OpenAI’s $600 Billion Bet Meets a Revenue Wall
April 28, 2026 — OpenAI missed internal targets for new users and revenue early this year, triggering a swift market selloff across AI infrastructure stocks and reigniting a fundamental debate: Have the economics underpinning the AI arms race finally hit a ceiling?
The Immediate Fallout: Markets Reprice the Ecosystem
The Wall Street Journal’s April 28 report landed like a depth charge across the semiconductor and cloud sectors. Oracle—carrying a $300 billion, five-year cloud commitment with OpenAI—slid more than 3% during the session, following a premarket plunge of as much as 7.7%. CoreWeave, which anchored its growth to an expanded $22.4 billion infrastructure pact with the AI giant last year, dropped 7.4% to roughly $104. SoftBank, a major equity backer, tumbled as much as 11% in Tokyo, while AMD shed approximately 3%.
The selloff was sharp, yet highly targeted. Analysts at Jefferies and D.A. Davidson specifically flagged Oracle as the most exposed counterparty. “OpenAI’s most significant counterparty is Oracle,” noted D.A. Davidson’s head of research. “If OpenAI is unable to fulfill that commitment, Oracle would feel the brunt of it.” That vulnerability is exacerbated by the heavy debt Oracle has already assumed to build out data centers ahead of the deal’s revenue start date next year.
What the Numbers Actually Say
Beneath the market reaction lies a stark operational reality. OpenAI failed to reach its internal target of one billion weekly active ChatGPT users by the end of 2025, reaching approximately 800 million late last year and inching to roughly 900 million by February 2026. While undeniably massive in absolute terms, this marks a visible deceleration from the explosive pace that saw users double from 400 million in early 2025.
Compounding the user slowdown, the company missed multiple monthly revenue targets in early 2026. OpenAI is actively bleeding enterprise and coding market share to Anthropic, while ceding consumer ground to Google's Gemini amid elevated subscriber churn. Internally, the math is triggering alarms. CFO Sarah Friar has privately warned leadership that decelerating revenue growth threatens the firm's ability to service its colossal future compute contracts. Simultaneously, the board has scrutinized CEO Sam Altman’s aggressive push for additional data-center capacity in the face of slowing momentum.
While Altman and Friar issued a joint statement denying any rift—dismissing rumors of division as “ridiculous”—the mere necessity of a coordinated public denial confirms that internal pressure has reached a boiling point.
The Structural Tension: Capex Commitment vs. Revenue Reality
OpenAI’s dealmaking spree has locked the company into approximately $600 billion in cumulative compute spending through roughly 2030—a staggering ledger that includes Oracle ($300 billion), CoreWeave ($22.4 billion cumulative), Microsoft, AWS, and others.
To fund this, the company closed a record $122 billion funding round in March 2026 at an $852 billion post-money valuation. Yet, internal projections suggest OpenAI could lose approximately $14 billion in 2026 alone, putting it on pace to exhaust a vast majority of that $122 billion war chest within three years—even if it hits its ambitious, revised revenue targets. Complicating matters, some of this funding is strictly conditional upon specific partner agreements.
The arithmetic is unforgiving: astronomical fixed compute costs are colliding with softening near-term revenue. With a looming IPO demanding the kind of profitability visibility that private-market hype easily obscures, Friar’s reported hesitation regarding a 2026 public offering—specifically flagging gaps in internal controls and reporting rigor—stands out as the most prudent signal in the entire saga.
A Stress Test, Not a Collapse
This is a severe but navigable recalibration of the AI capex thesis, not its structural refutation.
OpenAI’s "buy everything" compute strategy was entirely rational when ChatGPT’s growth trajectory looked invincible. Slowing user and revenue momentum inherently amplifies the risk, but the company’s moat remains formidable. It commands massive consumer distribution with ~900 million weekly users, sits on a fresh $122 billion capital buffer, boasts leading frontier models (including the newly released GPT-5.5), and holds a structural advantage by having locked in massive compute capacity ahead of an industry-wide crunch that is now squeezing rivals like Anthropic.
The genuine, long-term risk is far more subtle than a quarterly revenue miss. Internal evaluations of project "Orion" reportedly revealed that vastly larger models, trained on exponentially more compute, are no longer yielding proportionally smarter results. If scaling laws are truly stalling, the foundational rationale for $600 billion in infrastructure commitments weakens—not because demand is evaporating, but because the returns per dollar of compute are diminishing.
That is the pivotal question investors must ask, rather than fixating on whether this week’s selloff marks a definitive top. Anthropic’s rapid enterprise gains confirm that real, durable demand exists. OpenAI’s stumble simply confirms that execution and unit economics finally matter.
For institutional investors, this is a healthy valuation and sentiment reset. The playbook is clear: Buy quality, diversified AI infrastructure names selectively, and avoid pure-play OpenAI proxies until revenue re-acceleration is demonstrably proven. The AI revolution is fully intact—it has simply arrived with a P&L.
not investment advice