news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

New Intelligence Report

Edit: My ears are so sleepy

【New Wisdom Introduction】Recently, foreign media did the math for OpenAI and found that it may suffer a loss of $5 billion this year.

OpenAI is one of the fastest-growing companies in history, but also one of the most expensive.

Recently, The Information predicted based on internal financial data and analysis from relevant people that OpenAI's losses this year could be as high as $5 billion.

This is based on a combined estimate of OpenAI’s expenses for running ChatGPT and training the next generation of LLMs, as well as hiring and staff costs.

If the judgment is correct, then this "cash-burning machine" valued at $80 billion will need to raise more funds in the next 12 months or so.

Dollar Incinerator

In terms of expenditure, OpenAI expects to spend $4 billion this year to rent Microsoft servers to support the operation of ChatGPT and the reasoning of its underlying LLM, according to people familiar with the matter.

Training costs, including paying for data, could swell to $3 billion.

This is because, as the AI ​​arms race becomes increasingly fierce, OpenAI has accelerated the training speed of new AI as early as last year.

In addition, OpenAI employs about 1,500 employees, and wages cost about $1.5 billion.

Of course, this is just an estimate.

According to internal sources, OpenAI's employee costs in 2023 were approximately $500 million, and the number of employees increased to around 800 by the end of the year.

Today, OpenAI’s headcount has nearly doubled again.

And, if the nearly 200 open positions listed on the website are anything to go by, more staffing is expected to be added in the second half of 2024.

By this calculation, OpenAI's operating costs this year may be as high as $8.5 billion.

Analytics India once analyzed that if OpenAI continues to burn money at a cost of $700,000 a day and fails to make a profit in time, it may go bankrupt as early as the end of 2024.

In terms of revenue, ChatGPT itself could provide around $2 billion per year, while the API could bring in over $80 million per month.

Combined with other sources, OpenAI's monthly revenue recently reached about $283 million.

Based on this, The Information estimates that its annual revenue could be between $3.5 billion and $4.5 billion (depending on sales in the second half of the year).

The good news is that annual revenue is expected to reach a record high of $4.5 billion; the bad news is that the cost will be at least $8.5 billion.

That means OpenAI's losses will be between $4 billion and $5 billion.

This analysis explains why OpenAI CEO Altman describes the company as “the most capital-intensive startup in the history of Silicon Valley.”

At the same time, this also means that OpenAI will soon need to continue raising funds because the tens of billions of dollars investment from Microsoft will soon be burned out.

Anthropic VS. OpenAI

Although OpenAI has been burning through cash, it is in a better position than some of its competitors.

Because it gets a discount when it pays Microsoft to rent those energy-hungry servers equipped with expensive Nvidia GPUs.

Financial sponsor Microsoft provided OpenAI with a heavily discounted leasing rate, which basically only covered the cost of operating these servers.

Compared to OpenAI, its biggest competitor among startups is Anthropic.

Anthropic's revenue is estimated to be one-fifth to one-tenth of OpenAI's.

Earlier this year, Anthropic executives predicted that the company would burn through more than $2.7 billion in cash for the full year, including $2.5 billion in costs alone.

Meanwhile, Anthropic expects to reach annualized revenue of about $800 million, or $67 million per month, by the end of this year.

However, Anthropic shares revenue with Amazon, which provides funding and technical support, while also selling AI models to Amazon cloud customers.

It's unclear how much of Anthropic's revenue comes from sales on Amazon, but it's likely a significant portion.

This means that after paying its share, Anthropic could expect annualized revenue of between $400 million and $600 million, or $33 million to $50 million per month.

In other words, although Anthropic has strong growth momentum, it is still inferior to OpenAI in terms of operational efficiency and cost control.

Microsoft friendly price: 4 billion

According to people familiar with the matter, the cluster used by OpenAI is equivalent to 350,000 Nvidia A100s, of which about 290,000 chips (or more than 80%) are used to support the operation of ChartGPT.

Recently, Microsoft signed an agreement with Oracle to rent a huge server cluster to support OpenAI's computing needs.

The cluster is expected to come online in mid-2025, with two-year rents of approximately $5 billion, The Information previously reported.

However, the discount Microsoft gave to OpenAI can be said to be very large - each A100 is only $1.30 per hour.

But because the inference clusters were running basically at full capacity, the full-year cost still reached $4 billion.

Turning the tide

There are two ways to reverse the situation of huge losses: increasing revenue or reducing expenditure.

OpenAI hopes to increase revenue by reducing costs. By updating model structure technology, the cost of running AI models has been greatly reduced.

In addition, OpenAI plans to launch more new products, including new tools such as search engines and computer agents.

If these products are successfully launched, they will be able to handle diverse, multi-step tasks, such as helping shoppers accurately search for products or automating the filling of expense reports and integrating them into accounting software.

OpenAI said it is optimistic about launching its next flagship model before the end of this year.

If this new model is better than the current flagship model GPT-4, which is expected to be launched in early 2023, OpenAI's business growth may make up for its losses.

Another positive sign is that as more and more companies begin to use ChatGPT, OpenAI's revenue growth has exceeded its cost growth.

References:

https://www.theinformation.com/articles/why-openai-could-lose-5-billion-this-year?rc=epv9gi