OpenAI：迫不及待的首次公开募股

OpenAI: The IPO That Institutional Research Group Harrison RolfesSenior Research Analyst,Late-Stage Research The $1.15 trillion question pbinstitutionalresearch@pitchbook.com PitchBook is a Morningstar company providing the most comprehensive, mostaccurate, and hard-to-find data for professionals doing business in the private markets. Published on April 28, 2026 Contents Key takeaways Key takeaways •A realistic IPO window for OpenAI has shifted from Q4 2026 to mid- to late 2027.Public market investors will need to see how $1.15 trillion in infrastructureobligations convert into free cash flow. That large of an obligation requiresseveral more quarters of clean execution before anyone can tell if OpenAI is ready The obligation stack tells us why The governance signal matters more thanthe revenue miss The competitive divergence is compounding AI Business Quality (AIBQ) pressure isbuilding across three dimensions •OpenAI’s infrastructure obligations are fixed, but its revenue is not.Thatmismatch is manageable when growth accelerates and is existential when it does What this means going forward References •CEO Sam Altman and CFO Sarah Friar’s joint denial of internal friction is moreinformative than the revenue miss.Strategic alignment on capital allocation isnot something that healthy pre-IPO companies need to publicly assert. When the •Anthropic generates approximately $6 million in annualized revenue peremployee; OpenAI generates roughly $5.6 million.The gap is narrow today, but thetrajectories diverge. Anthropic is scaling revenue faster than head count. OpenAI •AIBQ composite scoring pressure is building across three dimensionssimultaneously.A downward revision to the high threes at the next scoring window •Whichever company lists first defines the public market multiple for frontier AI.OpenAI is at risk of entering a valuation framework it did not set, on terms it cannotcontrol, behind competitors that got there on a fraction of the capital. That is the The IPO listing timeline is the first casualty OpenAI’s Q4 2026 IPO target was already ambitious given its corporate restructuring.In our view, it is now unrealistic. The company missed multiple monthly revenuetargets earlier this year after ceding share to Anthropic in coding and enterprise, andpublic market investors are going to want several more quarters of clean executionbefore anyone can credibly explain how $1.15 trillion in infrastructure obligationsconvert into positive free cash flow. Mid- to late 2027 is the more realistic window forOpenAI’s listing. This changes the AI IPO landscape. Anthropic and Databricks are The obligation stack tells us why OpenAI generates roughly $2 billion a month in revenue.1By any conventional measure, that is an extraordinary business. But it has committed over $1.15 trillionin infrastructure obligations across Oracle, Microsoft, Amazon Web Services (AWS),NVIDIA, AMD, Broadcom, and CoreWeave.2The Oracle contract alone requires $60 billion a year beginning in 2027,3which exceeds OpenAI’s own projected net revenuefor that year. These are multiyear commitments that do not flex downward. Revenue The governance signal matters more than the On April 27, CEO Sam Altman and CFO Sarah Friar issued a joint statement callingreports of internal friction over compute spending “ridiculous.” The friction reportedlyconcerned whether OpenAI could fund its future computing contracts if its growthdid not reaccelerate. When the two most senior leaders at a pre-IPO company feel the The competitive divergence is compounding Anthropic carries roughly one-twelfth of the infrastructure obligations, operates athigher gross margins, and is gaining share in the exact segments that OpenAI justlost. The number that deserves more attention is revenue efficiency. On an annualizedrun-rate basis, Anthropic generates approximately $6 million per employee on 5,000 AI Business Quality (AIBQ) pressure is building across In March,PitchBook scored OpenAI across five dimensions in its AIBQ framework.Governance optionality, already the weakest dimension for OpenAI at 3 out of10, absorbs the most direct pressure from these recent developments. A publicdisagreement between the CEO and CFO over spending philosophy during the pre-IPOwindow is precisely the type of signal this dimension was designed to capture. The What this means going forward OpenAI’s revenue miss will be forgotten in a quarter. The cost architecture will not. Butthe less obvious implication is what this does to the sequencing of the entire frontierAI IPO class. The longer OpenAI waits to go public, the more likely it becomes thatAnthropic and Databricks will set the valuation framework for the sector. Whichever References 1:“OpenAI Raises $122 Billion to Accelerate the Next Phase of AI,” OpenAI, March 31, 2026.2:“OpenAI’s $1 Trillion Infrastructure Spend,” Tomasz Tunguz, November 3, 2025.3:“Oracle & OpenAI’s $300B Deal: AI Infrastructure Analysis,” Intuitio