409A Valuations for AI Startups: What Founders Need to Know
Author: Redwood Valuation Content Team
Published: September 15, 2026
A standard startup guide will tell you to get a 409A valuation before your first option grants and refresh it every 12 months or after a material event. That guidance is directionally right. For many AI companies, it is also incomplete. Valuations can move between funding rounds for reasons that are not captured by revenue alone, including model performance, compute access, proprietary data rights, strategic partnerships, or a shift in investor demand for a particular category of AI company.
The 409A analysis is still a common-stock valuation exercise. AI does not create a separate 409A regime. Instead, it creates a fact pattern where company value may depend heavily on intangible assets, technical milestones, and market signals that need to be understood and documented carefully.
What a 409A Valuation Actually Determines
A 409A valuation establishes the fair market value (FMV) of a private company’s common stock as of a specific valuation date. Section 409A governs nonqualified deferred compensation more broadly, but this article focuses on how common stock FMV supports option strike prices. For a nonqualified stock option or stock appreciation right (SAR) to fall outside Section 409A under the stock-right rules, the exercise price generally may not be less than the FMV of the company’s common stock on the grant date, and the regulation’s other stock-right conditions must also be satisfied.
For tax purposes, the grant date is generally the date when the company completes the corporate action necessary to create the legally binding option right. That action is not complete until the maximum number of shares and minimum exercise price are fixed or determinable. The class of stock and the service provider must also be designated. Backdating a grant to capture an earlier, lower FMV does not solve the problem. The strike price has to be supportable as of the actual grant date.
Why AI Companies Face Particular Complexity
An AI company’s value can change between financing rounds for reasons that are not captured by revenue alone. A major model release, a compute partnership, a proprietary data agreement, or a benchmark result can change the company’s competitive position before the next 12-month valuation cycle expires.
The assets that define an AI company’s value may include model weights, proprietary training datasets, data licensing rights, technical know-how, compute access arrangements, and enterprise deployment relationships. Those facts matter when they change ordinary valuation inputs: a model or product milestone may change revenue expectations; data or IP rights may affect the durability or exclusivity of an advantage; compute arrangements may affect cost, capacity, or capital needs; and unusual business models can make public-company comparisons less informative. The appraiser’s job is to connect the AI-specific fact to the valuation assumption or market evidence it changes rather than treat “AI” itself as a source of value.
When an AI-related development may materially affect value, an appraiser check-in before the annual refresh date can help determine whether the prior valuation remains reasonable.
When Your AI Company Needs a 409A Valuation
Before issuing options, SARs, or other awards that depend on common-stock FMV, the company needs a supportable determination of FMV as of the grant date. Section 409A does not require an independent appraisal in every case; the regulation permits reasonable valuation methods and provides several safe harbors. The practical sequence is to establish and document supportable FMV first, then approve the grants.
When to Reassess an Existing Valuation
After closing a priced funding round. The new transaction data may make the prior valuation no longer reasonable for later grants.
After another development that may materially affect company value. New information can make the prior valuation stale before 12 months.
A 409A valuation is generally usable for up to 12 months from the valuation date. Twelve months is a ceiling, not a guarantee. If new information arises that may materially affect the company’s value, the prior valuation may no longer be reasonable for grant-pricing purposes.
You may have heard of a 90-day rule after a funding round or other material event. That is a common practice convention, not a separate regulatory deadline. The regulatory question is whether the prior valuation still reflects the information available when the company wants to use it.
The following AI-related developments do not automatically require a new 409A valuation. They are situations where the prior valuation may need to be revisited, so they are worth discussing with your appraiser:
Closing a major compute partnership, such as GPU cluster access or dedicated cloud infrastructure at scale
Signing a significant API licensing deal or enterprise foundation model deployment
Releasing model performance results that change the company’s competitive position
Signing a large exclusive training-data agreement
Receiving a strategic investment from an AI infrastructure, cloud, or platform company at a meaningfully different valuation
The practical question is whether the new information changes an input the valuation actually relies on. A development is more likely to matter if it changes expected cash flows, the value or durability of key technology or data rights, cost or capacity assumptions, competitive position, or the market evidence an appraiser would weigh. Routine progress that does not meaningfully change those assumptions may not require a new valuation. If the new information may materially affect value, the prior valuation should be reassessed before later grants. A board decision to keep using the prior valuation does not by itself make that valuation reasonable.
The Three Safe Harbor Methods
For private-company common stock that is not readily tradable on an established securities market, FMV must be determined through the reasonable application of a reasonable valuation method. Treasury Regulation Section 1.409A-1(b)(5)(iv)(B) also provides three methods that can create a rebuttable presumption of reasonableness. If the presumption applies, the IRS can rebut it only by showing that the valuation method or its application was grossly unreasonable.
The three safe harbor methods are commonly described as Method 1, Method 2, and Method 3:
| Method | Regulatory Citation | Key Conditions | Most Common For |
|---|---|---|---|
| Method 1: Independent appraisal | Treas. Reg. Section 1.409A-1(b)(5)(iv)(B)(2)(i) | Appraisal by an independent appraiser meeting §401(a)(28)(C), the Code provision used for nonpublic ESOP stock valuations; no more than 12 months before the relevant transaction | Venture-backed AI companies; later seed, Series A, and beyond |
| Method 2: Formula-based valuation | Treas. Reg. Section 1.409A-1(b)(5)(iv)(B)(2)(ii) | Formula price qualifying under the nonlapse-restriction rules, used consistently for specified transfers to the company and certain >10% shareholders; additional transferability limits apply | Rare for venture-backed AI companies |
| Method 3: Illiquid start-up valuation | Treas. Reg. Section 1.409A-1(b)(5)(iv)(B)(2)(iii) | Reasonable, good-faith written valuation of illiquid start-up stock prepared by a qualified person, with all start-up conditions satisfied | Certain early-stage companies before institutional processes make an independent appraisal preferable |
Board approval of the valuation conclusion is a governance best practice. It is not what creates the Section 409A presumption. The presumption depends on satisfying the applicable regulatory valuation method and applying it reasonably.
Method 3 permits an internal valuation in limited early-stage circumstances. The valuation must be reasonable, made in good faith, and documented in a written report. The corporation or a predecessor cannot have conducted a material trade or business for 10 years or more, and no class of the corporation’s equity securities can be traded on an established securities market. The company also cannot reasonably anticipate a change in control within 90 days or a public offering within 180 days, and the stock cannot be subject to a disqualifying put, call, or other purchase right.
For Method 3, the company must reasonably determine that the person performing the valuation is qualified based on significant knowledge, experience, education, or training. The regulation states that significant experience generally means at least five years of relevant experience. That experience may come from business valuation or appraisal, financial accounting, investment banking, private equity, secured lending, or comparable experience in the company’s industry.
How Appraisers Value an AI Company’s Common Stock
The appraiser’s work has two main parts: estimating the company’s equity value and determining how much of that value belongs to common stock after accounting for preferred-stock rights and preferences. Common stock is typically worth less than the preferred price from the last financing round because preferred stock often carries liquidation preferences, participation rights, anti-dilution protections, and other economic rights that common stock does not have.
For AI companies with a recent priced round, appraisers often use a backsolve, also called the Subject Company Transaction Method, to infer the equity value consistent with the observed preferred-stock price. A preferred-round backsolve commonly uses an Option Pricing Model (OPM) framework to solve for that value.
In plain English, OPM runs total equity value through the company’s capital structure instead of applying a fixed discount to the preferred price. Technically, it treats each share class as a series of call options on total equity value. The strike prices are set at the breakpoints where each class’s marginal payoff changes. In a preferred-round backsolve, the model is calibrated to the observed price of the recently issued preferred stock. OPM is commonly used when exit timing is uncertain, which describes many seed and Series A companies.
The round price is market evidence, not a number to copy mechanically into common stock. Financing terms, investor rights, strategic relationships, secondary components, or post-round developments can affect how much weight the appraiser gives the transaction. In a backsolve, the model incorporates the preferred stock’s rights and preferences rather than treating its price as the value of common stock.
PWERM can be useful when the company has identifiable exit paths that can be assigned supportable probabilities. When there is no recent financing to support a backsolve, the appraiser may rely more heavily on other approaches, including market, income, or cost methods. For an AI company, the available evidence determines the fit: limited comparables can weaken a market approach, early or rapidly changing projections can make an income approach harder to support, and a cost approach may not fully capture value when proprietary technology, data rights, or know-how create economic value that historical development cost does not measure.
The Discount for Lack of Marketability (DLOM) reflects the fact that private-company common stock generally cannot be sold as readily as public-company stock. It is a liquidity adjustment, not a penalty. The magnitude depends on company stage, expected time to liquidity, transfer restrictions, volatility, secondary-market activity, and the valuation methodology being used. For early-stage AI companies with limited secondary activity and longer expected times to liquidity, the DLOM may be meaningful. It should still be developed from the facts rather than applied as a formula.
In a preferred-stock backsolve, a DLOM is often applied to the resulting common-stock value. The appraiser should still consider whether the transaction evidence already reflects some illiquidity so the same effect is not counted twice. The double-counting risk is clearest when the reference transaction is a common-stock secondary sale, where illiquidity may already be reflected in the observed price.
Common stock therefore does not need to match the preferred share price. The difference usually reflects the rights and preferences in the capital structure, the allocation method used, and the marketability adjustment applied to common stock.
Penalties When Options Are Granted Below FMV
If options are granted with a strike price below FMV on the grant date, the person who received those options can face the Section 409A tax consequences. The primary taxpayer is the employee, advisor, or contractor who received the award, not the company.
Principal Consequences for the Service Provider
Income inclusion. Amounts deferred under the noncompliant arrangement may be included in income to the extent they are not subject to a substantial risk of forfeiture and have not previously been included.
20% additional income tax. Section 409A imposes an additional 20% income tax on amounts required to be included in income.
Premium interest. Section 409A also imposes an interest-based additional tax. It is calculated using the underpayment rate plus one percentage point on hypothetical underpayments dating back to the year of initial deferral or, if later, the first year the amount was no longer subject to a substantial risk of forfeiture.
The tax calculation is not simply the grant-date discount multiplied by the number of options. For a discounted option that becomes subject to Section 409A, the taxable amount can depend on later FMV, vesting, exercise, and year-end measurement rules. Tax counsel should be involved if a potential discounted-option issue is discovered.
Although the service provider is the primary taxpayer, the company can still face reporting and withholding obligations, audit and cap-table cleanup, and employee-relations issues. If key employees receive unexpected tax bills because grant prices were wrong, the company may still have to manage the consequences.
Selecting an Appraiser for an AI Company
Even if the company qualifies for Method 3, an independent appraisal may become more useful as the company adds institutional investors, its capitalization becomes more complex, or it anticipates a financing, acquisition, or audit. The provider decision should therefore reflect the company’s expected review needs as well as the regulatory route it plans to use.
Sector experience matters for AI companies because the valuation may depend on facts that are not captured by generic SaaS comparables. A valuation provider does not need to be an AI engineer. The provider should understand how to ask the right questions about model performance, training data rights, IP protection, compute commitments, technical milestones, and commercial traction.
Common valuation credentials include Accredited Senior Appraiser (ASA), Accredited in Business Valuation (ABV), and Chartered Financial Analyst (CFA). Credentials can be useful, but they are only one way to evaluate competence. The more practical question is whether the valuation provider has the experience, independence, documentation discipline, and judgment needed for the company’s stage and complexity.
Useful selection criteria for AI companies include:
Startup valuation experience, including AI, deep-tech, or IP-heavy companies
A process for evaluating technical milestones and intangible assets without overclaiming their value
Familiarity with complex capitalization structures and preferred-stock rights
Willingness to explain methodology rather than merely deliver a report
Experience with audit-firm review if the company expects financial statement scrutiny
An independent appraiser meeting the safe-harbor requirements if the company plans to rely on the independent appraisal method
What to Expect From the Process
Most of the requested information is straightforward if the cap table and corporate records are organized. Delays can arise when projections, recent cap table changes, intellectual-property or data rights, compute contracts, or technical milestones require additional documentation or clarification.
Information the Appraiser Typically Requests
Three or more years of financial statements, or statements since inception if the company is newer
Business plan and financial projections
Fully diluted cap table
Prior valuation reports, if any
Corporate documents, including certificate of incorporation and investment agreements
List of outstanding grants and their terms
Recent business update and key milestones
For AI companies specifically: IP documentation, data licensing agreements, compute contract terms, and model performance or product milestone support
At the end of a provider-led appraisal process, the company typically receives a written valuation report stating the appraiser’s concluded FMV of common stock as of the valuation date. The board typically relies on that value when approving option grants unless later information may materially affect value or the valuation becomes more than 12 months old.
What Founders Should Take Away
AI does not change the Section 409A rules, but it can change the valuation facts faster than a calendar-based refresh cycle suggests. The key is to connect AI-specific developments—model performance, data and IP rights, compute access, commercial traction, strategic relationships, and financing evidence—to the valuation assumptions they actually change. Establish a supportable FMV before grants, and revisit the valuation when a development may materially change those assumptions rather than waiting mechanically for 12 months to pass.
About Redwood Valuation
Redwood Valuation provides 409A valuations for private companies at all stages, including AI companies with complex intangible asset bases, data-rights issues, compute arrangements, and rapid valuation changes. If your company is approaching its first equity grants, closing a financing round, or evaluating whether a recent AI milestone affects your current valuation, contact Redwood to discuss the right valuation approach.