Who Actually Controls the Company

Key takeaways

• Control and economics separate legally and permanently. Alphabet reports Class B stock at 10 votes per share, Class A at one and Class C at none; its 2025 Form 10-K states that as of December 31, 2025 Larry Page and Sergey Brin beneficially owned about 89.3% of Class B, representing roughly 52.7% of total voting power.
• Delaware §141(a) puts the business under the direction of the board, and §242(b)(2) gives a class a separate vote on any charter amendment that adversely alters its powers, preferences or rights, which is why a preferred series with 15% of the shares can stop a sale.
• The 83(b) election has a hard 30-day deadline from the transfer of the property; Form 15620, issued April 2025, is the IRS form for it. Miss the window on founder stock and you are taxed as it vests, on shares you cannot sell.
• Anti-dilution formula choice is worth ten times more than the headline term. On the illustrative Series A modelled below, a down round costs the investor-protective adjustment about 361,000 extra common shares under broad-based weighted average and about 3,673,000 under a full ratchet.

The cap table is the wrong document

Founders read their ownership percentage and believe they have read their position. They have read one of three documents, and the least binding one.

The cap table records economics: who holds how many shares of what class, and therefore what fraction of a sale price flows to each row. The certificate of incorporation records power: how many votes each class carries, which decisions require a separate class vote, and what the preferred stock is entitled to before anyone else is paid. The voting agreement records who sits on the board. Delaware General Corporation Law §141(a) is blunt about which of these runs the company: the business and affairs of every corporation shall be managed by or under the direction of a board of directors, except as otherwise provided in the chapter or in the certificate of incorporation. Not by the majority shareholder. By the board.

Public markets show how far the two can drift. Alphabet's certificate authorizes three classes whose rights, in the company's own words, are identical except with respect to voting: Class A carries one vote per share, Class B carries 10, and Class C carries none except where law requires. Its Form 10-K for fiscal 2025 states that as of December 31, 2025 Larry Page and Sergey Brin beneficially owned approximately 89.3% of outstanding Class B stock, representing approximately 52.7% of the voting power of outstanding common stock. The filing goes on to note that because Class C carries no votes, issuing Class C to fund acquisitions and employee equity programs can continue that relative voting power indefinitely. Economic ownership can be diluted every year while control is not diluted at all.

Private companies achieve the same separation with different tools: board designation rights, protective provisions and a signed voting agreement rather than a supervoting class. The mechanism differs; the lesson is identical. If you want to know who decides, count board seats and read the veto list. The percentage column tells you what you get paid, not what you get to do.

Vesting: the four-year clock and the one-page form

Vesting is the only term in a startup that protects founders from each other, and it is the one most often waived at the beginning because everyone is friends.

The convention is four years with a one-year cliff: nothing vests for twelve months, then a quarter vests at once, then monthly thereafter. Illustrative only: a founder holds 5,000,000 shares of restricted stock on that schedule. If she leaves at month 10, she has vested nothing and the company repurchases all 5,000,000. If she leaves at month 18, she has vested 5,000,000 x 18/48 = 1,875,000 shares and keeps them. The alternative, no vesting at all, means a co-founder who quits in month three walks away owning half the company forever, contributes nothing further, and blocks the cap table every investor will look at.

The tax mechanic underneath is where money is genuinely destroyed. Restricted stock is taxed as it vests, at the value on each vesting date, unless the holder files a section 83(b) election. That election has a hard 30-day deadline from the date the property is transferred; IRC §7503 extends it only when day thirty falls on a weekend or legal holiday. The IRS issued Form 15620 in April 2025 as the standard form, and a compliant written statement under Treas. Reg. §1.83-2 still works.

Illustrative only, continuing the same grant: she buys 5,000,000 shares at $0.0001, paying $500 when fair market value is also $0.0001, so an 83(b) election reports $0 of income. Without it, she recognises ordinary income every month as shares vest. If the company's valuation has risen to $1.20 per share by month 24, the tranche vesting that month, 5,000,000 ÷ 48 = 104,167 shares, produces roughly $125,000 of ordinary income on stock she cannot sell to pay the tax. The bill repeats monthly and grows with every markup. One page, filed inside thirty days, is the difference.

Options have their own trap. Under §409A a discounted option is deferred compensation, and non-compliance triggers an additional tax of 20% of the amount included in gross income plus interest at the underpayment rate plus one percentage point, and it falls on the employee, not the company. That is why boards buy independent 409A valuations before setting strike prices. Rule 701 is the related plumbing: a non-reporting issuer may sell compensatory securities in any consecutive 12-month period up to the greatest of $1,000,000, 15% of total assets, or 15% of the outstanding amount of the class, and must deliver risk factors and financial statements once sales exceed $10 million in that period.

Acceleration is where founders get careless. Single-trigger acceleration on a change of control makes the company less attractive to buy, because the acquirer inherits a fully vested team with no reason to stay. Double-trigger, meaning acceleration only if the company is acquired and the person is terminated without cause, is the term that survives negotiation. Negotiate it at the grant, not at the sale, when your leverage is gone.

Three rounds, watched from the board table

Dilution is arithmetic. Losing the board is a discrete event, and it usually happens at a moment when the founders still hold a majority of the shares.

Illustrative only, one company across three rounds. Two founders start with 10,000,000 shares, 5,000,000 each, and a two-seat board they both sit on. Each round is priced so the new investor takes 20% of the post-money company and the option pool is refreshed to 10% of post-money, carved out of the existing holders before the money comes in.

Seed. Existing holders are 70% of post-money, so total shares become 10,000,000 ÷ 0.70 = 14,285,714. The investor takes 2,857,143 shares, the pool becomes 1,428,571, and each founder's 5,000,000 is now 35.0%, or 70.0% between them. The board goes to three: two common seats, one Series Seed designee. The founders still outvote the investor two to one.

Series A. Non-pool existing shares of 12,857,143 are again 70% of post, so total becomes 18,367,347. The new investor takes 3,673,469, the pool is topped to 1,836,735, and each founder holds 27.2%, or 54.4% combined. Then the board goes to five: two common, two preferred designees, one independent director whom both sides must approve. This is the event. The founders control 54.4% of the equity and two of five board seats. Under §141(a) the board directs the company, so from this point a majority of the shares no longer wins a disputed decision. Nothing improper happened; it is in the term sheet, and almost nobody models it.

Series B. Non-pool existing of 16,530,612 is 70% of post, total becomes 23,615,160. The investor takes 4,723,032, the pool goes to 2,361,516, and each founder holds 21.2%, or 42.3% combined, against 47.7% held across the three preferred series and 10.0% in the pool. The founders are now a minority of the equity and a minority of the board.

The lever is the seat count, negotiated before the round rather than after. Keeping the board at three through Series A, or making the independent seat appointable only by unanimous board consent, changes the company's decision-making far more than a half-turn of valuation. The risk of ignoring it is specific: the board hires and fires the chief executive, and founders who are surprised by that fact are usually being told it in the meeting where it happens.

The veto list nobody reads until it bites

Protective provisions live in the certificate of incorporation, and they are the quietest source of real power in a venture-backed company. They are a list of actions the company may not take without the approval of the preferred stock: typically selling the company, amending the charter, authorizing a new senior series, increasing the option pool, taking on debt above a threshold, or paying a dividend.

Delaware supplies a floor beneath whatever the parties negotiate. Section 242 requires a majority of outstanding stock entitled to vote for a charter amendment, and §242(b)(2) additionally requires a separate class vote where the amendment would increase or decrease that class's authorized shares, change its par value, or alter or change its powers, preferences or special rights so as to affect them adversely. Where an amendment adversely affects only some series within a class, only the affected series vote as a separate class. So even a series holding a modest stake has a statutory blocking right over anything that reshapes its own terms, before the negotiated protective provisions are read at all.

Stack three rounds and the arithmetic of consent gets uncomfortable. A seller who needs a charter amendment to close may need the common majority, a majority of preferred voting together, and a separate majority of each series. The NVCA publishes the industry's model Certificate of Incorporation and Voting Agreement precisely so these terms are recognisable across deals. Read them before your first term sheet, because they are the vocabulary the other side is negotiating in.

Drag-along rights are the intended counterweight. Sitting in the voting agreement, they bind minority holders to vote for a sale approved by a defined majority, so a single 2% holder cannot hold up a transaction everyone else wants. But a drag drafted to require preferred consent as part of that defined majority is not really a protection against holdouts. It is a second veto with a friendlier name.

The failure mode is concrete and common: a company with a good acquisition offer discovers that the Series A investor, who is at fund end-of-life and needs a bigger outcome, holds the series vote that gates the charter amendment required to close. Nobody is acting in bad faith. The document simply says what it says. The lever is to know which series can block, alone, before you sign the term sheet that creates it.

Down rounds: where the terms stop being theoretical

Every protective term is dormant until the company raises at a lower price. Then anti-dilution provisions convert the preferred stock at a better ratio, and the choice of formula, which reads as boilerplate on a term sheet, decides how much of the company the common holders lose.

Broad-based weighted average is the standard. The new conversion price is the old price multiplied by (A + B) ÷ (A + C), where A is the fully diluted shares outstanding before the new issuance, B is the number of shares the new money would have bought at the old price, and C is the number of shares actually issued.

Illustrative only, carrying forward the Series A above. The Series A bought 3,673,469 shares at $2.00, and fully diluted shares after that round were 18,367,347. The company later raises $4,000,000 at $1.00 per share, issuing 4,000,000 shares. So A = 18,367,347, B = $4,000,000 ÷ $2.00 = 2,000,000, and C = 4,000,000. The new conversion price is $2.00 x (20,367,347 ÷ 22,367,347) = about $1.821. The conversion ratio becomes $2.00 ÷ $1.821 = about 1.098, so the Series A's 3,673,469 shares now convert into roughly 4,034,000 common, about 361,000 extra shares, taken from the common holders.

Now run the same event under a full ratchet, which resets the conversion price to the new price regardless of how few shares were sold at it. The conversion price becomes $1.00, the ratio 2.0, and the same 3,673,469 preferred shares convert into 7,346,938 common, about 3,673,000 extra shares. Same round, same money, same investor: roughly ten times the transfer away from founders and employees. A full ratchet also means a tiny down-round issuance triggers the full reset, which is why it is rare outside distressed financings.

Pay-to-play sits alongside it and cuts the other way. It converts a preferred holder's shares to common if they do not participate pro rata in the down round, which punishes investors who will not support the company and rewards those who do. Founders should want it; it is one of the few terms that improves the common holders' position in bad weather.

What breaks first in a down round is not the cap table. It is the option pool. Employees holding options struck above the new price own nothing worth exercising, and the repricing or fresh grants needed to keep them create another carve-out from the same shrinking common pool. Model that before you agree to the round, not after the departures start. General education, not legal advice: a real financing deserves a real startup lawyer.

Put it to work

Open your charter and voting agreement beside the cap table. Write down who holds each board seat, which protective provisions exist, and which series can veto alone. Confirm every founder grant has a vesting schedule and a filed 83(b). Model one down round under weighted-average and full-ratchet anti-dilution. Take the gaps to a startup lawyer before the next term sheet.

Sources & references

Linked entries open the named source directly. Entries without a link say exactly what kind of reference they are — and how to check them yourself.

• Delaware General Corporation Law §141 — board of directors
• Delaware General Corporation Law §242 — amendment of certificate of incorporation and class votes
• Alphabet Inc. — Form 10-K for fiscal year 2025 (filed February 5, 2026)
• IRS Form 15620 — Section 83(b) Election (April 2025)
• IRS — Update to the 2024 Publication 525 confirming Form 15620 as a way to make an 83(b) election
• 26 U.S.C. §409A — additional tax on noncompliant deferred compensation
• 17 CFR §230.701 — exemption for compensatory benefit plans
• National Venture Capital Association — model legal documents
• 26 U.S.C. §1202 — qualified small business stock