The Dual-Class Sunset Loophole: How Founder Control Gets Rebuilt

A dual-class sunset sounds like a wonderfully adult compromise. Public investors fund the company. Founders keep extra votes long enough to protect the long-term plan. Then a date, ownership threshold, or departure event reunites economics and control.

Everyone gets to leave the IPO roadshow believing they won. The founder hears "vision." The institutions hear "temporary." Counsel hears the gentle closing of a very expensive binder.

Cloudflare has now shown why the compromise is incomplete. Its existing Class B stock carries ten votes per share and converts when specified ownership, death or disability, or holder-election triggers occur. There is no fixed calendar sunset. The company's special committee estimated that continued historical selling by co-founders Matthew Prince and Michelle Zatlyn could have caused the structure to end as early as 2034. It also acknowledged that the founders could extend it by simply retaining enough Class B.¹

The ownership trigger still imposed a real discipline. Selling economic stock eventually consumed voting control. Cloudflare's proposed restructuring breaks that link.

Shareholders approved the necessary charter changes on June 30, 2026. The company now plans to exchange substantially all founder Class B shares for a new non-economic preferred share carrying nine votes, plus a regular Class A share carrying one vote. It also plans to create non-voting Class C stock for future compensation, acquisitions, and financing. As of August 6, Cloudflare expected to implement the changes as early as September, subject to pending litigation.²³

The old ten-vote security disappears. Ten votes remain. Governance has discovered the benefits of changing outfits in the car.

The proposal is easier to understand as a sequence than as a collection of share classes.

First, every outstanding Class A and Class B share receives one new Class C share through a stock split. Class C has the same economic rights as Class A, including equal treatment in dividends and specified corporate transactions, but generally has no vote.¹

Second, substantially all Class B shares held by Prince, Zatlyn, and related entities are exchanged as follows:

The Series FF share is almost pure governance. It does not participate in dividends, cannot convert into another security, and receives no common-stock consideration in most change-of-control transactions. Its job is to vote. The Class A and Class C shares carry the money.¹

Nine plus one restores the old ten. Cloudflare says the founders' relative voting power will not increase immediately after the transaction. That statement is accurate and carefully framed. The point of the structure is what happens after immediately.

Under the old system, issuing Class A to employees or acquisition targets diluted the founders' percentage of votes. Founder sales also required Class B to convert into lower-vote Class A. Under the new system, Cloudflare expects to issue Class C for equity awards, M&A, and equity financing. Those shares expand the economic denominator while adding zero to the voting denominator.²

This is dilution with the awkward social part removed. Employees and sellers can receive more of the company. The control group does not receive more competition.

Ownership sunsets work because they connect two founder desires that eventually conflict: liquidity and control. A founder may sell, but enough selling causes the high-vote class to convert. The mechanism does not forbid wealth. It sends a bill for keeping the votes.

Cloudflare's Class C makes a founder sale less expensive in governance terms. The proxy states that the co-founders may sell up to all their Class C shares without affecting their voting power or whether Series FF remains outstanding. Their capital-at-risk covenant instead requires them to retain more than 40% of a baseline number of Class A shares. Class C is excluded from that test.¹

The special committee estimated that the retained Class A commitment represented about $1.4 billion at the proxy's May 2026 price and post-split assumptions. That is meaningful economic exposure. It is also a fraction chosen inside a package that lets the founders monetize an entire separate class without surrendering one vote.

The consolidated shareholder plaintiffs describe the package differently. They allege that it preserves founder control and significant liquidity without sufficient consideration for everyone else. Bloomberg Law reported claims that the founders could sell approximately $3.6 billion in new Class C and retain control with an economic stake falling as low as roughly 2%. Those are allegations, and Cloudflare says the claims lack merit.⁴

The broader point does not depend on the litigation's outcome. A voting structure that survives future stock compensation and stock-funded acquisitions changes who pays for growth. Employees still experience economic dilution. Acquisition targets still receive equity exposure. Founders have arranged for their voting percentage to enjoy a more selective wellness plan.

The easy criticism is that every dual-class company is an aristocracy with an investor-relations page. That misses the reason sophisticated investors sometimes accept these structures.

Founder control can protect a company from short-term pressure while a strategy is still compounding. Former SEC Commissioner Robert Jackson acknowledged that dual-class control may benefit a company for a defined early period even as he argued against perpetual structures. The real dispute concerns when the value of insulation becomes smaller than the cost of entrenchment.⁵

Cloudflare's special committee made a company-specific case. Revenue grew from about $192.7 million in 2018 to about $2.2 billion in 2025, and the stock rose from its $15 IPO price to more than $200 by May 27, 2026. The committee concluded that Prince and Zatlyn remained central to long-term strategy. It also negotiated protections that the old structure lacked:¹

• Series FF is subject to the 40% Class A capital-at-risk requirement
• service and board-role conditions can trigger redemption, sometimes after an 18-month period and an independent-director decision
• a majority of the board must satisfy independence and disinterestedness requirements
• acquisitions using more than $100 million of Class C require approval from a majority of independent directors

These are genuine concessions. A service-based condition can be more rational than an arbitrary calendar if the purpose of extra votes is to protect a particular founder's judgment. Research on non-time-based sunsets also finds that ownership and event triggers can preserve strategic benefits while limiting some agency costs.⁶

Then the negotiation record supplies the uncomfortable part. The special committee proposed a 15-year sunset. The founders rejected it. The committee ultimately dropped the fixed term, excluded Class C from the capital-at-risk requirement, and accepted a structure where continued board service can preserve Series FF after year five. The founders also rejected the committee's proposed commitment not to rely on controlled-company governance accommodations. The committee held 18 formal meetings and hired independent advisers. Process was abundant. The founders still received the two terms that matter most: Class C liquidity does not consume votes, and Series FF has no fixed expiry.¹

This does not prove the bargain was unfair. It identifies the bargain. "Founder retention" is the respectable description. The consideration supplied by public holders is continued limits on their ability to replace the people being retained. Most retention packages at least pretend the employee might leave for a competitor.

Cloudflare's Class A and Class B voted together. Each Class A had one vote. Each Class B had ten. On the record date, the co-founders beneficially owned 50.3% of the total voting power. The proxy said they could approve the charter changes without another shareholder, and the restructuring was not conditioned on a majority vote of unaffiliated holders.¹

The same proxy page states what that 50.3% cost the co-founders in stock: their Control Wedge shows votes running far ahead of the shares they actually hold.

Proposal 4A, which authorized Class C, passed with 382,323,567 votes for and 200,557,125 against. The related Class C split passed by a similarly large stated margin.³

"Shareholders approved" is therefore a correct description of the statutory tally. It is weak evidence that outside holders wanted the deal. The meeting asked the shareholder base for permission after identifying the two shareholders who already carried enough permission in their pockets.

That distinction sits at the centre of the lawsuits. As of August 5, eight complaints had been filed. Seven were consolidated and alleged breaches of fiduciary duty connected to preserving founder control and liquidity without adequate consideration. The eighth alleged that Cloudflare's charter required a two-thirds vote for Proposal Four. Cloudflare and its board dispute both theories, and no court outcome was reported in the company's August 6 filing.²

A founder should separate three questions that proxy headlines often compress:

1. Did the proposal receive the votes legally required?
2. Could the controller supply those votes alone?
3. Did unaffiliated holders independently approve the controller's new benefit?

Only the third question tests the bargain with the people paying for it.

Swatch Group supplied a cleaner demonstration in May 2026.

Bearer shareholders first voted on whom to nominate as their representative to the board. They backed activist investor Steven Wood with 80.4% of the bearer-only vote. The nomination then went to the full general meeting, where registered and bearer shares voted together. Wood received 19.2% support and was rejected by 79.6%. The constituency chose him. The capital structure overruled the constituency.⁷⁸

The result came from Swatch's share math. At the end of 2025, the company had 116,919,500 registered shares with CHF 0.45 par value and 28,936,000 bearer shares with CHF 2.25 par value. Every share carries one vote, while dividend entitlement follows nominal value. Equal nominal capital held through registered shares therefore produces five times as many votes as bearer shares.⁹

The Hayek pool and related parties controlled 44.5% of votes with 26.4% of the capital. Many outside investors and funds held bearer shares. Proxy advisers ISS and Glass Lewis backed Wood, and governance foundation Ethos supported him while arguing that the board needed greater independence.⁸¹⁰

Swatch can still say every share receives one vote. The sentence is pristine. One class simply receives five shares for the same nominal capital. Luxury is often a matter of presentation.

This is why vote percentages require a denominator and a constituency. An 80.4% mandate and a 79.6% rejection can describe the same candidate on the same day. Neither count is fabricated. One reveals preference among the bearer holders. The other reveals control.

Council of Institutional Investors supports one-share, one-vote and recommends that unequal voting structures sunset within seven years, with any extension approved by all shareholders voting on an equal basis.¹¹ That extension requirement matters more than the elegance of seven years.

A robust sunset must constrain the result, regardless of which certificate, preferred series, split, recapitalization, or unusually creative alphabet delivers it. Before accepting one, map these tests:

Run the table against four scenarios: the founder sells half their economic stake, the employee pool doubles, a large acquisition is paid in stock, and the founder leaves management while keeping a board seat. Then calculate economics and votes separately after each event. A cap table that stops at ownership is being asked to conceal the interesting half.

There is a useful historical comparison. In a 2017 Delaware case involving NRG Yield, a controller pursued low-vote acquisition currency to prevent its control from eroding. The transaction was conditioned from the outset on both an independent committee and approval from a majority of unaffiliated shares. The court treated perpetuation of control as a controller-specific benefit, then dismissed the claims after finding the procedural protections satisfied.¹²

That decision does not decide Cloudflare's pending cases. It supplies the better negotiating question: if preserving control benefits the controller, why does the controller get to cast the deciding vote on the price?

A dual-class sunset causes enhanced-vote shares to convert into ordinary shares after a specified trigger. The trigger may be a fixed date, the founder's ownership falling below a threshold, a transfer, death or disability, or departure from the company. Read the exact trigger and the amendment rules. "Sunset" does not always mean a calendar expiry.

No. Shareholders approved the charter proposals required for the restructuring on June 30. In its August 6 Form 10-Q, Cloudflare said it expected to implement the capitalization changes as early as September 2026, subject to the related litigation.²

The initial split is proportional, so every existing common holder receives the same new economic security. The asymmetry appears later. Cloudflare expects to use Class C for compensation, acquisitions, and financing, which can dilute economic ownership without diluting existing voting percentages. The founders may also sell their Class C without reducing their Series FF votes.

It proves the applicable voting threshold was met, subject to any legal challenge about that threshold. Fairness and independent support are separate questions. If a controller can approve the proposal alone, ask for the result excluding the controller and whether the transaction was conditioned on unaffiliated approval.

A sunset works only when the people holding control cannot replace it unilaterally before the trigger arrives. Put the limit on the voting outcome. Make any extension or substitute pass a one-share, one-vote ballot of the holders who do not receive the extra power.

Otherwise the sunset is scenery. The control machine has already gone indoors.