Salary Cash Tradeoffs at Series A Versus Series B AI Startups and What They Reveal About Your Total Offer
A Series A or B offer prices risk, so compare ownership, dilution and role.
An ML engineering offer at a Series A or Series B AI startup is a risk-sharing agreement. Series A usually offers more ownership for accepting more uncertainty, while Series B offers more cash and a smaller stake in a more-proven business. The right comparison turns on functional ownership, dilution, strike price and the probability of getting paid.
Why the cash-equity split at Series A and Series B encodes risk transfer, not just compensation
Cash rises slowly as startups mature, while equity percentages compress fast. Late-stage startups pay mid-level engineers roughly 15–18% more cash than early-stage companies do, but the share of ownership falls much faster from one stage to the next. The company is asking the engineer to exchange some dependable pay for a claim on an uncertain outcome. How much ownership it offers shows how it prices that exchange.
Startups almost never match big-tech cash, so ownership is where much of their competitive argument lives. Equity is the proposed payment for accepting business risk and waiting for a possible return. The split also reveals the company's position on valuation, because it asks you to accept today's price as a credible starting point for future growth. Treating only salary as negotiable leaves the largest uncertainty in the offer unexamined.
What the numbers actually look like at Series A versus Series B for engineers
The 2026 SF AI Startup Pay Guide places Series A AI engineers in the SF market at $200K–$280K base salary plus 0.15%–0.5% equity. The guide gives Series A backend engineers 0.1%–0.35% equity and frontend engineers 0.1%–0.3%, with base salaries described as competitive market ranges. For comparison, its Seed AI engineer benchmark is competitive base pay plus 0.5%–1.5% equity.
The guide also gives a Series A Head of Sales, a functional owner, roughly 0.5%–1.5% equity with competitive base pay. That row describes a different kind of responsibility. Across the Series A-to-Series B comparison, the broad pattern is that cash rises by roughly one bracket while equity percentages are cut roughly in half. The available figures do not establish a complete Series B salary range for every engineering role.
Pond AI's research finds that equity has split into two distinct markets by Series A. Senior individual contributors move toward a narrower band, while first functional leaders still receive meaningfully larger grants. The research draws on Carta data across thousands of grants and finds that median initial equity falls sharply from the first hire to the fifth. An engineer who joins at headcount 40 is in a different position from one who joins at headcount 8, even when both offers say Series A.
How dilution silently cuts the equity percentage you were actually granted
The percentage in an offer describes a starting position. Each later funding round issues new shares, so an existing grant becomes a smaller share of the total even if the engineer's nominal grant does not change. A 0.5% grant at Series A can therefore become materially smaller after Series B and Series C. Nothing has to disappear from the grant itself. The denominator grows, and the ownership percentage shrinks.
Option pool refreshes add another source of dilution. At Series A, investors commonly require the company to top up its option pool before the round closes, diluting existing holders, including an engineer who has just joined. Round dilution and pool dilution can compound before any exit. The engineer therefore needs to ask how large the pool is and when it was last refreshed. Without that context, the offer's percentage can look more durable than it is.
A Series B grant starts smaller but generally has fewer funding rounds ahead of it than a Series A grant. To compare them, an engineer needs to ask how many rounds the company expects before an exit and model the resulting dilution. That exercise explicitly compares promised ownership today with possible ownership at liquidity.
What the stage of a company signals about valuation confidence and the probability your equity pays off
Series A asks the engineer to believe in a less-proven thesis at a lower entry price. Series B offers more proof at a higher price. Cooley's Q1 2026 Venture Financing Report, together with PitchBook-NVCA and Carta data, places Series B median valuations substantially above Series A medians. With that higher valuation, a smaller Series B percentage can still be worth as much or more on paper. The percentage alone leaves you with no way to tell how much growth the offer already assumes.
AI startups also receive a valuation premium over non-AI peers, and the funding environment has improved the prospect of a meaningful exit for well-capitalized AI companies. But a premium is not a payout. A higher valuation makes a stake worth more on paper and raises the price at which the engineer joins. The decision therefore depends on the specific company's trajectory. More confidence and a better bargain are not the same thing.
Timing within that trajectory can be as important as the stage label. Someone who joins before a sharp rise in valuation gets growth that someone who joins afterward has already missed. OpenAI shows how much timing matters, though its pay is a poor benchmark for most Series A or Series B jobs. In the past, early-stage equity's low chance of paying out much has often made it hard to justify taking much less cash. Well-funded AI companies now have a better chance of an exit, which changes that trade-off, but paper equity still needs a discount.
How your role title and functional ownership position determine which benchmark table applies to you
Stage does not decide which compensation row belongs to you. The company's expectations come first: execution inside an existing function or ownership of building that function. Two similar engineering titles can hide that difference in pay and ownership. Pond AI finds that Series A individual-contributor grants compress toward a lower band, while first functional leaders can still command 0.5% and above. Grants can vary more by functional ownership than by stage at the same level.
For an ML engineer evaluating an offer, the practical test is what the role owns. Does the job mean building the core AI capability that makes the company different, or working within a capability someone else owns? Is the grant paying for foundational intellectual property or another engineering hire? Those questions locate the benchmark more reliably than seniority alone. A founding or functional-owner role should be compared with the benchmarks for those roles instead of the ordinary individual-contributor range.
Specialization also changes the comparison within a stage. AI engineers have higher cash floors and can get larger equity grants than backend generalists because a few builders with specialized skills can create the company's technical moat. Founding AI/ML engineers form a distinct category at both Series A and Series B, with grants meaningfully above senior individual contributors. But the title alone does not settle the comparison. The relevant distinction is whether the engineer is creating the core capability and taking responsibility for it, which is what the higher ownership grant compensates.
How do you evaluate and negotiate a Series A or Series B ML engineering offer?
The lesson for negotiation is to price the risk the company is asking you to accept. At Series A, greater uncertainty needs to be matched by more equity or more cash. Low cash paired with a mediocre equity grant is risk without enough compensation. A functional owner offered an ordinary individual-contributor grant has a concrete argument for more ownership based on the scope of the role.
At Series B, pushing for cash becomes more defensible because the company offers less remaining upside and should pay more for an equivalent role. A Series B offer that stays at a Series A cash floor deserves scrutiny against the later-stage market. Strike price is the amount paid to exercise an option, and it changes the option's effective value even when the ownership percentage stays the same.
At Seed and Series A, incentive stock options typically have a low strike price, with a meaningful gap between strike and current valuation. By Series B and later, new grants are set against a much higher 409A valuation. That reduces the effective value of the option at the same percentage. Therefore, negotiate not just how much ownership is offered, but at what exercise cost, against which valuation and with how much dilution still ahead.
What employee number would the engineer be, and what is the engineer's equity percentage on a fully diluted basis? What is the strike price, and what is the latest valuation? How much runway remains, and when is the next round expected? How large is the option pool, and when was it last refreshed? Pond AI's research identifies these as questions that move the money because they reveal the position behind the headline grant.
A cash discount can be rational when the grant meets or exceeds the functional-owner benchmark, fewer rounds are expected before exit and the valuation premium is credible. The reverse is an offer at or below the individual-contributor equity floor, a recently refreshed pool and no clear answer on runway. A refresh establishes dilution context that cannot be ignored. Combined with weak ownership and unclear financing prospects, it leaves the engineer absorbing risk without commensurate upside.
The comparison against liquid compensation that most engineers get wrong
A startup offer and a liquid compensation package cannot be compared by adding up their headline totals. Big-tech RSUs or stock in a later-stage public company are on the liquid side of that comparison, while startup equity needs a large discount to its paper value. Start with base against base and cash against cash. If a Series A offer shows $200K base plus a large paper equity number, that number does not erase the cash gap. Its contribution depends on the probability and size of an eventual payout.
The opposite mistake is to treat all startup ownership as worthless. The AI valuation premium and improved exit prospects for well-capitalized companies give the upside a basis for evaluation, but not a guarantee. Expected value is the useful frame. It allows the engineer to consider ownership after dilution, the option's strike price and the chance of liquidity rather than taking the current valuation at face value. The cash discount is then a visible cost of choosing that uncertain return.
Someone with high risk tolerance, a founding or functional-owner role and confidence in the company's trajectory may rationally accept less cash at Series A. Someone with less conviction who is optimizing for expected value should seek to close the cash gap at Series B. Neither stage wins by label alone. The stronger offer is the one whose cash, ownership and terms pay adequately for the risk the engineer is actually taking.