OpenAI's third-quarter revenue surged by nearly 50%, reaching 10 billion U.S. dollars.
On November 10, Bloomberg cited sources familiar with the matter saying that in the three months ending September, OpenAI's quarterly revenue crossed the $10 billion mark for the first time. Compared with the previous quarter, its revenue increased by nearly 50%; compared with the same period last year, its revenue was nearly three times that of the same period last year, representing a year-on-year increase of nearly 200%. These undisclosed financial details were disclosed during an investor conference call on Tuesday.
With quarterly revenue reaching $10 billion, this is a milestone figure for any technology company that has been established for only ten years. After its revenue climbed to about $13 billion throughout 2025, OpenAI showed an astonishingly steep growth curve in the first three quarters of 2026: $5.7 billion in the first quarter, $6.7 billion in the second quarter, and a direct jump to $10 billion in the third quarter.
Judging only from this report card, OpenAI seems to have completely shattered the outside world's doubts that "large AI models cannot be commercialized". However, in the current context of Wall Street, behind the $10 billion revenue lies a more ambitious and more dangerous capital game. This is not just a story about how much money a company can make, but also the ultimate test of whether the entire industry can achieve self-rescue before the computing power black hole completely devours its cash flow.
The "steepening" of the revenue curve and the essence of commercialization
The explosive surge of OpenAI's revenue in the third quarter mainly benefits from the resonance of several core variables.
The first is the full-scale explosion of the enterprise-level market. As Greg Brockman, President of OpenAI, previously admitted at the WSJ Tech Conference, the company went through a "quite difficult" period in the first half of this year, and was once suppressed by Anthropic's Claude Code especially in the field of agent programming. But in the third quarter, according to public reports, with the launch of GPT-5.6 Sol and ChatGPT Work around July 2026, OpenAI regained part of its lost ground in enterprise-level APIs and productivity tools. The proportion of revenue contributed by enterprise customers has rapidly jumped from about 30% a year ago to the majority. In-depth cooperation with large financial institutions such as JPMorgan Chase has become an important driving force for the growth of its enterprise-level revenue.
The second is the steady expansion of the C-end subscription ecosystem. Despite the increasingly fierce competition in the global AI application market, ChatGPT's status as a super entry is still unshakable. Its huge user base not only contributes stable subscription cash flow, but also provides OpenAI with extremely valuable reinforcement learning data.
The nearly 50% quarter-on-quarter growth rate from $6.7 billion to $10 billion illustrates one thing: after years of "trial period", large models are truly being embedded into the production processes of the global commercial system. Enterprises are willing to pay high fees for computing power and agents, and AI is no longer just a toy for tech giants, but is turning into infrastructure like water, electricity and gas.
The quality of the $10 billion revenue: Profits swallowed by costs
However, on the other side of the booming revenue, there is OpenAI's equally suffocating cost bill.
Quarterly revenue of $10 billion is still a drop in the bucket compared with OpenAI's current computing power commitments that amount to hundreds of billions of dollars. According to previously disclosed information, OpenAI's committed cloud computing infrastructure expenditure by 2030 is about $750 billion. Bernstein estimated after reconstructing the model based on the assumption that the contracts will be delivered on time that the actual expenditure is about $614 billion, with the peak appearing in 2031 at about $184 billion.
This extreme leverage ratio means that even if OpenAI can maintain a quarterly revenue of $10 billion and an annual revenue growth rate of $40 billion, the operating cash flow it generates will absolutely not be able to cover the huge capital expenditure in the short term.
The training and inference of AI models is a bottomless arms race. In order to maintain the leadership of its models, OpenAI must spend heavily on NVIDIA's new generation of chips; in order to meet the inference needs of ChatGPT and APIs, it has to lease massive data centers from new cloud vendors such as Microsoft, Oracle and even CoreWeave; in addition, self-developed chips (such as the project in cooperation with Broadcom) and the salaries of its huge top R&D team are consuming cash at an alarming rate.
According to the company's internal forecast, every $1 of revenue earned corresponds to about $1.8 of negative free cash flow. This financial model has not been fundamentally reversed by the surge in revenue in the third quarter. The absolute growth of revenue only brings an increase in the "financing limit" rather than the realization of "self-sufficiency". Its operating loss and negative free cash flow are still difficult to reverse in the short term, and there is no public timetable for net profit to turn positive. According to the internal forecast in its financing materials, the cumulative negative free cash flow from 2026 to 2030 will reach $278 billion.
The patience of the capital market and the $1.25 trillion "anchor"
Facing such a company with surging revenue but continuous huge losses, the attitude of the capital market is undergoing subtle differentiation.
Just this week, OpenAI just completed a new $11 billion financing, which is the initial part of a round of financing with a valuation of $1.25 trillion. However, compared with the epic private placement of $122 billion with a valuation of $852 billion in March this year, the size and valuation expectation of this new fund have shrunk. The original plan was to raise $30 billion with a higher valuation expectation, but it was finally adjusted to raise about $20 billion at a valuation of $1.25 trillion.
If the private market's pricing of OpenAI still retains some of the "founding team aura" and "AGI faith", the attitude of the public market is much more ruthless. Anthropic, which is about to land on NASDAQ, is acting as the "touchstone" for the public pricing of this round of AI assets.
Anthropic's IPO pricing document implies a maximum implied valuation of about $1.4 trillion; but Bloomberg reported that the valuation range discussed by potential buyers after the recent market turmoil was once lowered to $1.2 trillion to $1.3 trillion. What is more thought-provoking is that Anthropic plans to allocate as much as 20% of its IPO shares to retail investors. Some analysts pointed out that this is because "retail buyers may be less price-sensitive than institutional buyers". In other words, in the face of the valuation of about $1.4 trillion, smart institutional funds are becoming cautious, and some views hold that underwriters need to rely on retail sentiment to fill the subscription gap.
This "downward" financing strategy targeting retail investors exposes the liquidity anxiety of AI giants when they sprint for trillion-dollar valuations. When Anthropic needs to expand the retail quota to finance its $895.4 billion computing power bill, OpenAI's $1.25 trillion valuation also needs to undergo strict scrutiny from the public market.
The $700 billion bill and the countdown to the 2027 IPO
As early as October 2025, OpenAI completed a key organizational restructuring, removing the legal obstacles for IPO. The non-profit foundation holds about 26% of the equity and retains control, and the for-profit entity is restructured into a PBC. In September 2026, Sam Altman explicitly ruled out the possibility of launching an IPO in 2026 on the grounds that "AI safety and alignment work has not been completed" and the current time is not appropriate; Brockman publicly stated at the Wall Street Journal Tech Conference in early November 2026 that the company expects to be ready for listing in 2027.
But the clock of the capital market does not move at Altman's will.
According to OpenAI's current quarterly revenue ($10 billion) and cash burn rate, the hundreds of billions of dollars raised in the last round of financing are expected to run out around 2028. On the premise that the internal forecast is fulfilled, the 2027 IPO is more like a "life-sustaining battle" that must be won rather than an optional choice.
In this battle, the $10 billion revenue is both a weapon and a weakness.
It proves that OpenAI's commercialization has achieved a closed loop, and proves that ChatGPT is not a bubble. But this $10 billion is too small in the face of the $750 billion computing power commitment. When Silicon Valley's patience is gradually exhausted by high depreciation and interest, and institutional investors are no longer willing to write checks for "science fiction novels", OpenAI must prove to the public market that it can convert the growth rate of revenue into the growth rate of profit at a predictable rate.
Conclusion
Quarterly revenue of $10 billion, nearly three times that of the same period last year, representing a year-on-year increase of nearly 200%. This is a glorious moment belonging to OpenAI, and it is also the most outstanding commercial achievement of the entire large AI model industry so far.
But behind this figure, we see an extremely steep revenue curve racing against an even steeper expenditure curve. OpenAI, Anthropic and other players are using unprecedented capital density in the history of human commerce to build a future centered on agents, computing power and artificial general intelligence.
How much premium the public market is ultimately willing to pay for this future determines the end point of this race. $10 billion is an excellent start, but in the face of trillion-dollar valuations and hundreds of billions of dollars in bills, it only means that the real test has just begun. Before the computing power black hole is completely filled, no AI giant can stop its galloping pace.
This article is from the WeChat official account "Tech Business", written by Zisen and Bowen, and authorized for release by 36Kr.