Education
409A Valuations for AI Startups: Why a Record Round Doesn't Set Your Strike Price
AI startups raise at record valuations, but a huge round doesn't set your strike price. Why your 409A common-stock value comes in far lower, and what to do.
By 409.AI Team - 2026-08-05
# 409A Valuations for AI Startups: Why a Record Round Doesn't Set Your Strike Price
The money going into AI is hard to overstate. Global venture funding hit a record $510 billion in the first half of 2026, and AI took the lion's share of it. Total capital invested in AI reached roughly $256 billion in 2025, close to three times the 2024 figure, according to Crunchbase. Seed-stage AI companies are now raising at valuations about 40% higher than comparable non-AI startups, and Series B AI rounds clear a median near $143 million.
If you're a founder riding that wave, here's the part that trips people up. You just closed a round that values your company at, say, $400 million. Then your 409A report comes back and prices your common stock at $3.10 a share, a small fraction of what your investors just paid. Your first instinct is that someone botched the math. Nobody did. The distance between what a VC pays for preferred stock and what your employees pay to exercise an option is not a rounding error, and for AI companies it tends to be wider than almost anywhere else in the market. Understanding why saves you real grief at grant time, at audit, and at exit.
The round price and the strike price are two different numbers
Investors in a priced round buy preferred stock. Your employees get options on common stock. Those are not the same security, and they aren't worth the same amount per share.
Preferred stock carries rights that common stock doesn't: a liquidation preference that pays investors back first, anti-dilution protection, sometimes a dividend, often a board seat. In a modest exit, preferred holders can take most or all of the proceeds before common sees a dollar. A 409A valuation exists to answer a narrower question than "what's the company worth." It asks what a single share of *common* stock is worth today, on its own terms, with no market to sell into. That number sets the minimum strike price you can grant options at under IRC Section 409A. If you'd rather start with the basics, our [comprehensive 409A guide](https://409.ai/articles/what-is-a-409a-valuation-a-comprehensive-guide) walks through the whole framework.
So the preferred price your investors paid is the ceiling, not the answer. The appraiser strips out the preferences, applies a discount for the fact that the stock can't be sold, and lands on a common value well below the headline. We've written a full explainer on exactly [why your 409A comes in lower than your post-money](https://409.ai/articles/why-is-your-409a-valuation-lower-than-post-money-valuation), and for AI companies the spread is usually larger, for reasons worth unpacking.
The model has to guess how volatile you are
Most 409A appraisals of a venture-backed company use an option pricing model, or OPM, to split the company's total equity value across preferred and common. The OPM treats each share class as a series of call options on the company's value and prices them with a Black-Scholes framework. To do that, it needs two inputs that are more art than science: the expected time to a liquidity event, and volatility. Our piece on [OPM, PWERM, and the backsolve method](https://409.ai/articles/409a-allocation-methods-opm-pwerm-backsolve) covers how appraisers pick between these approaches.
Volatility is where AI companies get interesting. Appraisers derive it from the historical volatility of comparable public companies. If your comp set is a basket of stable enterprise SaaS names, you get a moderate volatility figure. If your comps are the handful of public AI and semiconductor stocks that swing 60% or more a year, your volatility input climbs, and higher volatility pushes more value toward the common stock's upside optionality. Two AI startups with identical cap tables can land on different common values purely because their appraisers chose different comparable companies. That's not a flaw you can eliminate. It's a judgment call you should understand and be able to defend.
What makes AI valuations genuinely harder
AI revenue multiples are running anywhere from 10x to 50x, with a median somewhere around 20x to 30x. Those multiples exist because investors are pricing a future, not a present. Plenty of the most richly funded AI companies have thin revenue, or none, relative to what they've raised. That breaks the two valuation approaches an appraiser would normally lean on.
The income approach discounts projected cash flows. When a company has burned through a nine-figure raise on compute and research talent with revenue still ramping, those projections are so speculative that the model's output swings on assumptions nobody can pin down. The market approach compares you to transactions and public multiples, but the public AI comp set is small, expensive, and volatile, so it produces a wide range rather than a tight number. Much of an AI company's value also sits in things that resist clean measurement: model weights, proprietary training data, and a research team that could walk out the door. Appraisers can't ignore those assets, and they can't precisely quantify them either.
Then there's your cap table. Many AI startups raise fast, stacking SAFEs and convertible notes ahead of and between priced rounds. Those instruments don't just sit quietly until they convert. A SAFE with a $200 million cap changes the math the moment your appraiser has to decide how and when it becomes equity, and it can pull your common value in either direction depending on the terms. We've broken down [how SAFEs affect your 409A](https://409.ai/articles/how-safes-affect-your-409a-valuation) and [what convertible notes do to your strike price](https://409.ai/articles/convertible-notes-effect-on-409a-valuation) in detail, because the effect surprises founders more often than not.
Raise fast, revalue fast
A 409A valuation is good for a maximum of twelve months, but it expires the instant a material event occurs, whichever comes first. That's a core condition of the IRS safe harbor: an independent appraisal buys you a presumption of reasonableness, meaning the IRS would have to prove your strike price was grossly unreasonable rather than you having to prove it was right. Lose the safe harbor and that burden flips back onto you and your employees.
A new priced round is the textbook material event. So is a major model release that changes your commercial trajectory, a large customer win or loss, or a pivot. AI companies hit these triggers more often than most, simply because they're moving faster and raising more frequently. If you close a Series A in March and a Series B in September, one 409A won't cover both grant windows. Our guide to [how often you actually need a 409A](https://409.ai/articles/409a-valuation-frequency-how-often-should-you-get-one) lays out what counts as material and what doesn't. The practical takeaway: grant options against a stale valuation and you're pricing them on numbers the IRS no longer considers current.
When the boom cools
Valuations this stretched invite an obvious question. What happens to your 409A if AI sentiment turns and your next round comes in flat or down?
The common value moves with it. A down round usually produces a lower 409A, which is fine for new grants but painful for existing option holders whose strike prices were set at the peak. Those options can end up underwater, worth less to exercise than they cost. There are real, compliant ways to respond, from repricing to exchange programs, and there are wrong ways that create fresh 409A problems. We walk through the tradeoffs in our piece on [down rounds and underwater options](https://409.ai/articles/down-round-409a-underwater-options-repricing). Planning for the downside while you're raising at a premium isn't pessimism. It's the difference between a clean repricing later and a scramble.
What founders should actually do
Use a qualified independent appraiser, not a spreadsheet you built the night before a grant. The safe harbor is the whole point, and it only attaches to a defensible, independent valuation. Keep board minutes that show the valuation was reviewed and relied on when options were approved.
Revalue on every material event, not just annually. If you're raising every six to nine months, budget for a 409A on that cadence too. Be ready to explain your appraiser's comp set and volatility inputs, because those are the assumptions an auditor or the IRS will probe first for an AI company. And if an IPO is anywhere on your horizon, get disciplined early: the SEC scrutinizes pre-IPO option grants for signs of underpriced "cheap stock," and AI companies with steep valuation ramps draw exactly that kind of attention. Our post on [cheap stock and the pre-IPO 409A](https://409.ai/articles/cheap-stock-pre-ipo-409a-sec-option-grants) explains what that review looks for.
None of this requires you to slow down. It requires you to treat the 409A as a live document that keeps pace with how fast you're moving. A record round is a great problem to have. Just don't let it convince you that your common stock is worth what your investors paid, and don't let a stale valuation turn a routine option grant into a tax mess for the people building the thing. If you're ready for an expert-reviewed, audit-defensible number, [409.ai's 409A valuation](https://409.ai/products/409a) is built for exactly this pace.
*Sources: Crunchbase News venture funding data (H1 2026 and 2025); IRC Section 409A and Treasury regulations on the independent-appraisal safe harbor; AICPA Accounting and Valuation Guide, Valuation of Privately-Held-Company Equity Securities Issued as Compensation. This article is educational and not legal, tax, or accounting advice.*