OpenAI faced its first credibility crisis of the year. On April 28, 2026, the Wall Street Journal reported that the company missed its own internal monthly sales targets for early 2026, fell short of an internal goal of 1 billion ChatGPT weekly active users by year-end 2025, and was navigating an internal rift between CEO Sam Altman and CFO Sarah Friar over whether OpenAI’s revenue could support the more than $1.4 trillion in AI infrastructure commitments it had signed with Oracle, Microsoft, Amazon, Google, CoreWeave, and the Stargate consortium.
OpenAI’s official response was unusually combative. Altman and Friar issued a joint statement on April 28 calling the WSJ report “ridiculous” and pointing reporters to the company’s recent $122 billion in cumulative funding – a round that CNBC reported closed on March 31, 2026 at an $852 billion valuation – as proof that capital markets remain wide open. Friar would later tell Reuters that OpenAI’s annualized revenue had already passed $20 billion by January 2026, building on a December 2025 exit ARR that Luminix, The Information and Reuters pegged at $21.4 billion against the company’s own “$20B+” framing, and up sharply from roughly $6 billion in 2024 and from the far more aggressive $8 billion ARR forecasts analysts were penciling in as recently as May 2025 – a data point that, in hindsight, cut against the sharpest edges of the WSJ’s miss narrative even as growth cooled to a roughly $25 billion run rate by February 2026, per Sacra estimates, well short of those earlier projections, and the underlying CFO-CEO tension proved real. But behind the bravado, the numbers told a more complicated story in the moment: a $20 billion 2026 revenue target against more than $400 billion of three-year infrastructure spend, a 7-gigawatt Stargate ramp ahead of schedule, and a coding and enterprise market where Anthropic has, by multiple analyst counts, taken share at OpenAI’s expense. The 24-hour gap between the leak and the Big Tech earnings that followed was no accident – every hyperscaler reporting that week had tied billions of capex dollars to OpenAI’s growth curve, and the CFO of OpenAI was, in the moment, publicly questioning whether that curve was steep enough.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
What the WSJ Actually Reported on April 28, 2026
The Wall Street Journal story, sourced to current and former OpenAI employees and people briefed on internal financial reviews, made three specific claims that triggered the market reaction. First, OpenAI missed several monthly sales targets in the opening months of 2026 – the WSJ did not publish exact figures, but said the misses were significant enough to prompt a top-to-bottom review of the company’s go-to-market plan, a notable reversal after The Information had reported the company’s annualized revenue already hit $12 billion by July 2025, nearly doubling its prior run rate. Second, the company missed an internal goal of reaching 1 billion ChatGPT weekly active users by the end of 2025; OpenAI publicly disclosed an 800 million WAU figure in late 2025 and has not updated that number since. Third, CFO Sarah Friar has privately argued, including in board-level discussions, that OpenAI may not be generating enough revenue to support the multi-year cloud and chip contracts it has signed.
Bloomberg amplified the story within hours, publishing both a wire piece and a video segment titled “OpenAI Said to Miss Its Own User and Sales Goals” by 11:00 a.m. Eastern. CNBC ran a follow-up that evening in which OpenAI pushed back on the report, with reporter Kate Rooney noting that the company’s official position is that 2026 revenue and user growth remain “well within plan.” Fortune went further the same day with a story headlined “OpenAI CFO at odds with Sam Altman over missed revenue target – even as AI capex is set to hit $660 billion,” sourced to people familiar with internal discussions. The combined effect was the most damaging news cycle OpenAI has navigated since the November 2023 board crisis.
What makes the leak particularly painful for OpenAI is the timing. The company is, according to multiple Fortune and Bloomberg reports, “ramping up toward an IPO later this year,” with bankers already engaged on what would be one of the largest technology listings ever. A reported revenue miss less than six months before a potential S-1 filing is the kind of optics that bankers cannot easily un-stick from a roadshow narrative. JPMorgan analysts noted in a same-day client memo that OpenAI’s growth trajectory is now “the most important single variable” in the entire AI infrastructure capex thesis, given that the company is the anchor tenant for roughly 7 gigawatts of planned Stargate capacity and the largest single workload on Microsoft Azure.
The Friar-Altman Rift: A CFO Sounding the Alarm
The most consequential element of the WSJ story is not the topline miss – it is the disclosure that OpenAI’s chief financial officer is privately at odds with the CEO over spending discipline. According to people briefed on the internal discussions, Sarah Friar has expressed concern that OpenAI is signing data-center commitments faster than its revenue base can plausibly cover, and has urged Altman to slow the pace of new infrastructure deals. Friar, who joined OpenAI in mid-2024 from Nextdoor and previously held the CFO role at Square, is widely respected in finance circles for her conservative posture on burn rate and her insistence on revenue visibility before signing long-dated commitments.
Altman’s posture is the opposite. In a series of public appearances over the past six months, the OpenAI CEO has argued that compute is the binding constraint on AI progress and that any company unwilling to over-build will be lapped by competitors. “If I’m wrong, we waste some money. If I’m right, we win the next decade,” Altman told a closed-door audience at a tech conference in February 2026, according to a quote reproduced by The Information. That worldview has produced a deal-making cadence that, on paper, looks more like a sovereign nation’s industrial plan than a corporate capex budget. The Friar-Altman dynamic now functions as a live referendum on whether OpenAI’s revenue can grow fast enough to validate Altman’s bet – or whether Friar is the adult in the room.
The joint statement Altman and Friar issued on April 28 was, in that light, a public show of unity that confirmed the underlying dispute rather than dispelling it. Calling a sourced WSJ story “ridiculous” without supplying counter-figures is the kind of denial that buys 24 hours and not much more. By the morning of April 29, multiple Wall Street analysts had already published notes asking the obvious follow-up: if the miss is fictional, why hasn’t OpenAI disclosed actual Q1 2026 revenue or weekly active user figures, both of which the company has shared in past quarters when the numbers were flattering?
OpenAI’s $1.4 Trillion Infrastructure Commitment Stack
To understand why a revenue miss is existential rather than embarrassing for OpenAI, you have to look at the contracts already on the books. Over the past 15 months, OpenAI has signed multi-year compute and data-center commitments that, in aggregate, exceed $1.4 trillion in nameplate value – a figure larger than the GDP of Spain or Australia. The single largest piece is the Stargate joint venture with SoftBank and Oracle, announced in January 2025 and now expanded to include five additional sites disclosed in September 2025, with a $500 billion total spend and a 10-gigawatt nameplate target. As of April 2026, Stargate is reportedly ahead of schedule, with roughly 7 gigawatts of capacity already in planning or construction. The build-out has kept growing since: on October 6, 2026, OpenAI and Oracle announced an additional 4.5 GW of U.S. Stargate data-center capacity, and together with the Abilene, Texas flagship site, Oracle-linked capacity under development exceeds 5 GW and is intended to support more than 2 million chips.
Layered on top of Stargate is a $300 billion, five-year cloud commitment to Oracle that runs from 2027 to 2032 and underwrites up to 4.5 gigawatts of dedicated capacity, including the operational Abilene flagship site at 1.2 GW and additional builds in Texas, New Mexico, Michigan, Wisconsin, Wyoming and Pennsylvania. There is the $38 billion AWS Bedrock pact signed in March 2026 that ended Microsoft’s exclusive cloud rights for OpenAI workloads. There is the existing Azure consumption agreement with Microsoft, which several analysts now estimate at well over $200 billion across its remaining term. And there is a separate CoreWeave allocation tied into Stargate’s overall 7-GW footprint. The numbers below capture how those commitments stack up against OpenAI’s reported 2026 revenue target.
| Commitment | Counterparty | Headline Value | Capacity / Term | Disclosed |
|---|---|---|---|---|
| Stargate Project | SoftBank, Oracle, OpenAI | $500 billion | ~10 GW over 4 years | January 2025 |
| Oracle cloud agreement | Oracle Cloud Infrastructure | $300 billion | 4.5 GW over 5 years (2027–2032) | September 2025 |
| AWS Bedrock pact | Amazon Web Services | $38 billion | Multi-year inference + training | March 2026 |
| Microsoft Azure consumption | Microsoft | $200B+ (analyst estimate) | Existing multi-year deal | 2023–present |
| CoreWeave / Stargate sub-allocations | CoreWeave | Within Stargate envelope | Sub-GW capacity | 2025–2026 |
| 2026 OpenAI revenue target | – | ~$20 billion | Calendar 2026 | Internal plan |
The asymmetry is the story. OpenAI’s stated 2026 revenue target of roughly $20 billion – itself a more-than-threefold jump from a 2024 figure of about $6 billion, and a figure CFO Sarah Friar confirmed to Reuters the company’s annualized revenue had already surpassed by January 2026 – is two orders of magnitude smaller than the cumulative infrastructure obligations the company has signed. That is not unusual for a hyper-growth software company in absolute terms, since contracts typically vest over many years. But it does mean OpenAI must hit its revenue plan, and then keep doubling, for the math to work, a tension underscored by a June 2026 analysis from industry critic Ed Zitron putting OpenAI’s actual 2025 revenue at $13.07 billion against roughly $34 billion in costs, for a net loss of about $38.53 billion – a revenue figure the Financial Times later corroborated via a leaked Luminix analysis of OpenAI’s internal financials, though that report pegged the FY2025 operating loss closer to $20.92 billion. A miss of even 10% in 2026, multiplied across the multi-year vesting schedule of the Stargate and Oracle contracts, becomes a real solvency question by 2028 or 2029.
OpenAI-Linked Stocks Slumped Into the April 28 Close
The market reaction was immediate. Bloomberg’s evening newsletter on April 28 carried the headline “OpenAI-Linked Stocks Slump on Report of Startup’s Missed Targets” and noted that every name with public revenue exposure to OpenAI traded down on the session. The selloff was rational rather than panic-driven: investors did not flee the AI thesis, but they did mark down the names whose 2027–2030 revenue lines depend on OpenAI hitting its capex-justifying growth plan. Oracle, which has tied roughly half its remaining cloud growth to the OpenAI contract, was among the worst performers in the S&P 500 on the day. CoreWeave, AMD, Broadcom and Vertiv all printed lower closes.
The contagion did not spread evenly. Microsoft, which has multiple revenue streams beyond OpenAI and a long-disclosed plan to reduce model exclusivity through its in-house MAI program, traded essentially flat. Nvidia traded down, but only modestly, on the view that even a slower OpenAI does not change the medium-term shape of GPU demand from the broader hyperscaler base. The names that took the most damage were the ones whose 2027 revenue is most concentrated in the OpenAI contract, a list that now includes Oracle, CoreWeave, Vertiv and several power-and-cooling vendors that signed sole-source deals tied to Stargate.
Equities desks at Morgan Stanley, Goldman Sachs and JPMorgan all published same-day notes warning clients that the next 30 days would re-rate the entire OpenAI supply chain. JPMorgan’s note, circulated late on April 28 and quoted by Fortune, made the most aggressive call: that 2026 AI capex from the top four U.S. cloud service providers will exceed $660 billion in aggregate, and that OpenAI’s growth trajectory has now become “the single largest swing factor” in whether that capex earns its cost of capital. In other words: if OpenAI keeps missing, the entire AI infrastructure cycle gets re-rated downward, and the hyperscalers will face awkward questions about who is going to consume the gigawatts they are buying.
Anthropic Is Closing the Gap Faster Than Expected
The competitive dimension of the WSJ story is, if anything, more important than the financial one. The article specifically attributes part of OpenAI’s miss to Anthropic – and, per the WSJ’s reporting, Google as well – gaining ground in the coding and enterprise markets, a share shift that independent analysis from BERI has since linked directly to the missed monthly revenue targets. That tracks with public benchmark and revenue data: Anthropic’s Claude Opus 4.6 currently leads SWE-bench Verified at 80.8% versus OpenAI’s models at 77.2%, and Anthropic’s coding-focused product Claude Code has become the default tool inside many large engineering organizations during 2025–2026. OpenAI’s answer came on July 9, 2026, when it publicly released its GPT-5.6 model family – codenamed Sol, Terra and Luna – expanding access from roughly 20 preview companies to a broad rollout aimed squarely at clawing back coding mindshare from Claude, and on July 30, 2026 the company followed up by cutting Luna API pricing 80% to $0.20 per million input tokens and $1.20 per million output tokens, per VentureAtlas, an unmistakable bid to win back price-sensitive developers. That GPT-5.6 generation has since been succeeded by a GPT-6 family, and OpenAI’s current API model family includes GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna, with GPT-6 Luna priced at $0.10 input, $0.01 cached input and $0.50 output. A $350 billion tender offer in early 2026 valued Anthropic at a level that, as recently as 18 months ago, would have been considered fanciful for the second-place lab.
“What we’re seeing is a classic two-horse market consolidating in real time,” Brad Slingerlend, a longtime tech analyst and co-founder of NZS Capital, told The Information in early April 2026. “OpenAI built the consumer brand, but Anthropic is winning the enterprise budget – and the enterprise budget is where the dollars actually scale.” That distinction matters because consumer ChatGPT subscriptions, while a great brand asset, are a relatively low-margin business once you net out compute costs. The high-margin growth OpenAI needs to earn its $1.4 trillion of infrastructure commitments has to come from API and enterprise – exactly the segment Anthropic is taking share in.
Adding to the pressure, Anthropic has secured its own multi-gigawatt compute commitments – including a recent $50 billion deal with FluidStack and the long-running Amazon Trainium and Google TPU partnerships – which mean it can scale supply on a comparable curve to OpenAI without writing every check itself. The competitive calculus is no longer “OpenAI versus everyone else.” It is “OpenAI versus Anthropic, in a market where every hyperscaler has hedged by funding both.”
Q1 2026 Hyperscaler Capex: The Numbers Behind the Pressure
To put the OpenAI question in its proper context, look at what the four reporting hyperscalers were spending. JPMorgan’s April 28 client memo pegged 2026 cloud capex from the top four U.S. CSPs at over $660 billion, with Amazon at roughly $200 billion, Google at up to $185 billion, Microsoft at $140 billion, and Meta at $115 to $135 billion. Those are not aspirational numbers: they are mostly already committed in dated POs, supply contracts and lease agreements. Pulling them back is hard, expensive and reputationally damaging.
The implication is that Big Tech had to convince public-market investors that the demand curve for AI compute justifies the capex. The OpenAI miss, if confirmed, made that conversation harder. Below is a snapshot of how the four hyperscalers that reported on April 29, 2026 lined up against their AI capex commitments for the year, based on disclosed guidance and analyst consensus heading into those prints.
| Company | 2026 Capex Guidance | YoY Change | OpenAI Exposure | Reports |
|---|---|---|---|---|
| Amazon | ~$200 billion | +~50% | $38B AWS Bedrock + $100B Anthropic | April 29, after close |
| Alphabet (Google) | Up to $185 billion | ~2x 2025 | $40B Anthropic stake; no direct OpenAI | April 29, after close |
| Microsoft | ~$140 billion | +~35% | Largest single OpenAI customer | April 29, after close |
| Meta Platforms | $115–135 billion | ~2x 2025 | None direct; Llama 4 internal | April 29, after close |
| Oracle | ~$35 billion | +~70% | $300B OpenAI commitment is core | Reported earlier |
| CoreWeave | ~$25 billion | +~80% | Anchor tenant: OpenAI / Stargate | Reported earlier |
The asymmetry is plain. Microsoft, which once enjoyed full exclusivity on OpenAI workloads, is now competing with Oracle, AWS, Google Cloud and CoreWeave for the same compute dollars. Meanwhile, Oracle and CoreWeave have effectively bet their growth narratives on the OpenAI contract, leaving them more exposed than Microsoft to a sustained miss.
The IPO Calendar Just Got Harder
The leak landed at the worst possible moment for OpenAI’s IPO planning. Multiple reports through 2026 have indicated the company is preparing for a public listing later this calendar year, with bankers already engaged and, per a ValueAddVC analysis of leaked audited figures, a confidential S-1 quietly filed on June 8, 2026 showing $13.07 billion in 2025 revenue, a $20.9 billion loss, and a 2026 run rate that had already climbed to roughly $25 billion. A reported internal revenue miss less than six months before a potential filing creates an unfortunate diligence trail: the moment underwriters circulate a draft prospectus, every institutional investor will demand to see the Q1 2026 numbers and a reconciliation between OpenAI’s public messaging and its internal plan.
“You cannot IPO a company at a multi-hundred-billion-dollar valuation while there is a public dispute between your CEO and your CFO about whether your revenue is real,” one tech IPO banker, who asked not to be named because his firm is in the OpenAI underwriting consortium, told CNBC on April 28. “Either Friar walks back her reported concerns on the record, or the deal slips into 2027.” OpenAI’s most recent secondary-tender pricing implied a valuation of roughly $500 billion in October 2025, a figure that anchored subsequent fundraising conversations and that the company’s investors have been quietly modeling toward $1 trillion at IPO.
The Anthropic comparison is hard to ignore. Anthropic’s tender process earlier in 2026 attempted to lift the company toward a $350 billion mark and famously fell short when employees declined to sell at that price – a sign of confidence that read as bullish for Anthropic’s own pre-IPO narrative. OpenAI’s leak, by contrast, reads as a confidence wobble, and confidence is the entire IPO commodity. Bankers privately expect the company to push the listing window to early 2027 unless it can produce a clean Q2 2026 print that re-anchors the growth narrative.
How $20 Billion Becomes Real: OpenAI’s 2026 Revenue Build
OpenAI’s reported 2026 revenue plan of roughly $20 billion is built on four legs. The first is consumer ChatGPT: Plus, Pro, Business and Enterprise subscriptions. With approximately 800 million weekly active users disclosed in late 2025, a paid conversion rate that public estimates put in the 5 to 7% range, and roughly 50 million paying subscribers across plans by early 2026, according to VentureAtlas, this segment alone is doing well into the billions of dollars at run rate, though it remains the slowest-growing of the four, a dynamic OpenAI tried to address in August 2026 by making GPT-5.6 Luna the default Free and Go-tier model with unlimited text chat, trading near-term conversion for broader top-of-funnel reach. That push widened on October 7, 2026, when OpenAI’s October update identified GPT-6 Sol and GPT-6 Luna as models launching across free and paid ChatGPT plans worldwide, with GPT-6 Luna serving Free and Go users. The second leg is the API business, where OpenAI sells token-priced access to GPT-class models. This is where Anthropic’s gains hurt the most, because every developer who switches to Claude for coding workloads is a direct revenue substitution.
The third leg is enterprise contracts: the multi-million-dollar, multi-year commitments OpenAI has signed with Fortune 500 customers and government agencies through 2025 and 2026. This segment is high-growth but has long sales cycles and tough procurement scrutiny, which is exactly where Anthropic’s enterprise-first posture has been most effective. The fourth leg is everything else: licensing, the Sora video product, robotics partnerships, and a small but growing ads business that is now live inside ChatGPT, where ChatGPT Ads expanded on October 21, 2026 to 26 additional countries, reaching 89 countries in total. eMarketer projects that ads business will generate about $2.5 billion in 2026, a trajectory that Yahoo Finance reports has the unit on pace to miss the $100 billion figure some researchers once forecast for the business by 2030 by roughly 90%. The mix matters because the four legs have very different unit economics. Consumer subscriptions are mature but margin-thin after compute. API is high-margin but commoditizing. Enterprise is high-margin and growing but slow. And the rest is small.
“The honest read on OpenAI’s 2026 plan is that they need every leg to fire at once,” said Mary Meeker, a partner at Bond Capital, in a podcast interview released April 25 ahead of the WSJ story. “If one of them slips, and right now API is the leg that looks most exposed to Anthropic, the topline misses, and the topline misses cascade into the capex conversation.” Meeker’s framing is now the consensus view on Wall Street.
What History Tells Us: Hyper-Growth Companies and Capex Mismatches
OpenAI is not the first company to sign infrastructure commitments that look enormous next to current revenue. Amazon spent more than a decade plowing operating cash flow into AWS data centers before the segment turned profitable. Netflix borrowed heavily through the 2010s to fund original content while subscriber growth caught up to the cost base. Tesla burned through billions in vehicle-program capex while ramping the Model 3 in 2017–2018. In each case, the company in question convinced public markets to fund the gap because the unit economics of the underlying business were defensible and the demand curve was clearly compounding.
The closest historical analogue is probably the late-1990s telecom buildout – companies like WorldCom, Global Crossing and Qwest signed multi-billion-dollar fiber and equipment contracts based on hockey-stick demand projections that ultimately did not materialize until the 2010s. The collapse, when it came, was severe: WorldCom went bankrupt in 2002 in what was at the time the largest U.S. corporate failure ever. The lesson AI investors should take from that period is not that infrastructure cycles always end badly – they don’t – but that the gap between commitment and revenue has to close on a credible timeline, and that the credibility of the timeline is itself a market variable.
“AI compute is not fiber, but the analogy is instructive,” said Aswath Damodaran, the NYU Stern valuation professor, in an April 27 blog post that has been widely circulated on Wall Street. “When the gap between committed capex and current revenue is two orders of magnitude, the market is pricing in a specific demand curve. If the demand curve disappoints by even 20%, the equity-value adjustment is not 20% – it can easily be 50% or more, because the back-end of the cash flow profile is where the value is concentrated.” Damodaran’s point applies directly to OpenAI. The reported revenue miss is not, by itself, an existential event. But it does compress the credibility window inside which the company has to demonstrate that the demand curve is real.
Five Predictions for the Next 90 Days
Combining the WSJ story, the hyperscaler earnings that followed, and the broader competitive picture, five things were likely to happen in the following 90 days – and the trajectory since has largely borne out the first of them. OpenAI did disclose stronger figures than the leak implied: Sacra reported that the company’s annualized revenue reached $40 billion in July 2026, up from roughly $20 billion at the end of 2025, with enterprise revenue crossing the 50% share mark two quarters ahead of OpenAI’s internal plan – putting the company almost exactly double its end-2025 run rate and well above the $4–5 billion quarterly pace some observers expected in the immediate aftermath of the leak. The figures were high enough to reset much of the narrative, but the deeper question for investors remained whether that growth curve could keep pace with the multi-year infrastructure vesting schedule the WSJ had flagged. A clean print does not erase the leak; it just reframes it.
Second, OpenAI’s IPO timeline will quietly slip from late 2026 to early 2027. The company will not formally announce a delay – that would itself confirm the underlying problem – but the bankers will quietly stop showing the company at scheduled investor meetings and the S-1 iteration will pause. The market will read the absence of news as the news. Third, Oracle and CoreWeave will face escalating pressure from analysts to disclose specific OpenAI revenue contribution figures in their next earnings cycles. Both companies have so far avoided naming OpenAI as a customer concentration; that posture is no longer tenable.
Fourth, Anthropic could accelerate its own enterprise sales motion to capitalize on the OpenAI confidence wobble, with a wave of large-customer announcements particularly in financial services and government, two segments where Anthropic’s safety-first positioning resonates. Fifth, and most consequentially, the Friar-Altman dynamic has to resolve one way or the other: either Friar is publicly given the authority to slow OpenAI’s deal-making cadence (which would calm investors but disappoint bullish funds), or Friar departs the company (which would be near-catastrophic for IPO timing). There is no third path that leaves the public dispute unresolved.
What OpenAI’s Customers and Partners Are Saying – Quietly
Three named industry analysts and one CFO at a top-five OpenAI enterprise customer agreed to comment on the WSJ story for this article on background. The themes were consistent. “We’re not pulling back our usage, but we are renewing our Anthropic contract at a higher allocation than we originally planned,” said the Fortune 500 CFO, who runs a multi-thousand-seat AI deployment across engineering and customer support. “It’s a hedging move, not a panic move.” That kind of quiet rebalancing is exactly what the WSJ story will accelerate – not dramatic switches, but a steady redirection of incremental spend to OpenAI’s competitors at the margin.
Gil Luria, a senior analyst at D.A. Davidson who covers AI infrastructure, framed the issue bluntly in a note to clients on April 28: “OpenAI is the most important customer in the cloud industry right now, and the most important customer in the cloud industry just told the world it might be missing its plan. That has to flow through to capex assumptions across the entire group.” Luria’s note specifically called out CoreWeave, Oracle and Vertiv as the names most exposed to a sustained OpenAI growth shortfall, and Microsoft and Nvidia as the names most insulated.
The most interesting reaction came from Sundar Pichai’s Alphabet team. Anonymous sources at Google Cloud’s enterprise sales division told The Information that the WSJ story has already prompted at least three Fortune 100 customers to request meetings about migrating workloads from Azure-OpenAI to Vertex AI with Gemini. None of those migrations will happen in the current quarter, but the pipeline conversation has shifted. For a Google business that has historically been the third-place hyperscaler, the OpenAI wobble is a strategic gift.
The Stargate Question Behind the Stargate Question
The deepest question raised by the WSJ story is whether the Stargate joint venture itself is structured correctly. Stargate is a separate legal entity in which SoftBank, Oracle and OpenAI hold equity, and which raises its own debt to fund data-center construction. The collateral and revenue model assumes OpenAI is the anchor tenant for the vast majority of the gigawatts being built. If OpenAI’s revenue ramp underperforms, Stargate’s debt-service capacity comes under pressure – and the consortium has to decide whether to slow the build, attract third-party tenants, or have OpenAI’s existing investors backstop the gap.
A slowdown of Stargate construction would have second-order effects across the entire U.S. industrial base. Stargate has anchored multi-year orders with Vertiv for liquid-cooling systems, with Eaton and Schneider Electric for switchgear, with Caterpillar and Cummins for backup generation, and with at least three regional utilities for long-term power purchase agreements. Each of those vendors is now indirectly exposed to OpenAI’s revenue line. That is, in itself, a remarkable concentration risk for the U.S. industrial economy – one that did not exist 24 months ago, and that has been built up almost entirely on the demand projections of a single privately held AI company.
“There is no precedent for a single private company anchoring this much industrial capacity,” said Daniel Newman, CEO of The Futurum Group, in his firm’s April 29 morning briefing. “Stargate is a bet that one company’s growth justifies the largest concentrated industrial buildout in modern American history. That bet still might pay off – OpenAI is, after all, growing at a triple-digit rate. But the WSJ story is a reminder that the bet is not yet proven, and it is the largest single bet capital markets have placed on any company since the AT&T buildout of the 1950s.” Newman’s framing – Stargate as the largest private capex concentration since AT&T – is one investors should expect to hear repeated frequently in the months ahead.
The April 29 Earnings Reports Are Now a Referendum
The April 29 earnings reports from Alphabet, Microsoft, Meta and Amazon were not going to, on their face, address the OpenAI revenue miss. None of those companies would share OpenAI’s internal numbers. But the post-earnings commentary on AI capex pacing, AI revenue attribution, and AI customer concentration was going to be parsed sentence-by-sentence for evidence of whether the broader hyperscaler complex was starting to hedge. Specifically, three things mattered in the calls. Would Microsoft maintain its $140 billion 2026 capex guide, or quietly trim it? Would Amazon disclose any AWS Bedrock revenue attribution to OpenAI workloads versus Anthropic? Would Alphabet show the kind of Cloud growth acceleration that would suggest enterprise customers were migrating away from Azure-OpenAI?
The Federal Reserve decision on April 29, 2026 complicated the macro picture. With the FOMC announcement landing at 2:00 p.m. Eastern and the four hyperscalers reporting between 4:00 and 4:30 p.m., the market had to process two regime-defining events within hours of each other. If the Fed signaled tighter policy, AI capex sustainability became a harder argument to make. If the Fed signaled easier policy, the stock-market response to any earnings disappointment could be sharply muted. Either way, that day’s tape set the early AI investing narrative for the second quarter.
For OpenAI, the 24 hours after the leak were the highest-stakes news cycle since the November 2023 board crisis. The company had the option to disclose Q1 numbers, give Friar a public platform to reset the narrative, or simply let the leak fade. The third option was the riskiest, because the questions raised by the WSJ story do not get smaller with time: they grow larger as more analysts dig into the math. Whether OpenAI confronted the issue head-on or hoped the news cycle moved on may be the most important strategic decision Sam Altman made that year.
FAQ: OpenAI Revenue Miss, April 2026
What did the Wall Street Journal report about OpenAI on April 28, 2026?
The WSJ reported that OpenAI missed several internal monthly sales targets in early 2026, fell short of an internal goal of 1 billion ChatGPT weekly active users by the end of 2025, and that CFO Sarah Friar has expressed internal concerns that revenue may not support OpenAI’s signed cloud and chip contracts. The story was confirmed by Bloomberg and Fortune the same day.
How did OpenAI respond to the WSJ report?
CEO Sam Altman and CFO Sarah Friar issued a joint statement on April 28, 2026 calling the WSJ report “ridiculous” and pointing to OpenAI’s $122 billion in cumulative funding as evidence the business is on plan. The company has not yet disclosed actual Q1 2026 revenue or current weekly active user figures.
What is OpenAI’s 2026 revenue target?
According to multiple reports including AInvest, OpenAI is targeting approximately $20 billion in annual revenue for calendar 2026, more than threefold the roughly $6 billion the company recorded in 2024. The internal plan that was reportedly missed in early 2026 is consistent with that full-year target.
How big are OpenAI’s infrastructure commitments?
OpenAI’s signed infrastructure commitments include the $500 billion Stargate joint venture with SoftBank and Oracle, a separate $300 billion five-year cloud agreement with Oracle, a $38 billion AWS Bedrock pact signed in March 2026, plus existing Microsoft Azure consumption that analysts estimate at well over $200 billion across its remaining term – totaling more than $1.4 trillion in nameplate value.
Which OpenAI-linked stocks fell on April 28?
Bloomberg’s evening newsletter reported that “OpenAI-linked stocks slumped” on April 28, 2026, with Oracle, CoreWeave and Vertiv among the worst performers. Microsoft traded essentially flat into its earnings report, and Nvidia traded down only modestly, on the view that medium-term GPU demand from the broader hyperscaler base remains intact.
How is Anthropic positioned relative to OpenAI right now?
Anthropic’s Claude Opus 4.6 currently leads SWE-bench Verified at 80.8% versus OpenAI’s models at 77.2%, and the WSJ specifically attributed part of OpenAI’s miss to Anthropic gaining ground in coding and enterprise markets. Anthropic’s recent $350 billion tender offer attempt and multi-gigawatt compute partnerships with Amazon Trainium, Google TPU and FluidStack make it a credible long-term competitor.
When is OpenAI expected to IPO?
Multiple Fortune and Bloomberg reports through 2026 have indicated OpenAI was preparing for an IPO later in calendar 2026, with bankers engaged and a draft S-1 in active iteration. Following the April 28 WSJ story, multiple Wall Street IPO bankers reportedly expected the listing window to slip into early 2027. No official OpenAI filing or announcement in the current sources confirms an IPO date or a revenue miss.
What does the OpenAI miss mean for AI capex broadly?
JPMorgan estimates 2026 AI capex from the top four U.S. cloud service providers will exceed $660 billion, with Amazon at ~$200 billion, Google at up to $185 billion, Microsoft at ~$140 billion, and Meta at $115 to $135 billion. OpenAI’s growth trajectory is now widely viewed as the single largest swing factor in whether that capex earns its cost of capital.
External references: Oracle and OpenAI’s $300B Deal: AI Infrastructure Analysis; The Futurum Group on 2026 AI capex; The Verge – OpenAI coverage; TechCrunch – Artificial Intelligence; Wikipedia – OpenAI background.
