While everyone was still immersed in the Claude Code source code leak incident, OpenAI came out to grab the headlines again. Just now, OpenAI officially announced the completion of a $122 billion round of financing.

A single round of private placement worth RMB 122 billion is unprecedented in the history of human business. After the completion of the financing, OpenAI’s valuation fell to US$852 billion, just one step away from one trillion US dollars, and the company has only been established ten years ago.

It is worth mentioning that when this round of financing was first announced in February this year, the commitment amount was still US$110 billion, but it ended up being US$12 billion more at the end of the day, indicating that more institutions followed up than expected.

It is generally believed that this is OpenAI's last large-scale private placement before its IPO at the end of the year, and the pace of listing has become increasingly clear.

https://openai.com/index/accelerating-the-next-phase-ai/

Where does the money come from

The main investors in this round of financing are Amazon (50 billion), NVIDIA (30 billion), and SoftBank (30 billion). SoftBank also jointly led the investment with a16z, D.E. Shaw and other institutions.

Microsoft, as a long-standing partner, continues to invest, but this time the specific amount was not disclosed. It is only known that as of the end of last year, Microsoft's cumulative investment in OpenAI has exceeded US$13 billion.

In addition, OpenAI also opened its fundraising to wealthy individual investors through bank channels for the first time, raising approximately 3 billion. ARK Invest's $6 billion flagship innovation ETF also announced the inclusion of OpenAI, with a holding ratio of about 3%. This is also the fund's first investment in a non-listed company.

In fact, some funds managed by T. Rowe Price and Fidelity already hold a small amount of OpenAI shares. The addition of ARK this time further opens up channels for ordinary people to participate.

In short, almost the entire technology circle is supporting OpenAI.

But if you think about it carefully, the logic is actually very simple: after OpenAI takes the money, it still has to buy Nvidia chips and rent servers from Amazon and Microsoft. By investing money, the giants have locked in the world's largest computing customers in advance. This round of financing is not so much about being optimistic about OpenAI, but rather a business that is sure to make money.

For OpenAI, this money is more like the last major replenishment before the IPO.

The book data is really good: there are nearly 900 million weekly active users, more than 50 million paying users, last year's revenue was 13.1 billion U.S. dollars, and the highest revenue in a single month was 2 billion, and the growth rate was four times that of Internet giants such as Google and Meta at the same stage.

However, OpenAI has not yet made a profit, and the rate of money burning has not slowed down at all.

Why should we turn off Sora

Before and after this financing, OpenAI's product rhythm has not stagnated.

They released the most powerful GPT-5.4, which has significantly improved multi-tasking and workflow performance. Codex, a code generation tool, has also been upgraded from a function to an independent programming agent. It currently has more than 2 million weekly active users, which has increased fivefold in the past three months, and its monthly growth rate remains at around 70%.

The performance of the enterprise is also worthy of attention. At present, enterprise services account for more than 40% of OpenAI's total revenue, and it is expected to be equal to that of consumers by the end of 2026.

The number of tokens processed by the API per minute exceeds 15 billion, the use of search functions has nearly tripled in the past year, and the annualized revenue of the advertising pilot project exceeded $100 million in less than six weeks after going online. This is also the signal that OpenAI hopes to send to the outside world: Income sources are becoming more and more diverse, and the subscription fee of ChatGPT is only one of them.

However, right next to this piece of red data, Sora quietly went offline.

When Sora was first released, it indeed caused quite a shock in the film and television circles and creative industries. The video is generated in one sentence, and the picture quality is quite realistic. Many people think this is the most exciting thing about AI technology.

But the computing power consumption of video generation is much higher than that of text generation. Every AI inference, every piece of text generation, and every frame of video rendering actually consumes expensive GPU computing cycles and power. Without free intelligence, every call is a loss of real money.

On the user side, although they find it fun, not many are willing to pay a high price for it.

According to the Wall Street Journal, one of the reasons why OpenAI chose to shut down Sora was because it burned about $1 million every day, but the number of users plummeted from 1 million when it was launched to less than 500,000.

When the retained data is ugly and the commercialization path is unclear, there is naturally no reason to continue this expensive business. So, before reality was overturned, Sora ceased to exist.

Shutting down Sora is just the beginning. OpenAI is still looking at other directions that cost a lot of money and have slow returns, and are preparing to shrink further; concentrating computing power on text models, code generation, and enterprise services, which have stable cash flow, is OpenAI's statement to Wall Street: We know and need to make money.

From "changing the world" to "water, electricity and coal"

OpenAI was founded in 2015 with the original vision of ensuring that general artificial intelligence benefits all of humanity.

In 2019, in order to raise enough research and development funds, the company transformed into a "limited profit" model, established a for-profit subsidiary, and accepted a US$1 billion investment from Microsoft. Although the operating entity has been commercialized, the non-profit OpenAI Foundation still holds about 26% of the shares, nominally continuing the original public welfare mission.

There is a sentence worth noting in the official statement of OpenAI financing: "Building the infrastructure layer of intelligence itself."

A few words actually reveal the change in OpenAI's self-positioning. In the past, they were more concerned about refreshing the outside world's understanding of AI with stunning demos. Now, what they want to do more is to retreat behind the scenes and become a low-level tool that enterprises and individuals cannot live without.

They call this direction "super application" and plan to integrate ChatGPT, Codex, search, browser and other capabilities into a unified entrance, mainly for developers and enterprise users, so that people do not have to jump between a bunch of tools.

The logic behind this is to let the habits of consumers naturally drive the procurement of enterprises, and the two businesses will strengthen each other.

An ordinary user may find it new today and cancel his subscription tomorrow, but a company that runs its core business on the OpenAI model is unlikely to discontinue it at will. The latter is the kind of customer stickiness that Wall Street really wants to see.

In the past few years, OpenAI has seen eye-catching things appear every now and then, including new models, new products, and new possibilities, one after another.

But judging from this round of financing and the closure of Sora, the stage full of surprises may really be coming to an end. The next step may be more like a mature business: some people are in charge of computing power, some are in charge of data, and some are in charge of sales. Everyone takes care of their own business, pays attention to cost control, and pays attention to business implementation.

OpenAI can’t go back to its former self, but it may not have intended to go back in the first place.