Stock-Based Compensation in Tech IPOs

The most important line in a tech IPO prospectus is not the revenue growth rate. It is the gap between the GAAP loss at the top of the income statement and the non-GAAP profit management wants you to focus on. Stock-Based Compensation in Tech IPOs is almost always the single largest item bridging those two numbers.

The mistake is treating that item as a freebie. The label “non-cash” makes it sound costless. It is not. Stock-based compensation is a real cost, paid in dilution rather than dollars, and the right way to read an S-1 is to fold it back in and watch share count grow, not just admire the adjusted-margin slide.

This piece reads that claim through public filings, not press releases. Every company figure below ties to a specific SEC form and fiscal period. The framing is analytical: how to think about SBC as a cost structure, not what to do about any stock.

Key takeaways

• SBC is the bridge from GAAP loss to non-GAAP profit. For most newly public software companies it is the largest single add-back in the reconciliation table.
• Figma’s combined SBC charges hit roughly 129% of 2024 revenue, including a one-time RSU release, while reported revenue was $749M (Figma Inc. Form S-1/A, 2024).
• Snowflake spent $1.479B on SBC in fiscal 2025, about 43% of revenue (Snowflake Inc. Form 10-K, FY2025). At scale, Alphabet ran near 6% and Meta near 5% (Forms 10-K, FY2025).
• The cost is dilution, not cash. New shares issued to employees shrink each existing holder’s claim on future cash flow. That is the line non-GAAP margin hides.
• Nvidia stopped excluding SBC from non-GAAP effective Q1 fiscal 2027 (Nvidia Form 10-K, FY2025; company communications, 2026), a signal that the add-back convention is losing credibility.
• The operator read is two-step: strip SBC to see cash margin, then track diluted share-count growth separately to price the real economic transfer.

Why SBC matters at IPO: the GAAP-to-non-GAAP bridge

At IPO, stock-based compensation is the largest item separating a company’s GAAP loss from the non-GAAP profit it leads with. It reduces reported operating and net income under GAAP, but management adds it back in non-GAAP measures to show “underlying” profitability before that charge.

That bridge is where the narrative gets built. A company can post a deep GAAP operating loss and a clean non-GAAP operating profit, and the entire distance between them is often one line: SBC. The reader’s job is to decide whether that line is a genuine non-recurring distortion or a permanent feature of how the company pays its people.

For growth software, it is almost always the latter. Equity is not a one-time grant; it is the recurring currency these companies use to hire and retain engineers and salespeople. Treating a recurring cost as a one-time adjustment is how an unprofitable business starts looking profitable on a slide.

This is the same discipline that separates a durable model from a fragile one in why gross margin is destiny in SaaS: the headline number is only useful once you know what sits underneath it.

What is stock-based compensation? The accounting basics

Stock-based compensation is the expense a company records when it pays employees in equity, such as restricted stock units (RSUs) or options, instead of cash. Under U.S. GAAP (ASC 718), the grant-date fair value is recognized as a non-cash expense over the vesting period and allocated across cost of revenue, R&D, and SG&A.

So it is “non-cash” in a narrow, literal sense: no money leaves the bank when the expense is booked. But the company gave up something real, an ownership stake, to obtain the labor. The cash the company conserved by paying in stock is exactly the cash a shareholder’s future claim gets diluted to fund.

A few mechanics matter for reading filings:

• It hits the income statement, not just a footnote. SBC flows through operating expenses and reduces GAAP operating income and net income.
• It is spread across functions. The same grant shows up partly in COGS (which pressures gross margin), partly in R&D, and partly in SG&A.
• It is recurring for growth companies. New hires get new grants every year, so the expense renews rather than rolling off.

That last point is the whole tension. The accounting treats SBC as an expense; non-GAAP treats it as removable; the cap table treats it as permanent. All three are looking at the same grant.

How does SBC show up as a percentage of revenue across tech companies?

SBC intensity varies enormously by stage. Recently public and pre-IPO companies often run SBC at 20% to 40%-plus of revenue, while mature mega-caps run it in the low single digits because their revenue base dwarfs the grant pool. The percentage, not the dollar amount, is what reveals how reliant a business is on equity to operate.

Here is a cited snapshot across stages. Read it as cost intensity, not a ranking of quality.

Company	SBC expense	Period	SBC as % of revenue	Source
Figma	~$1.84B combined charges*	FY2024	~129% of $749M revenue	Figma Form S-1/A
Snowflake	$1.479B	FY2025 (ended Jan 31, 2025)	~43%	Snowflake Form 10-K
Alphabet	$24.95B	FY2025 (ended Dec 31, 2025)	~6%	Alphabet Form 10-K
Meta Platforms	$20.43B	FY2025 (ended Dec 31, 2025)	~5%	Meta Form 10-K
Nvidia	~$4.7B	FY2025 (ended Jan 26, 2025)	low single digits	Nvidia Form 10-K

*Figma’s 2024 charges combine $947.5M of ongoing SBC with a one-time $889.3M expense tied to its May 2024 RSU release (Figma Form S-1/A). The combined figure exceeds reported revenue, which is why the percentage reads above 100%.

Sources: Figma Inc. Form S-1/A (2024 disclosure); Snowflake Inc. Form 10-K, FY2025; Alphabet Inc. Form 10-K, FY2025; Meta Platforms Inc. Form 10-K, FY2025; Nvidia Corporation Form 10-K, FY2025.

The pattern is structural. At IPO, a company is still building its revenue base while paying competitive equity to attract talent, so the ratio is high. At Alphabet and Meta scale, the same dollar value of grants disappears into a revenue line measured in the hundreds of billions. The IPO investor is buying the company at the expensive end of that curve and underwriting the path down it.

The SBC Adjustment Ladder: from GAAP loss to non-GAAP profit

To read a reconciliation table without getting played, climb it one rung at a time. Call this framework the SBC Adjustment Ladder: a step-by-step view of how a company gets from a GAAP loss at the bottom to a non-GAAP profit at the top, with SBC almost always the tallest rung, plus what to verify at each step. It is an original analytical asset; the mechanics are general and the figures used to illustrate it are cited to filings.

Rung	Step in the reconciliation	What it does	What to check
1	GAAP operating loss/income	The honest starting point	Is the company profitable before any add-backs? Usually no at IPO
2	+ Stock-based compensation	Removes the largest non-cash cost	What % of revenue is this? Above 30% is a yellow flag
3	+ Amortization of acquired intangibles	Removes deal-related accounting	Is the company a serial acquirer leaning on this?
4	+ Other one-time items	Removes restructuring, IPO costs, legal	Are “one-time” items recurring every year?
5	= Non-GAAP operating profit	The number management leads with	Did SBC alone flip the sign from loss to profit?
6	Cross-check: diluted share count	The cost SBC was removed at	Is share count growing faster than the margin improved?

The discipline lives in rungs 2 and 6. If climbing from rung 1 to rung 5 turns a loss into a profit and the bulk of the climb is rung 2, then the company is profitable only after you ignore how it pays its workforce. Rung 6 is the one most readers skip: the SBC you added back to flatter the margin reappears as shares in the denominator of earnings per share. You cannot honestly delete the cost from one statement and ignore its arrival on another.

Two filings show the ladder in extreme form:

• Figma (Form S-1/A, FY2024): a GAAP operating loss of $877.4M on $749M of revenue became an adjusted operating profit of $127.2M once SBC and related charges were excluded. The entire sign flip, loss to profit, is the SBC rung.
• CoreWeave (Form S-1, FY2024): a GAAP net loss of $863M narrowed to an adjusted net loss of roughly $65M after add-backs, with stock-based compensation the primary gap between the two figures.

In both cases rung 2 is doing almost all the work. That is not necessarily damning. It is the thing you have to price.

The dilution myth: why “non-cash” doesn’t mean “costless”

The real cost of stock-based compensation is dilution, and dilution is not non-cash, it is non-current. When a company issues new shares to employees, each existing share becomes a smaller claim on the same future cash flows. The shareholder pays, just later and through ownership rather than through the income statement.

Think of it as a transfer. The company conserves cash by paying in equity. Employees receive that equity. The bill is sent to existing shareholders in the form of a thinner slice of the company. Removing SBC from non-GAAP earnings does not erase the transfer; it just relocates it from the income statement to the share count.

The measurement tool is the treasury stock method (TSM). Net dilution equals the gross dilutive securities (options and RSUs that would convert to shares) minus the number of shares the company could buy back with the hypothetical exercise proceeds. RSUs, which dominate modern grants, carry near-zero exercise proceeds, so they are close to fully dilutive. As Morgan Stanley’s Counterpoint Global research and Aswath Damodaran’s work on share count both argue, the cleanest read is to track net new shares issued, period over period, against the growth in cash flow.

Here is the practical consequence. A company can show flat or improving non-GAAP margins while its diluted share count climbs several percent a year. Earnings per share, the number that actually accrues to an owner, can stall or fall even as adjusted profit “grows,” purely because the denominator is expanding. Buybacks can offset the issuance, but a buyback funded to mop up employee dilution is capital spent to stay in place, not capital returned. The same “own your cost floor” logic that governs unit economics in usage-based pricing vs seat-based pricing applies to the cap table: dilution you do not measure is a cost someone else is setting for you.

How do you read SBC in an S-1 without getting played?

Read GAAP first and treat every non-GAAP add-back as a claim to test, sizing the SBC line against revenue and checking whether it alone flips a loss into a profit. The trap in an IPO prospectus is reading the non-GAAP profit as the “true” number and the GAAP loss as accounting noise. The correct order is reversed: GAAP is the starting truth, and each non-GAAP add-back is a claim you should test before accepting.

Work the reconciliation with the SBC Adjustment Ladder open beside it, and apply four checks.

1. Size the SBC rung against revenue. Convert the SBC dollar figure to a percentage of revenue. Above 30% means equity is a primary operating cost, not a footnote, and the non-GAAP profit is heavily manufactured.
2. Check whether SBC alone flips the sign. If GAAP is a loss and non-GAAP is a profit, and the SBC add-back is larger than the gap, the company is profitable only by ignoring how it pays people.
3. Look for the missing share count. A reconciliation that adds back SBC but never shows diluted shares outstanding is presenting half a transaction. Find the share count and its growth rate elsewhere in the filing and pair it with the margin.
4. Test whether “one-time” items recur. IPO-related RSU releases are genuinely one-time. Restructuring, “transaction costs,” and litigation that appear in adjustment tables every single year are not.

For SaaS specifically, Equity Methods’ research notes that stock-based compensation is the most common non-GAAP exclusion and frequently the single largest one for companies such as Workday, Splunk, Atlassian, and Okta, where it is the difference between a GAAP loss and a non-GAAP profit. That is a useful prior to carry into any software S-1. The same filing-literacy that reads cost structure properly catches inflated profitability here: the headline is an output, and you have to read the inputs.

The operator playbook: tracking share count vs. margin expansion

If you build software or read filings for a living, the transferable lesson is a two-track measurement habit: never let an adjusted-margin improvement be reported without the matching share-count change. Here is the playbook, five concrete moves you can run on the next prospectus or board deck.

1. Build the two-line dashboard. Track non-GAAP operating margin on one line and year-over-year diluted share-count growth on the other. Improving margin with rising share count is the signature of dilution-funded “profitability.” Improving margin with flat or falling share count is the real thing.
2. Convert SBC to per-share terms. Divide annual SBC by shares outstanding to get the per-share equity cost. It reframes an abstract billion-dollar add-back as the ownership each existing holder gives up that year.
3. Run the SBC Adjustment Ladder on every reconciliation. Start at GAAP, climb each rung, and write down which rung created the profit. If rung 2 (SBC) created it, your model should treat the company as pre-profit and price the path to real margin.
4. Treat IPO-cliff grants as a known event, not a surprise. Many companies recognize a large RSU charge at or near IPO (Figma’s $889.3M release is the textbook case, per its Form S-1/A). Separate that one-time spike from the recurring run-rate so you do not over-extrapolate either number.
5. Watch the policy signal. When a bellwether like Nvidia stops excluding SBC from non-GAAP (effective Q1 FY2027), it changes the peer comparison. A company still excluding a large SBC line while its largest peer no longer does is making a choice worth questioning.

For a founder, the same habit runs in reverse: model your own dilution before you celebrate your own adjusted EBITDA. The cleanest cap table is the one where equity grants are sized against the ownership you are willing to transfer, not against the cash you are trying to conserve this quarter. The broader filing-reading method, segment by segment and add-back by add-back, is the same one applied to infrastructure economics in the AI infrastructure market map and to platform margins in Apple Services as the margin engine inside iPhone.

The bear case: what the skeptics get right about excluding SBC

The strongest argument is not that SBC matters. It is that the people who exclude it have a real point, and a purist who simply re-adds every dollar can be just as wrong as the company that waves it all away.

The bear case for excluding SBC runs like this. Stock-based compensation is genuinely non-cash, and cash is what funds operations, services debt, and ultimately accrues to owners. A company with a deep GAAP loss driven entirely by SBC can be free-cash-flow positive and self-sustaining, which a naive reader of the GAAP loss would miss. CoreWeave’s narrowing from an $863M GAAP loss to a roughly $65M adjusted loss (Form S-1) is partly a story about a business that consumes far less cash than its GAAP statement implies. Excluding SBC, in that frame, is removing accounting noise to reveal the cash engine.

There is a second valid point. Not all dilution is permanent. A company that generates real free cash flow can repurchase shares to offset employee issuance, converting “dilution” back into a cash cost that the cash-margin view already captured. For those companies, double-counting SBC as both an income-statement expense and a share-count hit overstates the damage.

Here is the honest weighing. The skeptics are right that SBC is non-cash and that cash flow is real, and the right response is not to ignore the cash view, it is to hold both views at once. Strip SBC to see the cash margin (the skeptics’ valid point), then look at diluted share-count growth and any buyback spend to see what the cash margin cost in ownership (the dilution point). The two are not in conflict; they are two faces of the same grant. The position to avoid is the one that takes only the cash view and never opens the cap table, because that is exactly the read the non-GAAP slide is designed to encourage.

Where SBC analysis is vulnerable: what the numbers miss

A claim this strong deserves its own counterexamples. Folding SBC back into a valuation is the right default, and it can still mislead if applied mechanically.

Fair-value accounting is an estimate, not a fact. Under ASC 718, SBC expense is based on the grant-date fair value of equity, computed with option-pricing models and assumptions about volatility and vesting. Two companies with the same economic generosity can report different SBC depending on those inputs. The number is a model output, not a measured cash flow, so precision past the first significant figures is false comfort.

Timing distorts the percentage. A one-time IPO RSU release, like Figma’s, can push SBC above 100% of revenue in a single year and then fall sharply. Reading that spike as the run-rate would badly overstate ongoing cost. Conversely, a company that front-loads small grants can look cheap on equity right before a hiring wave inflates it.

Buybacks complicate the dilution story in both directions. A buyback that offsets employee issuance is a real cash cost, but it also means the share count understates how much equity was actually granted. A reconciliation that omits both the SBC line and the buyback can hide the full transfer. You have to read issuance and repurchase together.

SBC says nothing about retention or growth durability. A company can have low SBC and a collapsing business, or high SBC and a franchise worth diluting for. SBC measures one cost; it does not measure whether the company is worth owning. That judgment lives in the same place as customer concentration and retention quality, not in the reconciliation table.

None of this overturns the thesis. It bounds it. SBC is the most important add-back to scrutinize at IPO, and it is one input among several, best read as a range rather than a single decisive number.

Case study: the 2025 to 2026 IPO wave and the SBC question

The recent IPO and pre-IPO cohort makes the pattern concrete. Across very different businesses, SBC sits at the center of the GAAP-to-non-GAAP bridge, and the cleaner filings show the math both ways.

Cerebras Systems (Form S-1, filed April 17, 2026) reported a GAAP operating loss of $145.9M in 2025 and a non-GAAP net loss of $75.7M after adjusting for $49.8M of stock-based compensation. The SBC rung is roughly a third of the gap between the two figures, a more moderate ratio than the SaaS extremes, reflecting a hardware cost base where compensation is a smaller share of total spend.

Figma (Form S-1/A, FY2024) is the opposite extreme: an $877.4M GAAP operating loss flips to a $127.2M adjusted operating profit almost entirely on the SBC and RSU-release add-backs. The ladder’s rung 2 carries the whole climb.

For the largest private AI companies, the picture is necessarily less precise because the filings are confidential. Reporting around an OpenAI confidential S-1 filing (June 8, 2026) has pointed to very large operating losses before any SBC add-back, with projected SBC in the billions. Those figures are not in a public SEC filing and should be treated as illustrative, not verified: the takeaway is directional, that at the AI frontier the SBC line will be measured in billions and will dominate any non-GAAP bridge, not a specific number you can underwrite.

The throughline across the cohort, from a chip designer to a design-software company to a frontier AI lab, is that the reader’s first move is identical. Find the SBC line, size it against revenue, and check whether it is the rung that turns the loss into a profit. The economics that govern these reconciliations sit alongside the platform-cost pressures examined in AWS margin pressure and the cloud reset and the app-store take-rate dynamics in Apple’s App Store economics under pressure: in every case the headline number is downstream of a cost structure you have to read for yourself.

Methodology: how to read an SBC-driven reconciliation

When you take a GAAP loss and a non-GAAP profit and draw a conclusion about a company’s economics, here is the frame to keep it honest.

• Inputs: GAAP operating income or loss, the SBC add-back, other reconciling items, revenue, and diluted shares outstanding, all from the most recent S-1 or 10-K (e.g., Figma’s $877.4M GAAP loss and $127.2M adjusted profit on $749M revenue, Form S-1/A; Snowflake’s $1.479B SBC, Form 10-K FY2025).
• Assumptions: that grant-date fair value reasonably approximates the economic cost of equity granted, and that disclosed share count captures dilution. Both can drift when valuation inputs change or when large RSU cliffs land in one period.
• Sensitivity: SBC at 40% of revenue versus 10% changes whether a company is structurally pre-profit or merely optically so. A few points of annual share-count growth materially change per-share economics even when adjusted margin is flat.
• What this misses: public filings rarely isolate SBC by function cleanly, fair-value estimates carry model risk, and a single year’s percentage can be distorted by IPO-timing charges. The output is a range and a direction, not a target price.

This is a framework for understanding cost structure and dilution, not a model that outputs what a company is worth.

How the pieces fit together

Stock-based compensation in tech IPOs is not an accounting curiosity. It is the line that decides whether a company’s headline profitability is real or rented.

The logic stacks cleanly:

1. At IPO, SBC is usually the largest item bridging the GAAP loss to the non-GAAP profit.
2. It is “non-cash” only in timing. The cost is real and arrives as dilution.
3. The right read is two-track: strip SBC to see cash margin, then track diluted share count to see what that margin cost in ownership.
4. The SBC Adjustment Ladder is how you climb the reconciliation without getting played, with rung 6, the share count, as the rung most readers skip.
5. Nvidia’s move to stop excluding SBC is a signal the convention is under pressure, and a reason to fold the cost back in by default.

The companies competing to show the cleanest non-GAAP profit are competing on the one axis a disciplined reader can adjust for in a single step. The axis that actually decides the economics is dilution, and that one shows up in the share count, not the margin.

That is the whole read. The rest is matching each add-back to the cost it conceals.

---

Analysis, not investment advice. Figures are drawn from public SEC filings cited inline by company and fiscal period (Snowflake, Alphabet, Meta, and Nvidia Forms 10-K; Figma, CoreWeave, and Cerebras Forms S-1/S-1A) and from cited research from Morgan Stanley, Aswath Damodaran, and Equity Methods. Frameworks here are for understanding business models and tradeoffs, not for making buy or sell decisions.

Want the full toolkit for reading filings like this, the non-GAAP reconciliation worksheet, the dilution tracker, and the SBC Adjustment Ladder used above? It’s in the Tech Business Analysis Playbook.