The short version
OpenAI’s revenue just hit an incredible $10 billion for the third quarter. That’s a massive number, showing explosive growth.
But they also posted an operating loss of $10.9 billion in the same three months. They are spending money even faster than they are making it.
This whole situation reveals the brutal, astronomical cost of building frontier AI and makes you seriously question the long-term plan here.
So what are the actual numbers?
The latest figures are staggering. The recent investor disclosure from OpenAI paints a wild picture of a company growing at a breakneck pace while also burning cash at a historic rate. For the third quarter of 2026, revenue hit $10 billion. That’s up nearly 50% from the previous quarter and almost triple what it was this time last year. It’s hard to overstate how fast that is. The growth puts the company on an annual revenue run rate of roughly $50 billion.
That revenue is reportedly split pretty evenly, 50/50, between their enterprise clients using the API and dedicated services, and their consumer-facing products like ChatGPT Plus subscriptions. This is a healthy sign. It shows they are not entirely dependent on one market segment.
But then you see the other side of the ledger. The company’s adjusted operating loss widened to $10.9 billion. Think about that. For every dollar in revenue they brought in, they spent more than a dollar just keeping the lights on and the models running. This isn’t just a small loss in the pursuit of growth. It is a chasm. The loss grew from $10.1 billion in the second quarter, meaning that as revenue scales, their costs are scaling right alongside it, and even a little faster.
Why is OpenAI losing so much money?
One word: compute. The single biggest driver of OpenAI’s colossal losses is the mind-bending cost of the hardware required to train and run their models. Frontier models like the GPT series are not built on a handful of servers in a closet. They are built using massive clusters of tens of thousands of specialized GPUs, mostly from Nvidia. Each one of those chips is a power-hungry, expensive piece of equipment. Now imagine buying them by the truckload.
Training a new, state-of-the-art foundation model is an undertaking that costs hundreds of millions, or even billions, of dollars in compute time alone. And that’s just to get the model built. After training, you have inference costs—the cost of actually running the model to answer prompts from millions of users and enterprise API calls. Every time someone uses ChatGPT or an application built on OpenAI’s API, it consumes a slice of that computational power. With their scale, that adds up to an astronomical energy and hardware bill. This isn’t a software business with software margins. It’s a hardware-intensive business disguised as a software one.
Beyond the raw compute, there are other significant expenses. OpenAI is in a fierce talent war with Google, Anthropic, Meta, and others. They are paying top-of-the-market salaries and equity packages to attract and retain the world’s best AI researchers and engineers. This adds hundreds of millions in personnel costs. You also have the standard costs of a rapidly growing global company: office space, marketing, legal, and all the infrastructure needed to support thousands of employees. But make no mistake, the GPU clusters and the electricity to run them are the primary black hole for cash.
Is this growth model even sane?
On the surface, it looks like madness. A company that loses more than a billion dollars a month seems like a house of cards. But from another perspective, it’s a calculated, high-stakes bet that we’ve seen before in tech, just amplified to a terrifying new scale. This is the growth-at-all-costs playbook. The goal is to capture the market, become the default platform, build an ecosystem, and achieve a level of scale and data advantage that no one else can touch. Only then do you focus on optimizing for profitability.
OpenAI and its primary backer, Microsoft, are betting that artificial general intelligence, or something close to it, is such a transformative technology that owning the leading platform will be one of the most valuable positions in the global economy. In that context, burning through billions of dollars a year is the cost of entry. It’s the investment required to secure a trillion-dollar prize. The company’s new CFO, Sarah Friar, who came from Square and Nextdoor, is now tasked with navigating this financial tightrope. Her job is to fund this immense ambition without letting the burn rate spiral completely out of control.
Still, the question remains. What if the costs don’t level off? In most industries, scale brings efficiency. For AI, it seems that pushing the frontier to the next level of capability requires exponentially more data and compute. If building GPT-5 costs 10 times more than GPT-4, and GPT-6 costs 10 times more than that, you reach a point where the financial model simply breaks. At some point, the value generated by the models has to start dramatically outpacing the cost of creating them. Right now, that’s not happening.
What does this mean for the competition?
The numbers from OpenAI are a chilling message to every other company in the AI space. The price of admission to the frontier model club is billions upon billions of dollars in annual losses. This puts immense pressure on Anthropic, which is currently on its IPO roadshow. As Anthropic executives pitch their company to investors, they will inevitably be compared to OpenAI. They will have to explain how they plan to compete with a rival that is willing and able to sustain such massive losses, backed by one of the largest corporations on the planet.
This financial reality creates a powerful moat around the top players. It effectively ensures that only a tiny handful of companies—namely OpenAI/Microsoft and Google—have the resources to build these foundational models from scratch. Everyone else, from startups to mid-sized tech firms, is relegated to a different tier of competition. They can fine-tune existing open-source models or build applications on top of the APIs provided by the giants. They can innovate on product and user experience, but they cannot compete on the raw capability of the underlying model.
This dynamic could lead to a consolidation of power that we haven’t seen before. If the future of technology is built on a few core AI platforms, the owners of those platforms will have immense influence. It also raises the stakes for anyone betting on an open-source alternative. While models like Llama are impressive, they are still a step behind the cutting edge, and the cost to close that gap is proving to be a barrier for almost everyone.
What should you do about it?
If you’re a business leader using AI, these numbers should be a wake-up call. The AI services you’re using are likely heavily subsidized. The price you pay for an API call does not reflect the true, all-in cost of providing that service. This means you should be prepared for prices to change. As OpenAI faces pressure to get its finances in order, it’s logical to expect they will look for ways to increase revenue, which could mean higher API rates, new pricing tiers, or a stronger push towards higher-margin enterprise contracts.
Your strategy should be to focus on building real value and ROI from your AI implementations. Don’t just use AI for flashy demos. Use it to solve concrete business problems, save money, or generate new revenue. The more tangible the value you’re creating, the easier it will be to justify the cost, even if that cost goes up. It’s also wise to avoid being completely locked into a single vendor. While migrating from one AI platform to another is not trivial, understanding the landscape and having a potential Plan B is just good strategic planning.
For an individual user, this is a moment to appreciate the current AI boom. We are living in a period of incredible access to powerful technology at a very low cost, funded by venture capital and corporate strategy. That might not last forever. As the financial realities set in, we may see more services move behind paywalls or become more limited in their free offerings. The key is to use these tools to learn, build, and create while the getting is good. The skills you develop now will be valuable no matter how the market shifts.
FAQ
How much money did OpenAI make in Q3 2026?
OpenAI’s revenue was $10 billion in the third quarter of 2026, putting it on a projected annual run rate of about $50 billion.
Why did OpenAI lose more money than it earned?
OpenAI’s adjusted operating loss was $10.9 billion because its expenses, driven primarily by the immense cost of computing power for training and running its AI models, were greater than its revenue.
Is OpenAI going to go out of business?
It is highly unlikely in the near future. The company is heavily funded by Microsoft and other major investors who see the losses as a strategic investment to dominate the future of AI. The current financial model, however, is not sustainable in the long run.
What does this mean for OpenAI’s competitors like Anthropic?
It sets a very high bar for competition. It signals to the market and potential investors that competing at the frontier of AI requires a willingness to burn through billions of dollars annually, which may consolidate the market to a few giant players.
Are the AI services I use going to get more expensive?
It’s a strong possibility. The current prices for many AI services are subsidized. As companies like OpenAI face pressure to become profitable, they will likely raise prices or change their pricing models to better reflect the true cost of their services.