lakhani was "befuddled," and nobody could recall a word said in it. The court, not obliged to be as forgetful as a deponent, read the chat's probable purpose as getting everyone "on the bus" for what the founders wanted. Probable, it stressed, not proven.

Midnight, May 2

The committee's first proposal asked the founders to keep half their Class A shares or forfeit the preferred. The founders countered at ten percent. In between, the committee's lawyers had called the fifty "a placeholder." Then Lisbon, where Lakhani, having promised the committee he would not discuss the deal with Prince, told him over dinner on April 30 that he expected "the Committee was going to be able to get to where [the Founders'] proposal was." Nobody had authorized that. The committee's second proposal, drafted the same day, put the fifty back, and Sandell drew the job of previewing it.

The opinion reproduces what followed, over defense counsel's request that it stay out of open court. From 4:05 on the afternoon of May 2 until 12:22 the next morning, the chief executive of a $100 billion company wrote to a director of sixteen years' standing. "Silence is fun. Keep not replying. Builds trust." "You move nothing on the only thing we care about. Go. To. Hell." "Figure it out. 12 hours or I tender my resignation." And, at 12:16, with a clairvoyance the record now rewards, "This is going to make a helluva case study."

Sandell's side of that chat is the part of this record I would hand a new director. At 11:50 p.m.: "Hi Matthew, I have been putting kids to bed but am happy to talk if it's helpful." At 12:18: "If you want me to go to hell I going to bed and hope you will feel better in the morning." A man doing a hard job at midnight, politely. Whether he gave anything away is a different question, and the court found a reasonable probability that the tirade worked, "shifting Sandell into a conciliatory mindset." After he went to bed, Prince turned to Carl Ledbetter, a director since 2009 and not on the committee at all: "I'm tendering my resignation in 12 hours if this BS isn't resolved." Ledbetter answered immediately: "Let me fix it." He then forwarded Prince's next proposal to Sandell with a note that he had told Prince "we got everyone off 50%." Nobody had authorized that either. On May 10 the final proposal landed at 40 percent of the Class A shares alone, which the court observes was "where Prince had said he expected the negotiations to land."

The chats multiplied and the fuses got shorter: a "Progress Committee" chat on one-month deletion, another cut to a week by Lakhani, who could not say why. Nobody turned auto-delete off until the first litigation hold on June 5. The witnesses testified that nothing of substance happened on Signal, and the court, looking at the messages that survived, declined to believe them. It applied a rule from Smith v. Van Gorkom, that producing weak evidence when strong evidence "is, or should have been, available can lead only to the conclusion that the strong would have been adverse." The strong evidence here was the chats. The chats are gone.

Fifty-eight percent

Cloudflare put the nonvoting stock and the split to a vote on June 30, with the founders' 50.3 percent already committed. Both proxy advisors said vote no. The Class C proposal passed, 382.3 million votes to 200.6 million. Strip out the founders and between 15.6 and 17.7 percent of everyone else's voting power said yes. The statute offers a second safe harbor, an informed majority of the disinterested votes cast, and on those numbers the committee was never going to find it.

The statute, and the door in it

In 2025 Delaware rewrote Section 144 of its corporation law, in a bill signed March 25, to give conflicted controller deals a harbor from the entire-fairness trial that used to await them. If a committee of "2 or more" directors the board has found disinterested approves the deal "in good faith and without gross negligence," it "may not be the subject of equitable relief, or give rise to an award of damages." The defendants won every argument about what those words mean, and that matters for the next company. The injunction carve-out does not reach this recapitalization. Sandell's Thanksgiving with the Prince family was one of the "social-circle friendships" the board was not obliged to dissect, and Lakhani's stay at Prince's house in Lisbon was not material enough to require its attention. As statutory architecture goes, the committee was built to code.

And then the one requirement the drafters left in plain English did its work. The statute does not redefine good faith, and the legislature's synopsis says the law "does not displace the common law requirements regarding core fiduciary conduct." A director fails to act in good faith when she "intentionally acts with a purpose other than that of advancing the best interests of the corporation," and the purpose need not be greed. It can be as ordinary as liking the person across the table more than the strangers who own the stock.

"Humans often display loyalty to other humans, as well as respect and deference for those in power," the court wrote, and those traits "can lead directors to favor proximate and powerful flesh-and-blood humans, when their fiduciary duties require that they pursue the best interests of a metaphysical entity and its distant and dispersed stockholders."

The plaintiffs' theory is the "controlled mindset," then-Chancellor Strine's phrase for a diligent, well-advised committee with its "hands . . . on the oars" while its boat went "if anywhere, backward." The court found enough contemporaneous evidence of that mindset to require a trial, and was scrupulous about the other reading, in which a director honestly believes founder control is what makes a company valuable. Facebook's board, for what it is worth, abandoned a Class C stock dividend of its own in 2017. Meta's board declared a new one on October 19, three days after this ruling, on a special committee's unanimous recommendation, under an agreement that lets the founder sell a fifth of his new shares freely and restricts the rest. The court did not find the committee grossly negligent. It did not find the deal unfair, though it found a reasonable probability that the defendants could not prove it fair. And it offered the defendants a trial limited to good faith, ideally within about ninety days, if they asked within ten days. Counted the way the court's rules count, with weekends left out of any period under eleven days, those ten days end this Friday, October 30. As I write, the clock is still running.

Monday morning, for an Illinois company

What do I make of 142 pages? That you can legislate a safe harbor, and the harbor held on every point the legislature touched. What a legislature cannot write is a person's reasons. If Lakhani, Sandell and Suder take the stand and persuade the court that they believed, when they voted, that founder control was worth what it costs the people who hold one vote a share, the statute will do what it was written to do and the deal can close. If they cannot, Delaware will have learned that the hardest condition in its new law is the one it did not change. The people waiting on that answer are a Teamsters pension fund, California's public employees, and whoever bought at $15 in 2019 under a charter that promised the ten-vote shares would convert once the founders had parted with three-quarters of them. In June they were asked, and fewer than one in five of them said yes, and it did not matter.

Columns are the views of their authors. They are not legal advice and not the position of the firm.