OpenAI, why does it need to raise up to $100 billion in financing?

ChatGPT is on a global hot streak, fueled by burning money.

On May 5th, according to The Information, OpenAI's losses have roughly doubled to $540 million in 2022, mainly due to ChatGPT's huge expenses and poaching of many important talents from Google.

Without resources, even the cleverest housewife cannot cook. With the worsening situation of losses, more funds must be prepared. OpenAI CEO Sam Altman said that OpenAI may try to raise as much as $100 billion in the coming years to develop sufficiently advanced AGI and increase the company's profitability.

While other companies are thinking about entering the $100 billion valuation, OpenAI is directly raising funds to that level, which can be called bold.

In fact, OpenAI's financing has been ongoing, and Sam Altman also said that OpenAI may become the company that has raised the most funds in Silicon Valley history. Why is OpenAI burning so much money? Where will the $100 billion in funding go? What kind of capital gamble is behind the financing?

Where will the $100 billion go?

ChatGPT is top-notch in terms of popularity and burning money.

After OpenAI launched ChatGPT in November last year, it quickly created the fastest-growing consumer application in history. In January of this year, ChatGPT's monthly active users had exceeded 100 million, with millions of people using it every day, and the number is still growing rapidly.

With the rapid growth of users, ChatGPT's training costs continue to rise. OpenAI has never publicly disclosed specific cost consumption. But according to external estimates, it has been burning money in chip computing power, training databases, daily operations, and talent salaries.

In the future, the $100 billion that OpenAI wants to raise is likely to be used in these areas.

First of all, ChatGPT needs to create an excellent experience, which is highly dependent on high-performance hardware devices.

Research firm TrendForce pointed out in a report that when OpenAI trained GPT-3, the predecessor of ChatGPT, it used about 20,000 Nvidia A100 GPUs, and the current price of A100 chips is about $10,000, which is equivalent to a cost of $200 million. Some industry insiders estimate that ChatGPT will require more than 30,000 GPUs, as well as thousands of servers and hundreds of standard cabinets with chips.

Just the GPU chip alone, ChatGPT needs at least $300 million.

Secondly, ChatGPT is also highly dependent on software, which is cloud services. In 2020, OpenAI used Google Cloud services, which cost as much as $75 million. After getting closer to Microsoft, it gradually reduced its expenses in cloud service hosting by using the Azure platform, but it still needs to purchase cloud services from the Azure platform, which is also a huge cost. Beyond hardware and software, OpenAI also incurs significant expenses in its daily operations and talent recruitment.

Dylan Patel, chief analyst at research firm SemiAnalysis, has stated that when users use ChatGPT to write cover letters, generate academic plans, and polish personal profiles, OpenAI may burn up to $700,000 per day.

Note that this is per day, not per week.

Talent is the backbone of operations, and this is a huge expense.

Even before OpenAI launched ChatGPT, at least five former Google researchers were involved. In February of this year, media reports revealed that OpenAI had poached more than 12 AI technical experts from Google.

What is the salary expenditure for these high-end talents? According to The New York Times, as early as 2016, OpenAI paid more than $1.9 million to Ilya Sutskever, a top machine learning expert. At the ordinary employee level, according to OpenAI's official recruitment information, the salary range for ordinary ChatGPT software engineers is between $200,000 and $370,000, while the salary range for supervisors is between $300,000 and $500,000.

Peter Drucker, the father of modern management, once said, "The only true resource of an enterprise is people." OpenAI spends a lot of money to recruit talents, also for the subsequent realization of the AGI dream and to form a more sustainable development.

Opportunities attract talents, talents attract technology, technology attracts products, and products attract more and larger opportunities. These four forces pull each other to promote a virtuous cycle, which can accelerate the growth of enterprises, and OpenAI is also following this path.

The gamble behind financing

Under huge costs, another embarrassing situation that OpenAI faces is that its income is too little to cover its expenses.

Last year, OpenAI's revenue was only $28 million, mainly relying on the sale of professional AI software. Compared with the $540 million loss, it is undoubtedly a drop in the bucket. In February of this year, ChatGPT launched a paid version of $20 per month, which is expected to bring hundreds of millions of dollars in growth, but it still cannot offset the losses.

At this point, OpenAI must seek financing in advance to obtain funds to promote long-term development. TechCrunch reported that OpenAI completed a $300 million financing on April 28, with a valuation of $27 billion to $29 billion, becoming the newest and fastest-growing unicorn in the world. Financial investment was participated by top investment institutions such as Tiger Global Management, Sequoia Capital, California Andreessen Horowitz, and New York Thrive.

This is the second financing after Microsoft's $10 billion strategic investment earlier this year.

In fact, before OpenAI accepted Microsoft's $10 billion investment, it had already received $4 billion in investment in more than seven years since its establishment. FT Chinese once made a detailed interpretation of OpenAI's series of financing activities in the article "The Uniqueness of the OpenAI Equity Investment Agreement Design for Launching ChatGPT". According to the profit distribution, OpenAI is participating in a huge gamble, which is divided into four stages.

In the first stage, priority is given to ensuring that the first batch of investors such as Musk, Peter Thiel, and Reid Hoffman recoup their capital and ensure that they do not lose money.

In the second stage, Microsoft will have the right to receive 75% of OpenAI's profits until it recovers its investment of $13 billion;

In the third stage, when OpenAI's profits reach $92 billion, Microsoft's profit distribution ratio will drop to 49%, and the remaining 49% of profits will be shared by other venture capitalists and OpenAI's employees as limited partners;

In the fourth stage, when profits reach $150 billion, Microsoft and other venture capitalists' shares will be transferred to OpenAI's non-profit fund free of charge.

In other words, OpenAI has rented the company to Microsoft and will return it after the profits reach $150 billion. The length of the lease depends on OpenAI's profit speed. Before that, Microsoft will continue to inject various resources into OpenAI.

Some people may say that Microsoft's investment is too risky. Because according to OpenAI's own predictions, it will not reach $1 billion in annual revenue until 2024, and it is very difficult to achieve the profit target of $150 billion in the short term. Microsoft's $13 billion investment may take a long time to recover its cost.

In fact, Microsoft is the biggest winner in OpenAI's financing gamble. By tying strategic investment with ChatGPT, it not only keeps its market value rising and gains greater capital imagination space, but also implants ChatGPT into its various products to optimize product functions and gain new vitality. The total benefits of these are far greater than the investment cost.

If in the next few years, OpenAI can successfully raise as much as $100 billion, becoming the company with the most financing in Silicon Valley history, then Microsoft will be the one who earns the most.