Author：Wall Street CN

The internal tension between OpenAI's aggressive expansion strategy and financial prudence is coming to light.

According to The Information,CEO Sam Altman has privately expressed his hope to complete the IPO as early as the fourth quarter of this year, while CFO Sarah Friar had earlier this year expressed a completely opposite view to her colleagues, believing that the company was not yet ready to go public.

Friar's concerns point directly to the company's financial reality: OpenAI has committed to spending over $600 billion on server rentals over the next five years, while warning investors that the company's cash burn by 2030 will be more than double previous forecasts, potentially exceeding $200 billion.

According to a person who has spoken with Friar,She remains skeptical about whether the company's slowing revenue growth can support such a large spending commitment. Meanwhile, competitor Anthropic is eroding OpenAI's market share, further exacerbating external pressure.

In response to reports of a rift in their relationship, Altman and Friar issued a joint statement saying that they are “completely aligned” on the principle of making persistent computing power a core strategy of OpenAI, and that they have been “directly involved in every major computing power decision” over the past year.

Disagreements exist regarding the IPO timeline.

Despite Friar's reservations, OpenAI has begun pre-IPO preparations. The company has hired two law firms, Cooley and Wachtell Lipton Rosen & Katz, and has made informal contacts with the IPO underwriting teams of Goldman Sachs and Morgan Stanley.

Altman's desire to go public also carries a clear competitive motive; he has privately stated his hope that OpenAI's IPO will be completed before Anthropic, which is currently discussing a listing in the fourth quarter of this year. If the IPO goes ahead, the deal could potentially rank among the largest IPOs in history.

Friar's resistance to a quick IPO was not without precedent. In an interview with The Wall Street Journal last November, she publicly stated that an IPO was "not currently in the plans" because the company was still focused on "fitting the company to its current size."

The internal power structure is quietly changing.

At the organizational level, an unusual change had already occurred: since August of last year, Friar no longer reports directly to Altman, but instead reports to Fidji Simo, the head of applications. A CFO reporting directly to an executive other than the CEO is a rare arrangement in large companies.

According to multiple sources who worked closely with the two individuals,Altman has excluded Friar from several important discussions involving financial planning.

One example is that Altman did not invite Friar to a recent meeting with a major OpenAI investor to discuss server spending, even though Friar had attended a previous meeting on the same topic.

One attendee said her absence was "obvious and embarrassing." Another source indicated that Friar was also not invited to an executive meeting earlier this year involving significant financial decisions.

Exorbitant computing power promises conceal hidden financial risks.

OpenAI's current computing power spending plan is unprecedented in scale.

OpenAI has signed server leasing contracts totaling approximately $665 billion, covering:

Oracle (approximately $300 billion, five-year term, starting in 2027);

Microsoft ($250 billion, by 2032);

Amazon Web Services (approximately $138 billion, eight-year term);

CoreWeave ($22 billion, five-year term);

Cerebras ($10 billion) and other partners.

These commitments are not typical cloud computing contracts. Friar explained that AI data center construction cycles can take several years, and OpenAI must book capacity in advance.

"I must make decisions today to ensure we have sufficient computing power in 2028, 2029, and even 2030. If I don't place orders today, data centers will never exist."

In one case, as previously reported by The Information, OpenAI and Oracle signed a risk-sharing agreement regarding cost overruns in data center construction, a rare clause in cloud computing customer contracts. OpenAI has since shelved its plans to build its own data centers.

The dual pressures of revenue growth and cash burn

The external competitive landscape is tightening rapidly. Anthropic has surpassed OpenAI in selling AI models to enterprises and application developers, while Google Gemini continues to erode ChatGPT's dominance in the consumer chatbot market.

OpenAI raised its revenue forecast for the next five years by 27% this year, but also privately disclosed to investors that its cash burn by 2030 will be more than double the forecast made last summer.

In addition, the company informed investors that last year's gross margin was lower than expected because user demand exceeded expectations, forcing the company to temporarily purchase computing power at higher prices.

In February of this year, Anthropic co-founder and CEO Dario Amodei spoke candidly about the dangers of over-investing in data centers on the Dwarkesh Patel podcast, his words implicitly referring to OpenAI.

"If I misjudge, even by just one year... you'll go bankrupt. My impression is that some companies haven't seriously considered this; they don't truly understand the risks they're taking."

This echoes the concerns Friar expressed internally. One person who worked closely with the two described the CFO's situation this way: "She's dealing with an ambitious founder who wants to go all out on spending; it's a tough job."