The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Short answer: never treat an AI company’s ARR, annualized revenue, bookings, or press-reported run rate as interchangeable with audited annual revenue. First identify exactly what the number measures, its date and accounting basis, and whether cloud or reseller sales are shown gross or net. Then test whether revenue per customer, seat, task, or token covers the variable cost of serving that usage, while keeping model training, research, sales, and financing costs separate.
This framework is especially important for private frontier-AI companies. Their public figures may be company-reported or estimated by investors and journalists, while audited income statements, cash flow, customer retention, and contract terms remain unavailable. A fast-growing run rate can therefore coexist with weak contribution margins or heavy cash burn.
Start by naming the number
The phrase “revenue” is often used for several different measurements. Before comparing two AI companies, write down the metric name, measurement date, period covered, currency, geography, products included, and who reported it.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →| Metric | What it measures | What it does not prove |
|---|---|---|
| Recognized revenue | Revenue recorded under the company’s accounting framework for a completed reporting period, such as a quarter or fiscal year. | That the same sales level will recur, or that the company is profitable. |
| Annual recurring revenue (ARR) | A point-in-time estimate of recurring subscription or contract value, usually multiplied to a one-year equivalent. | That it is GAAP revenue, cash collected, or a forecast of next year’s recognized revenue. |
| Annualized revenue or run rate | A recent month, quarter, or other short-period sales rate multiplied into a year. | That the company earned that amount during a completed year. |
| Bookings or billings | Contracted orders or invoices, which may be recognized as revenue over time. | That the service has been delivered or the cash is collectible. |
| Remaining performance obligations | Contracted consideration for goods or services not yet recognized as revenue. | That every dollar will be recognized soon or without cancellation. |
| Gross transaction value | Total customer spend flowing through a marketplace or partner channel. | That the AI company keeps the full amount as revenue. |
For every quoted figure, label it explicitly—for example, “company-reported ARR at December 31, 2025” or “press-reported annualized sales based on a recent period.” If the source does not disclose the basis, mark it as undisclosed instead of silently assuming GAAP or IFRS treatment.
#1 Best Overall
Understand what ARR means—and what it does not mean
ARR is most informative when subscriptions are stable, contracts renew predictably, and usage does not swing sharply. It becomes less reliable for usage-led AI products whose demand, token consumption, pricing, or model mix can change within weeks.
Amplitude’s fiscal-year 2025 Form 10-K gives a useful public-company example. It defines ARR from subscription agreements at a point in time and states that ARR is not U.S. GAAP revenue on an annualized basis. Contract start and end dates and renewal rates can change the measure. The filing also reports 698 customers above $100,000 of ARR and 56 above $1 million as of December 31, 2025; those customers represented 78% of ARR, while no single customer exceeded 10% of 2025 revenue. These figures illustrate how a serious ARR disclosure combines a definition, a date, customer cohorts, concentration, and retention context. They are company-specific and should not be generalized to frontier AI laboratories.
Ask whether ARR includes only committed subscription fees or also expected usage, expansion, discounts, services, and contracts that have not started. Ask how downgrades, cancellations, non-renewals, and consumption-based plans are handled. Two companies can both publish “ARR” while measuring materially different economic bases.
Separate a run rate from a completed year
An annualized figure is usually a multiplication, not a historical accounting result. A simple calculation is annualized sales = sales in the selected period × periods per year. If a company generated $1 billion in an unusually strong month, multiplying by 12 assumes that launch timing, discounts, enterprise deals, usage spikes, and seasonality will repeat.
- Identify the source period. Record the exact month, quarter, or trailing window and whether it was actual sales, an estimate, or a forecast.
- Find the underlying sales. Request monthly or quarterly revenue rather than accepting only the multiplied headline.
- Check for timing effects. Look for a major contract, model release, price change, promotional credit, or unusually large usage event.
- Compare with the last completed fiscal year. Reconcile the run rate with recognized revenue, deferred revenue, billings, and remaining performance obligations when available.
- Use the right label. Call the result “annualized revenue” or “run rate,” never “annual revenue,” unless a completed reporting period supports that description.
Subscription revenue may be recognized over a contract term. A large new sale can therefore increase bookings immediately while appearing gradually in reported revenue. C3.ai’s 2026 Form 10-K explains that identifying distinct performance obligations in multi-element contracts requires judgment, which can affect how much revenue is recognized in a particular period.
Normalize cloud and channel sales before comparing companies
Partner distribution can create a large difference between what a customer pays and what the AI company reports as revenue. Determine who contracts with the customer, who controls and delivers the service, and whether the cloud provider’s or reseller’s share is deducted from revenue or recorded as an expense.
| Question | Why it matters |
|---|---|
| Who is the customer’s contractual counterparty? | The answer helps determine whether the AI company is principal or agent for accounting purposes. |
| Who controls delivery and sets key terms? | Control can affect whether sales are presented gross or net. |
| What percentage goes to the cloud provider, marketplace, or reseller? | The same customer spend can produce different reported revenue and expense totals. |
| Are credits, minimum commitments, and usage charges included? | These terms can shift both timing and amount of recognized revenue. |
Axios reported on October 8, 2026, and in a September 3 comparison, that OpenAI records only its share of certain partner sales while Anthropic’s method can report the full customer payment as revenue and record the cloud provider’s share as an expense. This is a presentation difference, not evidence by itself that one company has more demand or superior economics.
For an apples-to-apples comparison, show two columns whenever evidence permits: customer spend through the channel and company-recognized revenue. Keep the gross-versus-net policy visible. Do not “correct” one company by guessing a partner share that has not been disclosed.
Use current AI examples without overstating them
OpenAI’s 2026 company-authored post states ARR of $2 billion in 2023, $6 billion in 2024, and more than $20 billion in 2025. It also reports compute capacity of 0.2 gigawatts in 2023, 0.6 gigawatts in 2024, and approximately 1.9 gigawatts in 2025. These are company statements about ARR and capacity, not audited annual revenue or an independent profitability measure.
Axios reported two different annualized-revenue figures for OpenAI: $50 billion and $70 billion. The October 8, 2026 coverage explained that the higher figure reflected an attempt to gross up partner-mediated sales so the comparison would align more closely with Anthropic’s presentation. Treat both as reported estimates with different bases; neither is audited revenue recognized over a full year.
Rank #3
Growth alone cannot establish healthy unit economics. Axios reported on October 13, 2026, that public information about major private AI companies’ expenses, cash burn, and profitability is limited. Venture capitalist and blogger Paul Kedrosky described startup ARR figures as “inherently arbitrary and capricious,” a criticism of inconsistent definitions rather than a formal accounting rule.
Recommended Free Tools
Test whether growth is durable
A rising ARR or customer count can hide churn, discounts, one-time workloads, or concentration. Evaluate the quality of growth with the following evidence:
- Paid versus free usage: Separate active users and tokens from paying accounts.
- Retention and expansion: Seek gross revenue retention, net revenue retention, renewal rates, seat expansion, and usage expansion.
- Contract structure: Note term length, minimum commitments, cancellation rights, and renewal dates.
- Discounting and credits: Distinguish list-price demand from heavily subsidized consumption.
- Customer concentration: A few large contracts can accelerate ARR but increase renewal risk.
- Mix: Separate subscriptions, usage-based API sales, professional services, one-time licenses, and hardware or deployment fees.
- Repeat behavior: Check whether customers renew and increase usage after initial trials or pilots.
Customer count alone does not demonstrate durable monetization. Require the company to state how it counts a customer, whether related entities are combined, and whether inactive or free accounts are included.
Define the unit before calculating unit economics
“Unit economics” has no meaning until the unit is specified. Depending on the product, a unit might be one API request, one million input or output tokens, one completed task, one seat, one deployment, or one customer contract.
- Choose a measurable unit. For an API, separate input and output tokens or requests by model. For an enterprise product, a seat, task, or deployment may be more useful.
- Calculate realized revenue. Use collected or recognized price after discounts, credits, refunds, and partner deductions—not the published list price.
- List variable delivery costs. Include inference compute, storage, network egress, cloud or marketplace shares, usage-based licenses, support tied directly to consumption, and safety or moderation costs that scale with volume.
- Compute contribution margin.
contribution margin per unit = realized revenue per unit − variable cost per unit. State exactly which costs are included. - Test cohorts and models. Compare new and mature customers, model variants, context lengths, geographic regions, and peak versus average usage.
- Calculate payback where relevant. For customer acquisition,
payback period = acquisition cost ÷ monthly contribution profit, using a clearly defined contribution profit.
Do not call contribution margin gross margin. Gross margin may include a broader cost-of-revenue category; operating income also includes research, sales, administration, stock compensation, and other expenses. A company can have positive contribution on each API call while losing substantial cash on model development and infrastructure commitments.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errorsKeep inference, training, and corporate costs separate
Inference is the cost of answering users’ requests. Training and research create or improve models but may not rise in lockstep with current usage. Sales, support, safety, data acquisition, facilities, stock-based compensation, and financing costs affect overall profitability but should not be silently assigned to a single token or seat.
Stanford HAI’s AI Index 2026, drawing on Epoch AI estimates, reports annual compute spend of $16.3 billion for OpenAI and $6.8 billion for Anthropic in 2025. Its chart separates inference, research and development, and unattributed compute spending. Do not treat those totals as inference-only costs, and do not substitute them for audited income statements.
OpenAI said, “Compute is the scarcest resource in AI.” In the same company-authored discussion, it linked lower-cost infrastructure and high-volume workloads to costs measured in cents per million tokens and said, “That is what makes AI viable for everyday workflows, not just elite use cases.” Present these as OpenAI’s claims about its cost trajectory, not as independently verified margins.
Read public filings for accounting evidence
For a public company, begin with audited financial statements and the revenue-recognition note in its SEC filing. Then inspect:
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware match- Contract assets, contract liabilities, and deferred revenue.
- Remaining performance obligations and expected recognition timing.
- Customer concentration and related-party transactions.
- Stock-based compensation and non-GAAP reconciliations.
- Cloud, data-center, lease, and purchase commitments.
- Cash flow, capital expenditure, debt, and liquidity.
- Accounting judgments and critical audit matters.
C3.ai’s 2026 Form 10-K shows why the performance-obligation note matters: multi-element contracts require judgment about which promises are distinct and when control transfers. That judgment can change the period in which revenue appears without changing the customer’s total contractual price.
Best Value
A financial-statement analysis book can help readers learn these concepts, but it is optional; the checklist above can be applied directly to filings.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Evaluate private-company numbers by evidence quality
Private AI companies often disclose only selected metrics. Label each figure according to its provenance:
| Evidence label | How to use it |
|---|---|
| Audited filing | Strongest basis for recognized revenue, liabilities, cash flow, and accounting policy, subject to the auditor’s scope. |
| Company-reported metric | Useful when the definition and date are clear; still not equivalent to audited revenue unless stated. |
| Investor or lender estimate | Record assumptions, date, and whether the estimate is gross or net of partner sales. |
| Press-reported figure | Check the underlying source and avoid presenting it as a company-confirmed result. |
| Undisclosed | Leave the comparison blank rather than estimating margins, churn, or cash burn. |
No comparable audited profitability or unit-margin statistic for private frontier AI companies is established by the figures above. Any conclusion about their margins must therefore be qualified as incomplete.
Free tools Windows power users keep installed
One-click scans. No signup required.
Use a comparison grid instead of one headline
When comparing companies, put the same axes in the same order:
| Axis | Company A | Company B |
|---|---|---|
| Recognized revenue and period | State amount, currency, period, and accounting basis. | State amount, currency, period, and accounting basis. |
| ARR or annualized sales | Definition, date, and source. | Definition, date, and source. |
| Partner-sales presentation | Gross or net; explain channel treatment. | Gross or net; explain channel treatment. |
| Subscription versus usage mix | Disclose or mark not disclosed. | Disclose or mark not disclosed. |
| Retention, expansion, and concentration | Report cohort metrics or mark not disclosed. | Report cohort metrics or mark not disclosed. |
| Gross and contribution margins | List included costs and period. | List included costs and period. |
| Compute spend and commitments | Separate inference, R&D, and unattributed amounts. | Separate inference, R&D, and unattributed amounts. |
| Cash burn and capital needs | Use reported cash flow or mark not disclosed. | Use reported cash flow or mark not disclosed. |
| Stock-based compensation | State amount or mark not disclosed. | State amount or mark not disclosed. |
If an axis is unavailable for either company, write “not disclosed.” A complete-looking table built from guesses is less informative than an incomplete table that shows the evidence boundary.
Red flags that invalidate a quick conclusion
- A headline says “revenue” but the underlying statement uses ARR or annualized sales.
- The measurement date is missing or falls in the middle of a quarter.
- A recent month is multiplied by 12 without a seasonality or contract-timing discussion.
- Customer spend is compared with company revenue even though one company reports partner sales net.
- ARR growth is shown without churn, renewal, discount, or concentration data.
- Gross margin excludes inference, cloud, support, or safety costs that scale with usage.
- Compute spending is treated as entirely variable inference cost or entirely as research investment.
- Non-GAAP “profit” removes recurring stock compensation or infrastructure costs without reconciliation.
- A press estimate is presented with the precision of an audited statement.
A practical decision rule
You can regard an AI revenue claim as decision-grade only when you can answer five questions: What exactly is measured? For which period and date? Under what accounting basis? How are partner sales treated? Which costs are included in the associated margin?
If any answer is missing, downgrade the claim. Use recognized revenue for historical performance, ARR for clearly defined recurring contracts, and annualized figures only as directional snapshots. Then connect the revenue measure to a defined unit, variable delivery costs, compute commitments, retention, and cash needs. This prevents the most common error in AI-company analysis: mistaking a rapidly rising number for evidence that the underlying business is already durable and profitable.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
