#OpenAI第三季度营收升至100亿美元 $10 Billion in One Quarter, $50 Billion Run Rate: OpenAI’s AI Business Is Scaling Faster, but Profitability Is Still the Real Test

OpenAI has reached a milestone that would have seemed extraordinary just a few years ago: approximately $10 billion in revenue in a single quarter. But the most interesting part of its November 10 financial update is not the headline number. It is the contrast between accelerating customer demand and the enormous cost of building and operating advanced AI.

The business is getting bigger. Enterprise customers are spending more. Yet the losses are also increasing. That is the tension investors should focus on when evaluating the next stage of the AI industry.

The revenue acceleration is real

OpenAI reported third-quarter 2026 revenue of approximately $10 billion, representing nearly 50% growth compared with the second quarter. That is a significant acceleration from the approximately 18% quarterly growth reported for Q2.

The company’s revenue was also almost three times its level in Q3 2025, according to reports based on information shared with investors.

CFO Sarah Friar said OpenAI exited September with an annualized revenue run rate of roughly $50 billion. This figure needs to be understood correctly: it is a run-rate estimate based on the business's recent revenue pace, not $50 billion already earned during 2026 and not a guarantee of future revenue.

The distinction matters because a rapidly growing monthly or quarterly business can produce a very large annualized figure without having completed a full year at that level.

The bigger change is who is paying

OpenAI began with a powerful consumer product. ChatGPT created mass-market demand, but the latest figures suggest that enterprise customers are becoming an equally important part of the business.

According to Friar, enterprise revenue doubled quarter over quarter, bringing the revenue mix to approximately 50% enterprise and 50% consumer.

That is a major shift in the commercial profile. Businesses are not just experimenting with chatbots; they are buying tools designed to fit into actual workflows.

The reported growth drivers include coding products, cybersecurity applications and AI agents such as Dots. These products target tasks where companies may be willing to pay for productivity, automation and specialised assistance.

Small and medium-sized businesses also contributed to the expansion. OpenAI said its SMB offering grew more in Q3 than it had during the previous six quarters.

This matters because enterprise adoption can create recurring contracts and deeper integration into company operations. But customer retention, usage costs and the ability to deliver measurable business value will determine how durable that revenue becomes.

The uncomfortable number: $10.9 billion

Here is where the story becomes more complicated.

OpenAI’s reported adjusted operating loss increased to approximately $10.9 billion in Q3, compared with $10.1 billion in Q2. That is an additional $800 million in losses even as revenue rose sharply.

There is progress in the margin figure. Its adjusted operating margin improved from approximately -150% to -109%.

In simple terms, the business is generating more revenue relative to its adjusted operating costs than it did in the previous quarter. But a margin of negative 109% still means the company remains far from adjusted operating profitability.

The two numbers should be read together:

Revenue growth shows that customers are paying for AI.

The widening operating loss shows that the cost of delivering and expanding those services remains extremely high.

These are reported adjusted figures, not a substitute for a complete audited financial statement. They should not be treated as identical to standard GAAP operating results.

Why selling more AI does not automatically mean earning more

An AI subscription is not like selling a digital product once and walking away. Every new model, customer and workflow can require ongoing computing resources.

Training advanced models requires large-scale infrastructure. Serving millions of users and enterprise workloads requires inference capacity every time a model responds or an agent performs a task. Research and development must also continue if a provider wants to remain competitive.

That creates a difficult economic equation: revenue must grow faster than the combined cost of computing, infrastructure, research and customer delivery.

OpenAI has not provided a complete public breakdown of every cost category in the reported figures, so assigning an exact share of its losses to GPUs, training or inference would be speculation.

The business question is whether model efficiency, infrastructure scale and higher-value enterprise contracts can gradually improve the economics of each unit of AI usage.

Anthropic makes the comparison more interesting

OpenAI is not competing in an empty market. Anthropic reported approximately $17 billion in Q3 revenue, ahead of OpenAI’s $10 billion figure. Its quarterly revenue growth was about 47%.

However, comparing the two companies requires care. Anthropic’s updated filing reported a $567 million operating loss for Q3, while its adjusted operating income was approximately $2.5 billion. It also reported a much larger $23.3 billion net loss, substantially affected by accounting charges and other items.

These figures use different reporting and adjustment definitions from OpenAI’s reported adjusted operating loss. They cannot be compared as if they were identical measures.

Still, the contrast is important: both companies are attracting substantial commercial demand, but their reported operating economics look very different. It reinforces why revenue leadership alone does not determine which AI business has the stronger long-term model.

The wider AI value chain

OpenAI earns money from consumer subscriptions, enterprise products and API usage. Microsoft benefits from its strategic relationship with OpenAI and demand for cloud infrastructure. NVIDIA supplies GPUs and computing technology used across the AI ecosystem.

These are different business models. An AI application provider monetises software and services; a cloud provider monetises computing capacity; a chip supplier sells the hardware powering that capacity.

AI growth can create opportunities across all three layers, but each faces its own investment, competition and margin pressures.

My conclusion: demand is proven, efficiency is not yet settled

OpenAI’s Q3 results demonstrate that AI has moved beyond curiosity. A $10 billion quarter, nearly 50% sequential revenue growth and a doubling of enterprise revenue show that customers are paying for increasingly practical applications.

But the $10.9 billion adjusted operating loss prevents this from being a simple success story. A roughly $50 billion annualized run rate sounds impressive, yet scale only becomes a durable advantage when the business can convert that scale into sustainable margins.

The next AI race may not be won by the company with the largest model or the biggest quarterly revenue figure. It may be won by the company that delivers useful intelligence at a cost customers will pay for, while keeping its own computing and operating expenses under control.

OpenAI has demonstrated commercial momentum. Its next challenge is proving that this momentum can become a profitable business model.