Share the pie of OpenAI

Microsoft: $140 billion; other investors: $100 billion; employee teams: $125 billion...

After months of negotiations, the highly anticipated restructuring of OpenAI is finally coming to an end, and the total "cake" worth $500 billion has been divided up.

In May this year, OpenAI announced that it would restructure the company into a for - profit entity, removing obstacles for the next round of financing and future IPO. The attitude of Microsoft, as the largest investor, is crucial to the success of this restructuring. On September 11th, OpenAI and Microsoft issued a joint statement saying that they had reached a non - binding agreement, and Microsoft had agreed to OpenAI's restructuring plan. At that time, it was widely believed that Microsoft would obtain approximately 30% of OpenAI's shares, worth $150 billion.

Recently, The Information reported that according to information obtained from investors, Microsoft's final shareholding ratio in OpenAI after the restructuring has been set at 28%, making it the largest shareholder of OpenAI. Based on OpenAI's total valuation of $500 billion, the book value of Microsoft's shares is approximately $140 billion.

Meanwhile, the shareholding ratios of other OpenAI investors have also been determined. Investors from various rounds between 2019 and 2025 will receive shares ranging from 1% to 13%.

Of course, the final agreement has not been officially disclosed yet, and there is still a possibility of changes in the future.

What's going on with OpenAI's restructuring?

In 2019, in order to secure a $1 billion investment from Microsoft, OpenAI designed a complex non - profit - limited profit structure for the company. This structure has been widely criticized, and almost everyone believes that it cannot be maintained in the long run, and restructuring is inevitable.

Currently, what is commonly referred to as OpenAI actually consists of two completely different entities: OpenAI inc., a non - profit organization, and OpenAI LP, a so - called "limited profit company".

OpenAI LP is the operating entity for all of OpenAI's businesses. However, as the name suggests, it is not a real company but a limited partnership. You can think of it as an equity investment fund. Its only general partner (GP) is OpenAI inc., and investors in each round of OpenAI's financing, including Microsoft, are all limited partners (LPs) of this fund.

That is to say, all of OpenAI's investors are not actually shareholders of OpenAI. They do not have any decision - making power, only the right to receive returns. Moreover, OpenAI LP has set a cap on the returns of LPs in its articles of association. The return cap for the earliest investors in OpenAI is 100 times. This means that once their investment returns from OpenAI reach 100 times their investment cost, they will no longer have any rights. As OpenAI's financing rounds progress, this return cap has been continuously reduced. According to Sam Altman, the return cap for recent investors is only in the "single digits".

Pre - setting a cap on investment returns is unheard of in the VC industry. Naturally, investors are not satisfied with such an arrangement.

As OpenAI's financing amount and valuation have increased, this structure has increasingly become an obstacle to financing. Therefore, since November 2024, OpenAI has proposed to restructure the company into a traditional corporation.

According to the current plan, OpenAI LP will be restructured into a "public - benefit corporation". A public - benefit corporation is a company that must clearly state in its articles of association that it prioritizes the pursuit of public interests. Apart from this, it is basically the same as an ordinary company. OpenAI's investors will enjoy full shareholder rights, and the "tightening curse" of return caps will no longer exist. In fact, OpenAI's main competitors, including Grok by Elon Musk and Anthropic, are all registered as public - benefit corporations.

It is worth noting that the shareholding ratios of shareholders after the restructuring are not entirely determined by their investment amounts, and there is no unified calculation rule. Instead, they are finally determined through complex negotiations, a series of compromises, including consultations with regulatory authorities. This "cake - dividing" process is one of the biggest focuses of OpenAI's current restructuring.

It is said that during the entire process, OpenAI basically only negotiated with Microsoft, and other investors were just waiting for notifications.

How to divide the $500 - billion cake?

According to the latest report from The Information, the way to divide this "cake" has finally been determined.

Microsoft's shareholding ratio has been set at 28%, which is higher than its share of return rights under the current structure. In addition, Microsoft will extend its cooperation agreement with OpenAI beyond 2030 (the current agreement expires in 2030).

Based on OpenAI's total valuation of $500 billion, the value of Microsoft's shares can reach $140 billion. As of now, Microsoft's cumulative investment in OpenAI is approximately $13 billion. This means that Microsoft will achieve more than ten times the investment return in just a few years, which will be one of the most successful strategic investments in history.

The shareholding ratio of the non - profit company OpenAI inc. has been set at 27%, slightly lower than Microsoft's. The market value of OpenAI inc.'s shares is approximately $135 billion, which will make it one of the wealthiest charitable organizations in the world. In addition, to meet regulatory requirements, OpenAI inc. will be granted special voting rights and can still maintain control over the new OpenAI company.

OpenAI's employees, including Sam Altman, will receive a total of 25% of the shares, worth $125 billion.

Investors in OpenAI's Series F will receive 13% of the shares. It is reported that in March 2025, OpenAI completed a Series F financing of $40 billion, with a post - investment valuation of $300 billion. Investors include SoftBank, Thrive Capital, Coatue, Dragoneer, etc. Based on this calculation, the current return for Series F investors is approximately 62.5%.

Investors in OpenAI's Series E financing will receive 4% of the shares. In October 2024, OpenAI completed a Series E financing of $6.6 billion, with a valuation of $157 billion. Investors include Thrive Capital, NVIDIA, SoftBank Group, etc. Based on the current valuation, their return is approximately 200%.

Investors who were among the first to invest in OpenAI LP with Microsoft in 2019, including the University of Michigan Foundation, the Reid Hoffman Foundation, Paul Buchheit, the founder of Gmail, and Y Combinator, will also receive 1% of the shares. Although the shareholding ratio seems low, it is still worth $5 billion. Since their total investment is less than $200 million, their return multiple can reach more than 25 times.

So far, all of OpenAI's investors have made huge profits. Moreover, OpenAI's future listing path will be opened, and they can continue to obtain unlimited returns.

Has the dragon - slayer become the dragon?

After OpenAI proposed the restructuring plan at the end of last year, it was met with a deluge of doubts. After the California Attorney General launched an investigation into OpenAI's restructuring, there were even rumors that OpenAI would be forced to leave California.

In 2015, Altman, Musk, Peter Thiel and other leaders in the tech industry jointly founded OpenAI and established it as a non - profit organization. OpenAI's mission was defined as "ensuring that AGI benefits all of humanity". Looking back at these high - sounding statements made by OpenAI, it is hard not to find them hypocritical.

"We never want to make decisions that benefit shareholders. We are the only ones who want to be responsible for all of humanity." "After a certain point, the returns should not belong to a small group of people. Everyone should get a fair share in the AGI era." These are words once spoken by Sam Altman himself.

But now, OpenAI seems to have forgotten these ideas. Most of the wealth created by OpenAI now and in the future will belong to a small number of people and capital, just like any other company.

When Altman, Musk and others founded OpenAI, a direct reason was that they were worried that Google, which they considered an "evil" company, would be the first to invent and control AGI. Now, who can guarantee that OpenAI will not be more "evil" than Google?

From OpenAI's perspective, there are also reasons why it has to do so. When creating the non - profit structure, OpenAI seriously underestimated the future development of the situation and made promises that it was actually difficult to keep.

One of OpenAI's miscalculations was the scale of capital required for AGI.

Sam Altman wrote in a letter to employees that they initially did not think that building AGI would be a "capital - intensive" task. After receiving a donation commitment of $1 billion, they thought that this money would be enough.

But later, OpenAI's capital needs grew larger and larger. Relying on donations was just a drop in the bucket, and OpenAI had to raise funds in a more traditional way.

The second miscalculation was the intensity of competition in the AGI field.

When setting the return cap for investors in 2019, OpenAI thought that it would be the only, or at least the dominant, AGI company. In that case, OpenAI's value would be almost infinite, and it was necessary to limit investors' returns.

In the past few months, OpenAI has been repeating one sentence: "We are no longer the only company that can achieve AGI." When there are many great AGI companies in the world, limiting investors' returns is tantamount to suicide.

In other words, the promises made at the beginning were a bit too grand, and now it's hard to live up to them.

Will burn $115 billion in the next five years

For OpenAI at present, financing is still the lifeline of the company.

According to US media reports, the Series F financing of OpenAI has not been fully received. Investors such as SoftBank have only transferred the first installment of $10 billion to OpenAI, and the remaining $30 billion will not arrive until the end of 2025. Moreover, there is a pre - condition for this $30 billion, which is that OpenAI can successfully complete the restructuring.

If OpenAI fails to complete the restructuring before the end of 2025, SoftBank can choose to reduce the investment amount from $30 billion to $20 billion.

It can be seen that this restructuring is extremely crucial for OpenAI. If it fails, it will have a great impact on the company's cash flow.

Since 2019, OpenAI's expenses have been growing exponentially. Although OpenAI has never disclosed its financial data, it is certain that its expenditure is now an extremely terrifying figure.

In early 2023, OpenAI released GPT - 4. At that time, the industry estimated that the training cost was over $100 million, which was considered an astronomical figure. By July 2024, a financial document obtained by The Information showed that OpenAI had further increased its annual model training budget to $3 billion.

Model training is just a small part of OpenAI's expenditure. According to this document, OpenAI's total expenditure in 2024 can reach $8.5 billion. The specific composition is as follows:

· Training costs increased to $3 billion;

· Inference costs increased to $4 billion;

· Salaries increased three - fold to $1.5 billion;

Although OpenAI's revenue has also been growing explosively, reaching approximately $4 billion in 2024, it still lags far behind the growth of expenditure, and the annual loss can reach $4.5 billion. It is worth mentioning that OpenAI has received significant discounts from Microsoft Azure for inference and training. Even so, the gross profit of ChatGPT is still negative, and each visit makes OpenAI lose more money.

In 2025, OpenAI's revenue continued to grow significantly. At the beginning of June, OpenAI announced that its annualized revenue (the revenue of the most recent month multiplied by 12) had reached $10 billion, doubling compared to the end of last year. OpenAI's goal is to reach an annual revenue of $12.7 billion in 2025.

But this is still dwarfed by OpenAI's expenditure.

At the beginning of September, OpenAI told the media that its cash consumption is expected to reach as much as $115 billion by 2029. Next year, OpenAI's cash consumption will double to over $17 billion, continue to double to $35 billion in 2027, and reach $45 billion in 2028.

Where will this money be spent? In addition to building larger and more powerful AI models, OpenAI has recently been reported to be collaborating with Broadcom to develop its own AI chips, which are expected to be launched next year. It is said that Broadcom has received orders worth $10 billion. The most capital - intensive part is the construction of large - scale data centers and power infrastructure. Sam Altman said in August that OpenAI's computing power will double in just the next five months, not to mention the "Stargate" project with an alleged investment of $500 billion.

So far, OpenAI's total financing has exceeded $62 billion, more than any other competitor, but it is still far from enough. It is no exaggeration to say that OpenAI has become an unprecedented money - guzzling monster in human history, burning many times more money than the Apollo moon - landing program in the past.

The people who initially founded OpenAI once believed that "AGI should only be in the hands of a few trustworthy people who can handle it." Now it seems that AGI is so expensive that it will inevitably end up in the hands of the most powerful capital.

If AGI can really be achieved, what will OpenAI look like at that time?

This article is from the WeChat official account "China Venture Capital". Author: Tao Huidong. Republished by 36Kr with permission.