Microsoft Discloses AI Sales Metrics That Defy Easy Comparison

Key Takeaways

• Microsoft reported a $37 billion annual AI revenue run rate, while Amazon put AWS AI revenue above $25 billion, but the measures may cover different products and pricing models.
• GAAP and ASC 606 govern revenue recognition without defining "AI revenue," leaving companies broad discretion over attribution and disclosure.
• Microsoft's $24.1 billion in fiscal 2026 revenue from OpenAI suggests substantial customer concentration and raises questions about the rest of its reported AI business.

Hyperscalers are spending hundreds of billions of dollars on AI infrastructure, creating pressure to demonstrate that demand is catching up with investment. Revenue should offer a relatively straightforward answer. In practice, the numbers emerging from Microsoft, Amazon and other large technology companies are difficult to compare.

Microsoft has said its AI business exceeded a $37 billion annual revenue run rate. Amazon reported that its AWS AI business surpassed a $25 billion annual revenue run rate. Both figures represent substantial reported scale, but neither is a standardized GAAP revenue category.

That distinction matters. A run rate generally annualizes revenue observed over a shorter period, so it is not necessarily the same as revenue recognized during a completed fiscal year. It can also be affected by rapid growth, customer consumption patterns and management's decisions about which offerings belong inside the category.

GAAP explains when and how revenue is recognized, not whether revenue should be labeled "AI." ASC 606, Revenue from Contracts with Customers, addresses matters such as performance obligations, transaction prices and the allocation of consideration. It does not prescribe an AI segment or a common AI monetization calculation.

As accounting research published in Deep Quarry explains, companies can decide whether to disclose AI revenue measures, which products to include and how to calculate the resulting figure. That flexibility leaves room for internal reporting, but it limits external comparability.

Microsoft's fiscal 2026 disclosures illustrate the problem. The company reported $24.1 billion of revenue from commercial arrangements with OpenAI. By combining that amount with Microsoft's previously disclosed $37 billion AI annual revenue run rate and its projected growth, Bloomberg estimated that OpenAI represented roughly 50% to 70% of Microsoft's AI business during the year.

The estimate provides investors with an unusual view into the composition of a hyperscaler's AI activity. It also points to concentration risk. If OpenAI contributes more than half of Microsoft's AI sales, changes in that relationship, OpenAI's computing requirements or the commercial terms between the two businesses could materially affect the metric.

Still, the $24.1 billion requires careful interpretation. Revenue from Microsoft's OpenAI arrangements may include both OpenAI's consumption of Azure computing services and revenue-sharing elements. Separating those components would give investors a clearer picture of underlying infrastructure demand and partnership economics.

Then there is the other 30% to 50%. It could include Azure AI services sold to other customers, Microsoft 365 Copilot, GitHub Copilot and AI capabilities embedded across Microsoft's product portfolio. Microsoft has not provided enough detail to calculate each component independently.

Bundling makes the exercise even harder. If Microsoft sells a Microsoft 365 subscription containing AI functionality, should the entire subscription count as AI revenue? Only the incremental Copilot charge? Or an internal allocation based on the estimated value of each feature?

Amazon may make different choices when calculating AWS AI revenue. Usage-based model training, inference, specialized computing capacity and managed AI services could all qualify, but the breadth of the category determines what the $25 billion figure actually represents. Without consistent definitions, the apparent $12 billion gap between Microsoft and Amazon says less than it initially seems.

Other hyperscalers take still different approaches. Google discusses AI's contribution to Cloud growth and reports operating indicators such as token usage. Oracle emphasizes AI-related cloud demand and remaining performance obligations. Meta generally describes AI through its effect on advertising engagement and performance rather than presenting a separate revenue total.

Investor pressure is unlikely to fade. Reuters reported that J.P. Morgan raised its year-end S&P 500 target partly because of increased confidence that AI investment by Microsoft, Amazon and Alphabet would support faster revenue growth, alongside cloud expansion and larger order backlogs. Following Microsoft's July 2026 earnings, the company's chief executive officer also demonstrated an "ROIC Intelligence App" based on a Morgan Stanley report about hyperscaler returns on invested capital.

More consistent disclosure could help. Companies could identify included products, distinguish annualized run rates from recognized revenue, explain treatment of bundled subscriptions and provide period-to-period reconciliations. Customer concentration and gross-margin context would add another useful layer.

Until then, AI revenue will remain a management-defined measure rather than a common accounting yardstick. The figures show that monetization is occurring. They do not yet show, on a comparable basis, which hyperscaler is earning the strongest return from it.