From Shubhi K | Product & Market Analysis

Down Rounds Are Coming: How to Read a 2026 AI Cap Table

On this page

A headline valuation tells you almost nothing about what your equity is worth. Liquidation preferences, participation rights and anti-dilution ratchets decide who gets paid and in what order, and those terms are reappearing in rounds priced at multiples that make a flat outcome unlikely. If you hold equity in an AI company, the term sheet matters more than the number in the press release.

Key takeaways

  • Valuation and structure are traded against each other. Founders can usually have a higher headline number or cleaner terms, and rarely both.
  • A 1x non-participating preference is standard. Anything beyond that, including multiples or participation, materially changes common shareholder outcomes.
  • Ratchets transfer dilution to everyone else. Full-ratchet anti-dilution reprices an earlier investor's stake as if they had invested at the new lower price.
  • Employees are affected most and informed least. Option holders sit at the bottom of the payout order and rarely see the terms above them.
1xThe standard non-participating liquidation preference. Anything above it shifts outcomes materially toward preferred holders.
61%Share of global venture capital that went to AI in 2025, which is the condition that produced aggressive pricing.
Last inWhere common shareholders and option holders sit in the payout order behind every preference.

Why structure returns when valuations get stretched

Investors and founders negotiate two things that trade against each other. The headline valuation, and the terms attached to it.

A founder who insists on a high number gives the investor a reason to ask for protection, because the investor is underwriting a price they may not believe. That protection arrives as structure.

This is why aggressive pricing and aggressive terms travel together. The press release reports the first and almost never the second, which is how a company can announce a record valuation that leaves its employees worse off than a lower, cleaner one would have.

The conditions producing this are well documented. AI startups took roughly 61% of global venture capital in 2025, and the resulting competition pushed prices to levels where investors sought downside protection, a dynamic covered in the analysis of venture concentration.

Liquidation preference, explained without jargon

A liquidation preference determines who gets paid first when the company is sold, and how much they take before anyone else receives anything.

1x non-participating is the standard. The investor takes back their investment, or converts to common shares and takes their percentage, whichever is greater. They choose one, not both.

Participating means they take their money back and then also share in what remains according to their ownership. That is two bites, and it materially reduces what common shareholders receive at most exit values.

Multiples mean the preference is 2x or 3x the investment before anyone else is paid. At a modest exit, a 2x preference on a large round can consume the entire proceeds.

Four preference structures, same $100M invested Illustrative effect on common shareholders at a $150M exit Preferred takes Common receives Fairness 1x non-participating $100M $50M Standard 1x participating $100M plus share Reduced Aggressive 2x non-participating $150M Nothing Punitive 2x participating $150M plus share Nothing Punitive
Rows three and four produce identical outcomes for common holders at this exit value. Both are zero.

Ratchets and anti-dilution

Anti-dilution provisions protect an earlier investor if a later round prices lower. They vary enormously in severity.

Broad-based weighted average is the standard and the mildest. It adjusts the earlier investor's conversion price partially, taking into account how much new money came in.

Full ratchet reprices the earlier investment entirely as though it had been made at the new lower price. If someone invested at $50 per share and a later round prices at $10, their stake is recalculated at $10.

The dilution has to come from somewhere, and it comes from common shareholders. A full ratchet in a down round can move a founding team's ownership by a startling amount in a single financing.

Why this matters more in 2026 than it did in 2023

Ratchets only activate on a down round. In a market where every round priced up, they were dormant clauses nobody thought about.

The moment valuations reset, dormant clauses become active ones, and companies that agreed to them during the competitive peak discover what they signed.

What a payout actually looks like

The payout order, called the waterfall, is where all of this becomes concrete.

A $150M exit with $100M of 1x participating preferred Illustrative distribution, in $ millions $150.0B Exit proceeds -$100.0B Preference repaid $50.0B Left for everyone Illustrative example. Real waterfalls include multiple rounds, each with its own terms and seniority.
Then the participating preferred also shares in the final bar, alongside common and option holders.

In a real company there are several rounds, each with its own preference, and later rounds are usually senior to earlier ones. The most recent investor is paid first, then the one before, and so on down.

Common shareholders and option holders are last. Everything above them is paid in full before they receive anything, which is why a company can sell for a large number and leave employees with nothing.

Pay-to-play and the terms nobody discusses

Two further provisions deserve naming because they surface specifically in difficult financings.

Pay-to-play requires existing investors to participate in a new round or lose their preferred status, converting to common. It is aimed at investors rather than founders, and it can clean up a cap table quickly. It also means an investor unable to participate loses protection they paid for.

Redemption rights allow an investor to require the company to buy back their shares after a defined period. They are rare and they matter enormously when present, because they convert an equity investment into something closer to a debt obligation with a date attached.

Neither appears in a funding announcement. Both are visible in the documents, and both change what a headline valuation means for everyone below the preferred stack.

Questions to ask, whichever seat you are in

If you areAsk
An employee with optionsWhat is the total liquidation preference stack, and at what exit value does common start receiving proceeds?
A founder raisingWhat would this term do at a flat exit and at half the current valuation? Model both before signing.
An early investorIs my anti-dilution broad-based weighted average or full ratchet, and who bears the adjustment?
Joining a companyWhat is the preference overhang relative to the current valuation, and is the option strike recent?

The first row is the important one and the one least often asked. The number is knowable, companies can disclose it without revealing anything commercially sensitive, and many will if asked directly.

The fourth row matters most at the moment of joining, when you have the most leverage and the least information. A candidate who asks about preference overhang before signing is asking a normal question, and the reaction to it tells you something about the company independent of the answer.

Why employees are the most exposed

Option holders sit at the bottom of the payout order and typically have the least information about what sits above them.

They also face a second problem specific to periods of valuation reset. An option granted at a strike price set during a peak valuation can be worth nothing at an exit that would still be a good outcome for the company, because the strike exceeds what common shares actually receive.

Some companies address this by repricing options after a down round. Many do not, because repricing requires board approval, creates accounting consequences and dilutes existing holders further.

The practical consequence is a retention problem that arrives eighteen months after the financing rather than at the time of it. Staff who understood their equity as a meaningful part of compensation discover it is not, generally when a competitor makes an offer.

None of this is hidden and almost none of it is explained at the point of hiring. The information asymmetry is the problem rather than the terms themselves.

The signal a clean term sheet sends

Structure works in both directions as information. An investor willing to price at a lower number with a standard 1x non-participating preference is expressing more confidence than one paying a high number with protection attached.

Founders optimising purely for the headline should sit with that. The cleaner deal frequently reflects a higher genuine belief in the outcome, and it is the one that leaves your team whole across a wider range of exits.

Where this framing is unfair to investors

Three points in fairness.

Structure is not predatory by default. An investor asked to pay a price they consider high is entitled to seek downside protection, and structure is the mechanism for that. A founder who demanded the high number chose that trade.

Most rounds remain clean. The standard 1x non-participating preference is still the most common outcome, and aggressive terms cluster in specific situations rather than being universal.

And structure sometimes saves companies. A round that could only be completed with protective terms is still a round, and the alternative to a structured financing is frequently no financing at all. Employees of a company that survived on structured terms are better off than employees of one that did not survive.

The problem is not that structure exists. It is that headline valuations are reported as though it does not, which leaves the people with the least information holding the most exposed position.

A modest reporting change would fix most of this. Announcing a valuation alongside the preference multiple and whether it participates would take one additional sentence and would let readers interpret the number. Nothing prevents it except convention, and the convention benefits the party writing the press release.

Frequently asked questions

What is a liquidation preference?

It determines who gets paid first when a company is sold and how much they receive before anyone else. A 1x non-participating preference, the market standard, means the investor takes back their investment or converts to common shares and takes their percentage, whichever is greater. They choose one option, not both.

What does participating preferred mean?

The investor takes their money back first and then also shares in whatever remains, according to their ownership percentage. That is effectively two payouts from the same exit. It materially reduces what common shareholders and option holders receive, and the effect is largest at modest exit values.

What is a full ratchet?

A severe form of anti-dilution protection. If a later round prices lower, the earlier investment is repriced entirely as though it had been made at the new lower price. Someone who invested at $50 per share has their stake recalculated at $10 if a later round prices there. The dilution comes from common shareholders.

Why does a high valuation not always help employees?

Because valuation and terms are traded against each other. A founder insisting on a high headline number gives the investor a reason to seek downside protection, which arrives as structure. A record valuation with aggressive preferences can leave option holders worse off than a lower valuation with clean terms would have.

What should I ask about my stock options?

Ask for the total liquidation preference stack and the exit value at which common shareholders begin receiving proceeds. That number is knowable, it does not reveal anything commercially sensitive, and many companies will disclose it if asked directly. It is the single most useful fact about what your options are worth.

Are structured terms always bad for founders?

No. An investor asked to pay a price they consider high is entitled to downside protection, and a founder who pushed for the high number chose that trade. A round that could only be completed with protective terms is still a round, and the alternative is often no financing at all.

Where to start this week

One question, sent in writing.

If you hold options, ask your company what the total liquidation preference stack is and at what exit value common shareholders begin receiving proceeds. Phrase it neutrally and expect a straight answer, because it is a reasonable question with no confidential content.

If the answer is evasive, that is information too. Companies with clean structures generally answer this immediately, because the number reflects well on them.

If you are a founder rather than an option holder, do the modelling exercise instead. Take your current term sheet, or your last one, and calculate what common shareholders receive at three exit values: your current valuation, half of it, and twice it.

Most founders have modelled the third scenario and neither of the first two. The middle case is the one that determines whether your team is motivated in eighteen months, and it takes an afternoon in a spreadsheet to answer properly.

Whatever the answer, write it down and keep it. A cap table is a document that gets harder to reconstruct with every round, and the version of you negotiating the next term sheet will be grateful for a model that already exists.

References

  1. Crunchbase and published venture funding analyses for 2025, used for the AI share of global venture capital and the competitive pricing context.
  2. Long Angle, Software vs AI Q1 2026. Used for the market conditions driving valuation resets and downstream private mark pressure.
  3. CB Insights, State of AI Q1 2026. Used for AI funding volumes and round activity.

This post explains commonly used financing terms in general form. It is not legal or financial advice, term sheets vary considerably, and anyone evaluating a specific agreement should take professional advice on their own position.

SK
Shubhi K
Founding Member, Zan Digital. Writes about AI product economics, B2B software markets and what the numbers behind vendor claims actually say.

Related reading