OpenAI's annualized revenue reportedly reaches $50 billion as accounting differences complicate comparison with Anthropic

OpenAI's annualized revenue reached approximately $50 billion at the end of September 2026, according to financial information reportedly shared with investors, substantially below the $70 billion figure circulated only days earlier. The apparent $20 billion discrepancy has raised questions about the company's financial performance and the reliability of revenue comparisons across major artificial intelligence developers, although the available evidence indicates that the difference primarily concerns accounting presentation rather than a sudden deterioration in customer demand.

The larger figure originated from efforts to make OpenAI's revenue comparable with that of Anthropic, whose reported annualized sales include business generated through cloud distribution partners. OpenAI excludes certain partner-mediated sales from its own reported revenue measure, producing a narrower figure than estimates that incorporate the gross value of those transactions. When investors attempted to normalize the two companies' figures, the adjusted OpenAI estimate approached $70 billion, creating the impression that the company was generating substantially more revenue than its subsequent investor materials indicated.

The distinction is financially significant but should not be interpreted as $20 billion of revenue disappearing from OpenAI's business. The two figures reflect different measurement boundaries, and neither represents the company's actual revenue recognized over the preceding twelve months. Both are annualized indicators extrapolating recent business activity into a hypothetical full-year period.

The controversy arrives during an exceptionally rapid expansion of the commercial AI market. In March, OpenAI publicly reported generating approximately $2 billion in monthly revenue, equivalent to a $24 billion annualized pace, while Anthropic disclosed that its revenue run rate had exceeded $47 billion in May. Subsequent investor figures placed Anthropic above $65 billion by the end of July, suggesting that both companies were experiencing substantial demand growth even as their revenue reporting methodologies complicated direct comparisons.

The consequences extend beyond the relative financial positions of OpenAI and Anthropic. Their expected growth supports large commitments to GPUs, data centers, cloud infrastructure, electricity supply, and financing arrangements across the technology sector. Investors therefore need to distinguish genuine changes in AI demand from changes in how revenue is defined, particularly when private companies publish selected operating metrics rather than a complete set of audited financial statements.

··········

THE $50 BILLION AND $70 BILLION FIGURES REFLECT DIFFERENT REVENUE MEASURES.

The September figures describe approximately the same underlying period, but the higher estimate incorporated additional business activity to facilitate comparisons with Anthropic's reporting methodology.

The discrepancy became public on October 8, after information from OpenAI's investor materials indicated that its annualized revenue was approaching $50 billion at the end of September. Earlier indications had suggested a figure close to $70 billion, leading some investors to interpret the new information as a substantial shortfall.

The explanation concerns how sales generated through third-party cloud platforms enter the revenue measure. OpenAI's reported calculation excludes certain sales made through cloud partners, while Anthropic's comparable headline measure includes revenue associated with distribution through providers such as Amazon Web Services and Google Cloud.

Investors attempting to compare the two companies adjusted OpenAI's figures to include additional partner-related sales. That broader calculation produced an estimate of approximately $40 billion for July and subsequently contributed to reports that the September run rate was approaching $70 billion.

The narrower presentation contained in OpenAI's later investor information showed annualized revenue closer to $30 billion in July and $50 billion in September. Although the absolute figures were lower, they continued to indicate extremely rapid commercial expansion.

Using the rounded values, growth from $30 billion to $50 billion represents approximately 66.7%. The underlying investor figures reportedly showed an increase exceeding 70%, with the difference attributable to rounding and the more precise amounts used in the calculation.

........

Revenue indicator	Approximate annualized amount	Reporting basis
OpenAI, March 2026	$24 billion	Officially disclosed monthly revenue annualized
OpenAI, July 2026	$30 billion	Narrower reported revenue measure
OpenAI, July 2026	$40 billion	Broader investor-adjusted estimate
OpenAI, September 2026	$50 billion	Narrower reported revenue measure
OpenAI, September 2026	$70 billion	Broader investor-adjusted estimate
Anthropic, July 2026	More than $65 billion	Reported measure including cloud-partner sales

........

The table combines official company information and figures attributed to investor communications. It does not present an audited financial reconciliation, and the precise composition of the adjustments has not been disclosed publicly.

Two conclusions follow from these numbers.

First, a reported revenue run rate of $50 billion is consistent with substantial growth, even though it is below the earlier $70 billion estimate. The lower figure does not, by itself, establish a reduction in sales, customer spending, or enterprise adoption.

Second, the difference between the two September measures cannot automatically be interpreted as $20 billion of incorrectly recognized accounting revenue. It represents a gap between presentations, and an exact reconciliation would require information about customer contracts, distribution arrangements, the treatment of partner sales, and the measurement methodology used for each figure.

A further source of confusion is that $70 billion has also appeared as a reported target for OpenAI's annualized revenue by the end of 2026. A year-end forecast is conceptually different from an estimate of September's current business activity, even when both happen to use the same amount.

Separating the reporting date, accounting perimeter, and forecast period is therefore necessary before drawing conclusions about whether OpenAI has exceeded or missed expectations.

··········

ANNUALIZED REVENUE IS NOT THE SAME AS ACTUAL ANNUAL SALES OR RECURRING REVENUE.

An annualized revenue run rate extrapolates recent business activity over twelve months, making it useful for understanding growth momentum but unsuitable as a direct substitute for recognized annual revenue.

For a company experiencing rapid expansion, the revenue generated during the most recent month can be considerably higher than the monthly average achieved earlier in the year. Annualizing that month provides a snapshot of the business's current scale without requiring investors to wait for the completion of the financial year.

The basic calculation is straightforward:

Annualized revenue = Revenue during the reference month × 12

A reported annualized figure of $50 billion therefore corresponds to approximately $4.17 billion in monthly revenue, assuming that September is the reference period. The broader $70 billion estimate corresponds to approximately $5.83 billion per month under the same annualization convention.

Neither calculation establishes that the company actually generated $50 billion or $70 billion during the preceding twelve months. It describes the theoretical amount that would result if the selected monthly revenue level continued unchanged for a full year.

Consider a hypothetical company whose revenue increases from $1 billion per month in January to $4 billion in December, rising evenly by approximately $273 million each month. Its December revenue would imply a $48 billion annualized run rate, but the actual revenue earned throughout the year would total approximately $30 billion.

The difference is not an accounting error. Annualized revenue measures the latest commercial pace, while full-year revenue aggregates the amounts earned during the reporting period.

This distinction is particularly important for frontier AI companies because their revenue streams combine different commercial models. Consumer subscriptions, enterprise subscriptions, API usage, coding agents, platform integrations, and contracted services may exhibit different degrees of predictability.

A subscription customer paying a fixed monthly fee can generate relatively stable revenue until cancellation or renewal, whereas API consumption varies with application traffic, model pricing, workload characteristics, and customer demand. An enterprise contract may establish a minimum commitment without requiring all contracted usage to be recognized immediately as revenue.

Annualized revenue should therefore not automatically be labeled annual recurring revenue (ARR). Although the terms are sometimes used loosely in discussions of technology companies, ARR generally describes a normalized measure of recurring contractual or subscription revenue, while an annualized run rate may include variable consumption and other business activity that does not recur predictably.

The difference becomes more consequential when investors value companies using revenue multiples. A business generating most of its revenue through contracted subscriptions may have a different risk profile from one generating equivalent revenue through concentrated, highly variable API consumption.

For OpenAI, the composition of revenue matters as much as the absolute run rate. Enterprise adoption can increase contractual stability, while usage-based products can produce rapid expansion accompanied by greater variability in computing expenditure.

A meaningful financial assessment would therefore distinguish actual recognized revenue, recurring contractual revenue, usage-based revenue, annualized run rate, and remaining contractual commitments instead of treating them as interchangeable measures of commercial performance.

··········

GROSS VERSUS NET ACCOUNTING EXPLAINS WHY CLOUD DISTRIBUTION CAN CHANGE REPORTED REVENUE.

The accounting treatment of transactions involving cloud providers depends on contractual responsibilities and control over the service delivered, not simply on which company developed the underlying AI model.

Frontier models increasingly reach customers through third-party platforms. A business may consume an AI model directly through the developer's API, purchase access through a cloud marketplace, or use the model as part of a broader service operated by a cloud provider.

These arrangements introduce several economically distinct participants: the model developer, the cloud infrastructure provider, the distribution platform, and the ultimate customer.

A transaction may generate one amount paid by the customer while producing different recognized-revenue figures for the companies involved, depending on their contractual obligations.

Under ASC 606, Revenue from Contracts with Customers, a company generally recognizes revenue on a gross basis when it acts as principal and controls the specified good or service before transferring it to the customer. When it acts as an agent arranging for another party to provide the service, it generally recognizes its fee or commission on a net basis.

The principal-versus-agent assessment depends on the substance of the transaction, including which entity controls the promised service, bears responsibility for fulfilling it, and has discretion over pricing. Companies cannot simply choose whichever accounting presentation produces the highest revenue figure if the underlying contractual facts do not support it.

Consider an illustrative AI service purchased by an enterprise customer for $100, with a cloud distribution partner retaining $15 and the model developer ultimately receiving $85.

The accounting presentation can differ according to which entity controls the service provided to the customer.

........

Illustrative arrangement	Principal	Developer's recognized revenue	Developer's economics
Developer controls and sells the AI service; cloud partner acts as agent	Model developer	$100	$100 revenue less $15 distribution expense
Cloud platform controls and resells the service; developer supplies the platform	Cloud platform	$85	$85 revenue received from the platform

........

In both scenarios, the model developer receives $85 before considering other expenses, and the end customer pays $100. Nevertheless, the developer's reported revenue differs because the contractual roles differ.

The example isolates one accounting mechanism and assumes the amounts and services described meet the relevant recognition conditions. It is not a reconstruction of OpenAI's or Anthropic's actual agreements.

This distinction also explains why gross revenue is not gross profit, and net revenue is not net income. A company recognizing $100 in revenue and $15 in distribution expense may retain the same amount before other costs as another company recognizing $85 in revenue, but the two presentations produce different reported revenue and potentially different gross-margin percentages.

For frontier AI companies, cloud relationships can be particularly complicated because distribution and infrastructure procurement are separate economic activities. A developer may pay for computing capacity while also distributing models through a cloud marketplace, and those arrangements need not have identical contractual structures or accounting treatments.

OpenAI's relationship with Microsoft illustrates this complexity. Under the companies' amended April 2026 agreement, Microsoft remains OpenAI's primary cloud partner, while OpenAI can make its products available through other cloud providers. OpenAI's revenue-sharing payments to Microsoft continue through 2030 under specified conditions, even as other aspects of the partnership have changed.

Those revenue-sharing obligations should not be confused automatically with the gross-versus-net treatment of cloud marketplace transactions. Revenue sharing, infrastructure costs, reseller commissions, and the recognition of customer revenue involve different accounting questions, even where they arise within the same commercial relationship.

Microsoft's own financial reporting provides evidence of the relationship's scale. In its fiscal 2026 disclosures, the company reported $24.1 billion of revenue from commercial arrangements with OpenAI, including revenue-sharing payments, for the year ended June 30.

That amount is revenue recognized by Microsoft, not OpenAI's revenue, and cannot be inserted into or subtracted from OpenAI's annualized figures to produce a reliable gross-to-net reconciliation. The disclosure nevertheless illustrates the financial significance of the infrastructure and contractual relationships surrounding frontier-model developers.

The public reporting on OpenAI and Anthropic describes different treatments of cloud-partner sales, but it does not provide the underlying contracts or audited revenue-recognition schedules needed to establish precisely how each transaction is reflected in financial statements.

The appropriate conclusion is therefore narrower than suggesting that one company has inflated its revenue or that the other has understated it. Their reported headline measures are not directly comparable without normalizing the business included in each figure and identifying the accounting basis used.

··········

ANTHROPIC'S REPORTED $65 BILLION RUN RATE CANNOT BE COMPARED DIRECTLY WITH OPENAI'S $50 BILLION.

Anthropic's revenue measure includes business generated through cloud distribution partners, while the reported OpenAI figure excludes certain partner-mediated sales, creating an accounting difference alongside a two-month timing difference.

Anthropic's commercial growth has been substantial. In May 2026, the company publicly announced that its revenue run rate had exceeded $47 billion during the month, alongside a $65 billion funding round that valued the company at $965 billion after financing.

Subsequent investor information placed Anthropic's annualized revenue above $65 billion at the end of July, compared with approximately $9 billion at the end of 2025. Those figures indicate rapid expansion in Claude-related services, particularly enterprise adoption and usage through cloud platforms.

However, placing Anthropic's July figure directly beside OpenAI's September estimate of $50 billion would introduce at least two significant distortions.

The first concerns distribution accounting. Anthropic's headline measure includes the gross value of certain sales through partners such as AWS and Google Cloud, whereas OpenAI's narrower calculation excludes sales that do not fall within its reported revenue perimeter.

The second concerns measurement timing. July and September represent different points in a rapidly growing market, and a two-month difference can materially affect annualized results when revenue is increasing at the rates reported by both companies.

A third complication is the different composition of their businesses. OpenAI operates a large consumer subscription platform alongside enterprise products, APIs, and developer services. Anthropic has a substantial commercial focus on enterprise models, coding products, API consumption, and distribution through cloud platforms.

Even after normalizing revenue recognition and reporting periods, differences in customer concentration, product mix, revenue predictability, infrastructure costs, and commercial contract terms would remain relevant.

The fact that one company reports a higher annualized figure does not establish that it has a larger economically comparable business, higher margins, greater operating profitability, or stronger cash generation.

A proper comparison would require both companies to disclose revenue for the same period using consistent principles, including their treatment of partner-mediated sales. Investors would also need to distinguish revenue recorded by the model developer from amounts retained by cloud distributors.

Beyond accounting, customer spending should be classified according to its economic characteristics. Fixed enterprise subscriptions may offer greater visibility than discretionary API usage, while large cloud distribution channels can increase adoption but introduce dependence on platform pricing and contractual terms.

This is especially relevant as AI products become embedded within third-party software. A model developer may benefit from demand generated by an enterprise platform without controlling the complete customer relationship, while the distributor may capture a separate portion of the total expenditure.

The competitive question is therefore how much customer expenditure each company converts into recognized revenue, gross profit, and operating cash flow, rather than which company reports the largest standalone annualized figure.

Until comparable disclosures become available, the $50 billion and $65 billion figures should be treated as indicators of commercial scale under different reporting conventions, not as a definitive financial ranking.

··········

OPENAI'S REPORTED REVENUE GROWTH REMAINS RAPID, BUT PROFITABILITY REQUIRES SEPARATE ANALYSIS.

The narrower September revenue estimate still suggests substantial commercial expansion, while revealing relatively little about the costs necessary to support that growth.

OpenAI reported $2 billion in monthly revenue on March 31, equivalent to approximately $24 billion annually at that month's pace. If September's approximately $50 billion annualized figure were calculated on a comparable basis, it would imply that the company's monthly revenue had more than doubled over roughly six months.

The calculation is indicative rather than a verified like-for-like growth measurement, because OpenAI has not published a reconciliation confirming that the March and September amounts use identical reporting definitions.

OpenAI has also disclosed the increasing importance of enterprise customers. In April, the company stated that enterprise activity represented more than 40% of its revenue and was on track to approach parity with consumer revenue by the end of 2026.

That development matters because enterprise customers can generate recurring subscriptions, contracted usage commitments, and broader adoption across departments. At the same time, enterprise AI applications may require higher levels of support, reliability, data processing, and computational capacity than conventional consumer interactions.

The company has invested heavily in expanding its infrastructure through relationships with Microsoft, NVIDIA, Oracle, AWS, CoreWeave, Google Cloud, and other technology providers. Its ability to serve larger numbers of customers depends on obtaining sufficient computing capacity while improving the efficiency with which its models process requests.

Strong revenue growth does not automatically produce corresponding improvements in profitability.

An AI developer incurs costs associated with model training, inference, cloud infrastructure, personnel, research, product development, and commercial distribution. Some expenditures are directly associated with serving customers, while others support future model generations and infrastructure expansion.

The economics become particularly important when products are priced competitively while model capabilities continue improving. A more capable model may attract higher usage but require additional computation, and revenue growth can therefore coincide with elevated infrastructure expenditure.

OpenAI's reported annualized revenue also does not indicate how much cash the company has collected. Revenue recognition, customer billing, advance payments, supplier obligations, and capital commitments can produce substantial differences between accounting revenue and operating cash flow.

Large financing rounds further complicate the picture because proceeds from issuing equity are not customer revenue. OpenAI's March announcement of $122 billion in committed funding at an $852 billion post-money valuation represented capital raised from investors, not income generated by selling AI products.

For valuation purposes, these distinctions are fundamental.

A revenue multiple calculated using a $50 billion run rate will be higher than the same valuation divided by a $70 billion adjusted measure. At a hypothetical enterprise value of $1 trillion, the respective multiples would be 20.0× and approximately 14.3× annualized revenue.

The difference is approximately 40% in the calculated multiple, despite the assumed business valuation remaining unchanged. Neither figure would establish the correct valuation without considering the comparability of the revenue denominator, profitability, expected growth, and cash requirements.

This sensitivity helps explain why apparently small changes in reporting conventions can trigger substantial market reactions when AI-related companies are valued on expectations of exceptionally rapid expansion.

Following the October 8 reporting, several major technology and AI infrastructure stocks declined, reflecting renewed scrutiny of the financial assumptions supporting the industry's growth. Those market movements cannot be attributed exclusively to OpenAI's figures, but the episode demonstrated how closely investor expectations for semiconductors and data centers have become linked to the revenue trajectories of frontier AI developers.

The broader financial risk is not that one reported figure is necessarily incorrect. It is that investors may apply valuation multiples to metrics whose definitions, recurrence characteristics, and relationship with cash generation are insufficiently understood.

··········

INVESTORS NEED COMPARABLE REVENUE, MARGINS, AND CASH FLOW BEFORE DRAWING CONCLUSIONS ABOUT AI VALUATIONS.

The accounting discrepancy demonstrates why annualized revenue alone cannot support a comprehensive assessment of the financial performance of frontier AI developers.

OpenAI and Anthropic have both reached commercial scales that would be substantial for established technology businesses. Their financial information, however, remains less standardized and less publicly accessible than that of listed companies required to publish comprehensive periodic financial statements.

This creates a measurement problem for investors evaluating not only the model developers themselves but also the cloud providers, semiconductor manufacturers, infrastructure operators, and energy companies supplying their expansion.

A more complete assessment would begin with recognized revenue under clearly disclosed accounting policies, separating direct sales from third-party distribution and identifying the roles played by cloud partners.

It would then examine customer spending, gross profit, operating expenditure, capital requirements, and cash generation, rather than assuming that a rapidly increasing run rate necessarily implies a similar increase in economic value.

........

Financial indicator	What investors should examine
Recognized revenue	Actual revenue earned during a defined reporting period
Annualized revenue	Latest business pace, with the calculation period and included sales specified
Gross and net partner sales	Consistent treatment of third-party distribution arrangements
Recurring revenue	Contractual or repeatable income distinguished from variable usage
Gross margin	Revenue remaining after applicable costs of delivering services
Operating cash flow	Cash generated or consumed through ordinary business operations
Capital expenditure and commitments	Current investment and future infrastructure obligations
Customer concentration	Exposure to major enterprise, platform, and cloud customers
Cost per unit of AI service	Computing expenditure relative to delivered usage and value

........

These measures address different aspects of financial performance, and none can be inferred reliably from the difference between two annualized revenue estimates.

Gross margins are especially important because AI services are computationally intensive. Investors need to understand how much incremental expenditure is required to generate additional revenue and whether improvements in model efficiency are reducing the cost of serving customers.

Contractual commitments deserve separate attention. A long-term infrastructure agreement may involve future payments even before all the associated capacity is operational, while contracted customer spending may be recognized gradually as services are delivered.

Cash flow provides a further test of financial sustainability. A company can report strong revenue growth and still consume significant cash if computing expenses, working capital requirements, research expenditure, or infrastructure investments increase faster than collections from customers.

The timing of revenue also matters when evaluating year-end forecasts. OpenAI's reported expectation of reaching approximately $70 billion in annualized revenue by the end of 2026 would imply a substantially higher exit rate than September's narrower $50 billion estimate.

On a comparable annualization basis, that increase would require the monthly revenue pace to rise from approximately $4.17 billion to $5.83 billion, a further 40% increase by year-end.

This would be an ambitious but mathematically distinct target from the previously circulated $70 billion September estimate, which reflected a broader presentation of existing activity.

Whether OpenAI reaches such a target will depend on enterprise adoption, consumer monetization, API demand, product pricing, and available computing capacity. The company has not published a complete public forecast specifying every assumption behind the reported year-end expectation.

For Anthropic, similar questions apply to the sustainability of its cloud-mediated business, the economics of distribution partnerships, and the cost of serving growing demand for increasingly capable Claude models.

The emergence of these questions does not diminish the commercial significance of either company. Rather, it reflects the transition of frontier AI from a research-intensive industry financed primarily on technological expectations toward a large-scale commercial market whose participants must increasingly demonstrate operating efficiency and financial durability.

OpenAI's approximately $50 billion September annualized revenue indicates a rapidly expanding business, while the previously reported $70 billion figure illustrates the risks of comparing different revenue measures without reconciliation. The difference between them does not establish a collapse in demand, just as the higher number would not establish superior profitability.

As AI companies and their infrastructure partners compete for larger amounts of capital, the decisive financial evidence will be standardized revenue recognition, reproducible growth metrics, sustainable margins, and the ability to convert customer demand into cash after the costs of delivering increasingly sophisticated AI systems.

··········

FOLLOW US FOR MORE.

DATA STUDIOS

datastudios.org