According to news on April 28, according to the Wall Street Journal, OpenAI has recently failed to achieve its goal of 1 billion weekly active users and revenue as expected. According to people familiar with the matter, the slowdown in growth has caused some management to worry about whether OpenAI can still support its huge data center spending base with the current revenue growth rate.
Chief Financial Officer (CFO) Sarah Friar has sounded the alarm to the core executive team, people familiar with the matter said. She pointed out that if revenue growth continues to be lower than expected, the company's ability to perform on computing power procurement contracts may face challenges in the future.
Meanwhile, boards of directors have significantly tightened their scrutiny of data center-related deals in recent months. Against the background of slowing business growth, CEO Sam Altman is still going around to hoard computing power. This radical move is raising questions from the board of directors.
As OpenAI heads toward an IPO at the end of this year, internal scrutiny of capital expenditures is becoming increasingly stringent. According to people familiar with the matter, Freer and other senior management teams are working to control costs and strengthen financial discipline, which to some extent conflicts with Altman's expansion ambitions.
However, Altman and Freer tried to quell outside speculation in a joint statement: "We are completely aligned on the strategy of hoarding as much computing power as possible and are working hard to do so." The two emphasized that any rumors about management disagreements or intentions to slow down the deployment of computing power are "pure nonsense."
Altman has long regarded computing power as the core factor restricting the development of OpenAI, and has actively locked in data center capacity. In the last year alone, he has presided over a series of blockbuster deals. This not only burdens OpenAI with a long-term capital expenditure commitment of up to US$600 billion, but also deeply binds the fate of most of the technology industry to the success or failure of OpenAI.
This kind of "buy everything" computing power gamble at any cost was based on ChatGPT's overwhelming growth myth, and was fully endorsed by Freer and the board of directors. However, as ChatGPT's growth slowed at the end of last year, company executives began to have new doubts about this strategy.
According to people familiar with the matter, ChatGPT failed to achieve its internal goal of breaking 1 billion weekly active users (WAU) at the end of last year. The company has yet to announce it has reached that milestone, which has upset some investors. In addition, due to the rapid expansion of Google's Gemini model and eroding market share at the end of last year, OpenAI also failed to meet ChatGPT's annual revenue target. At the same time, people familiar with the data said the company is struggling with subscriber churn.
People familiar with its financial situation added that OpenAI missed its monthly revenue targets several times earlier this year as it faced fierce competition from Anthropic in the programming and enterprise markets.
OpenAI recently completed the largest round of financing in Silicon Valley history (raising $122 billion), providing it with a more solid financial foundation. However, analysis pointed out that based on the company’s huge computing power reservation scale, even if the ambitious revenue target can be achieved, the funds are expected to be exhausted in the next three years. In addition, some financing funds also come with gambling conditions, which are subject to specific agreements with partners.
Faced with financial pressure, OpenAI is adjusting its business focus. The popularity of its programming tool Codex is rising rapidly, and the company is cutting costs by cutting other projects such as its video-generating app Sora. Recently, OpenAI also released the blockbuster model GPT-5.5, which has topped multiple industry benchmark tests.
In fact, computing power anxiety has become a common problem in the entire industry. In recent weeks, many AI companies, including Anthropic, have encountered computing power bottlenecks, resulting in increased access prices to AI processors, service outages, and throttling. These problems have angered heavy users of AI products, especially programmers who are used to the system completing tasks smoothly in the past, but are now frequently frustrated.
A recent memo from OpenAI to investors obtained by the Wall Street Journal shows that the company emphasized that it is ahead of Anthropic in terms of computing power reserves, giving it an advantage in reaching users. The memo also responds to recent public innuendos from Anthropic CEO Dario Amodei. Amodei told a recent business conference that some companies have "turned the risk dial too high" on data center investments.
OpenAI counterattacked sharply in the memo: "In hindsight, the so-called 'caution' is more of a serious underestimate of the speed of demand than strict discipline."
In addition to the business-level game, the pace of IPO promotion is also undercurrent. Freer has expressed reservations in recent months about a timetable for a listing before the end of the year, according to people familiar with the matter.
She has repeatedly emphasized to senior management and the board of directors that OpenAI urgently needs to improve its internal control mechanism, and warned that the company is not yet ready to meet the stringent reporting standards of listed companies. However, according to some sources, Ultraman prefers a more aggressive listing timetable.
OpenAI must also solve a series of thorny issues before moving forward with its public listing. Earlier this month, the company's No. 2 figure, Fidji Simo, unexpectedly took a leave of absence due to illness, leaving a leadership vacuum. Separately, a lawsuit filed by Elon Musk, in which he seeks to oust Altman and undo OpenAI's transformation into a for-profit company, went to trial this week.