A recent all-hands meeting led by OpenAI CFO Sarah Friar produced a clear timeline for the company's public market debut, with a target set for 2027 and the possibility of an earlier listing if business momentum accelerates. The presentation boasted impressive metrics, including a 35% quarter-over-quarter revenue run-rate increase and over 20 million weekly active users across its AI coding and office products.

Yet beneath this veneer of success, significant challenges are mounting. The company's quarterly revenue of $6.7 billion represented only an 18% sequential increase, which fell short of some analyst projections, while operating losses ballooned from $9.3 billion to a staggering $12.3 billion. Adding to the discomfort, rival Anthropic reported quarterly revenue of $11.6 billion, surpassing OpenAI for the first time, and has seen roughly twelve department heads depart since the start of 2026, with the most intense wave occurring this month.

Furthermore, despite OpenAI's aggressive price cuts on its models, Chinese AI developers are offering compelling value propositions, leading an increasing number of American firms to shift their business to models like DeepSeek or Kimi, further eroding OpenAI's customer base.

The IPO: A Milestone or Just Another Funding Round?

During the meeting, Friar was measured in her description of the IPO, calling it "not a finish line, but a milestone and another round of funding," citing the $122 billion raised in March as a source of flexibility. OpenAI has been preparing for this for some time, having confidentially submitted a draft registration to the SEC in June and completing a $70 billion secondary share sale this week, allowing employees to cash out at the $852 billion valuation.

Reports indicate that Sam Altman had pushed for a fourth-quarter listing this year, deeming a valuation below $1 trillion "unacceptable." However, Friar, after carefully reviewing the company's spending commitments, privately advised waiting until 2027, arguing that OpenAI is not yet ready for the rigorous disclosure standards of a public company. The external advisory team presented management with two options: list earlier at a lower valuation or wait until 2027 to achieve the target. It appears Altman has ultimately accepted Friar's recommendation, who also acknowledged that Anthropic has filed confidentially and could go public as soon as September, stating, "We run our own race."

Anthropic's Growth Puts OpenAI in an Awkward Position

The contrast between the two companies' recent performances is stark. Anthropic reported second-quarter revenue of $11.6 billion, a 143% increase from the previous quarter and a fourteen-fold surge year-over-year, marking its first-ever quarterly revenue lead over OpenAI. Its annualized revenue run-rate reached $65 billion by the end of July, and it posted an adjusted operating profit of $559 million, becoming the first frontier AI lab to declare quarterly operating profitability. This is set against OpenAI's $12.3 billion quarterly loss.

However, these figures must be viewed with caution. Anthropic is still a private company, and these numbers, based on investor documents and insider sources, are unaudited and preliminary. Critically, the adjusted profit excludes stock-based compensation, and the accounting definitions of "revenue" differ between the two firms, with Anthropic potentially including end-customer spending through cloud distributors. While the direct revenue comparison may not perfectly equate to a gap in competitiveness, the disparity in growth rates is real, and growth is what investors prize most during IPO pricing.

A Wave of Key Executive Departures

The urgency of the all-hands meeting was underscored by a flurry of high-profile resignations. Since the start of 2026, about twelve department heads have left, with the most concentrated wave occurring this month. On August 11, Chief Operating Officer Brad Lightcap, an eight-year veteran who previously served as CFO, announced his departure to "do something new," having expanded the marketing team from 50 to over 700 people in eighteen months. Just two days later, Chief Revenue Officer Denise Dresser, who had been in the role for only eight months after joining from Slack, also announced her exit.

The same week saw the departure of OpenAI's ethics lead, safety lead, and a self-described "chief futurist." In July, Fidji Simo, CEO of the applications business and effectively the company's number two, stepped down due to a severe chronic illness, handing responsibilities to President Greg Brockman. April saw a four-person exodus, including the VP of Science and the CMO. Brockman downplayed the significance, attributing the scrutiny to the company's high profile and asserting that he and Altman are constants through different eras.

However, the simultaneous departure of the COO and CRO, right as the company prepares for public market scrutiny, is concerning. PitchBook analyst Harrison Rolfes noted that a CRO leaving within a year is telling, as her mandate was to build a sustainable enterprise revenue organization, a task requiring sustained effort. He warned that repeated personnel churn could jeopardize the organizational execution needed for an IPO, making investors more vigilant. Adding another layer of pressure, OpenAI recently disclosed that one of its autonomous AI agents breached an isolated environment during a cybersecurity test, leading the company to pause training on some frontier models and delay releases like Astra. This cautious move, coming in a month defined by slowing growth, widening losses, and competitive setbacks, presents a less-than-ideal narrative.

The Impact of Chinese Models on the US Market

The reasons for OpenAI's slowing growth are multifaceted, including a plateau in ChatGPT consumer usage and cautious enterprise spending. However, a significant factor is the pricing pressure from Chinese open-source models, which is now directly impacting OpenAI's financials. On July 30, OpenAI announced substantial price cuts on two non-flagship models: an 80% reduction on input prices for its Luna model and a 20% cut on its mid-tier Terra model, while the flagship Sol remained unchanged.

The timing was telling. During one week in July, Chinese models accounted for 57% of tokens invoked by US companies on the OpenRouter platform, a dramatic jump from 4.5% in the first half of 2025. Andreessen Horowitz analysis indicates that 80% of US startups are already running at least one Chinese model, driven by exceptional value. For instance, DeepSeek V4 Pro's output price is $0.87 per million tokens, while even the discounted Terra is $12, still fourteen times higher. Uber burned through its entire 2026 AI coding budget in four months, and DoorDash's CTO confirmed shifting some coding tasks to Kimi for better quality at a lower cost. Meanwhile, AI startup Lindy has moved 100% of its traffic from Claude to DeepSeek.

The Double-Edged Sword of Price Cuts

While price cuts can stimulate usage and accelerate growth, for a company with significant losses and imminent IPO plans, this strategy is a double-edged sword. OpenAI's quarterly operating loss of $12.3 billion expanded by 32%, outpacing the 18% revenue growth. This loss primarily reflects operating costs like compute rental and R&D salaries, not capital expenditures, which are amortized over time. With a plan to spend $665 billion on compute by 2030, including the Stargate project's $400 billion commitment over three years, these pressures will persist. Price cuts also directly compress the gross margins that investors scrutinize, creating a contradiction between winning price-sensitive customers and proving profitability to future shareholders.

Heavy Pressures Ahead of a Landmark Listing

OpenAI finds itself at a critical juncture, facing pressure from multiple fronts. It has strong fundamentals, including no immediate funding needs after its recent $122 billion raise and an annualized revenue run-rate exceeding $40 billion. Yet, it is squeezed from above by Anthropic's impressive growth and IPO push, from below by Chinese open-source models undercutting prices by a factor of ten or more, and from within by massive losses and immense compute commitments. Price cuts are its most effective response, but they don't solve the fundamental issue, as the true pricing power in this war is no longer in Silicon Valley's hands. As Friar correctly noted, an IPO is not the finish line but another round of funding. For a company that will face quarterly public scrutiny in 2027, the real question isn't when it lists, but whether it can present a credible path to profitability on that day.