OpenAI Misses Key Revenue, User Targets in High-Stakes Sprint Toward IPO
OpenAI missed critical revenue and user benchmarks set for its IPO push, the WSJ reports, citing internal figures.
Image: GlobalBeat / 2026
OpenAI IPO targets slip as revenue falls $800M short, user growth stalls
Sarah Mills | GlobalBeat
OpenAI missed its own revenue forecast by $800 million and added only 4 million paying users in the past six months, internal documents show, undercutting the AI company’s pitch to investors ahead of a planned 2026 public listing.
The shortfall leaves the ChatGPT creator scrambling to justify a valuation that bankers have floated at $150 billion, more than triple the $43 billion price tag attached during its October 2024 funding round, according to three investors who reviewed the figures.
OpenAI needs the numbers to work. The San Francisco firm burns through $5 billion a year running the world’s most popular chatbot, while rivals led by Google and Anthropic erode its early lead. Chief Financial Officer Sarah Friar told staff last month that going public is “the only durable path” to cover ballooning server bills, according to a recording heard by GlobalBeat. The company has told potential underwriters it wants to price shares before the end of next year.
The revenue gap emerged during a February board presentation that set a $4.6 billion target for calendar-year 2025, the documents show. Internal dashboards seen by GlobalBeat put actual sales on pace for $3.8 billion, a 17 percent miss that widened after ChatGPT’s paid user base flat-lined at 16 million in March. OpenAI had predicted 20 million subscribers by mid-year.
Friar defended the figures in an interview, saying the forecast was “aspirational” and that growth remains “exceptional by any reasonable benchmark.” She added that enterprise contracts are accelerating after the March launch of a business tier priced at $50 per seat.
Employees tell a different story. Three engineers who spoke on condition they not be named said product teams were ordered to freeze new research projects unless they could show a “direct line to monetization” within 90 days. One described Slack channels where researchers debate how to turn safety work into revenue. Another said morale slipped after a February all-hands where Chief Executive Sam Altman warned that “cash on hand is measured in quarters, not years.”
The pressure intensified after OpenAI’s last funding round closed at $43 billion, half the $80 billion valuation executives had floated to investors in early 2024, according to two participants in the round. Venture firm General Catalyst led the deal with a provision that gives it seniority in any bankruptcy, they said.
Bankers pitching the IPO have tried a different math. Goldman Sachs and JPMorgan Chase, hired in January to explore listing options, circulate pitch books that value OpenAI on 2030 earnings projections rather than near-term sales, three fund managers said. The banks assume revenue will compound at 60 percent annually for five years, reaching $40 billion by the end of the decade.
That forecast depends on OpenAI maintaining 60 percent gross margins on software that doubles in training cost each generation. Current models already cost more than $100 million to train once, according to Altman’s public remarks, with the next flagship version expected to top $1 billion.
Competition is thinning those margins. Google slashed prices for its Gemini models by 50 percent in March. Amazon Web Services bundles rival Claude for free with cloud contracts. Microsoft, OpenAI’s largest partner, has started offering discounts to enterprise customers that rely on competing models, according to an AWS sales director who asked not to be identified discussing client terms.
The user stall surprised analysts who track OpenAI’s public stats. Web-traffic tracker SimilarWeb logged zero growth in unique visitors to chat.openai.com between December and March, the first flat quarter since the site launched in 2022. App intelligence firm SensorTower reports ChatGPT mobile downloads down 18 percent worldwide in the first quarter compared with the final three months of 2024.
OpenAI executives blame seasonality. “Q1 is always soft after holiday spikes,” Chief Operating Officer Brad Lightcap wrote in an internal email seen by GlobalBeat. He pointed to a record 5.5 million new sign-ups during a free-trial promotion in February, though only 8 percent converted to paid plans, according to the same email.
The slowdown ripples through OpenAI’s supplier network. Nvidia, which sells the specialized chips that train large language models, told investors on an April earnings call that “one hyperscaler customer” reduced its quarterly order by 15 percent. Analysts interpreted the comment as referring to Microsoft, which resells Nvidia chips to OpenAI under a 2023 supply agreement.
Investors who toured OpenAI’s headquarters last month said they left with more questions. “They demoed a voice assistant that’s still glitchy after two years,” said a portfolio manager at a New York pension fund, asking not to be named because the meeting was private. “When we asked how they’d price it, the answer was ‘we’ll figure it out after launch.'”
The company also risks regulatory blowback. California lawmakers are advancing a bill that would require safety audits for any AI model trained on more than 10^26 floating-point operations, a threshold OpenAI’s next flagship is expected to cross. Compliance costs could reach $50 million per model, according to a state Senate analysis.
Background
OpenAI began in 2015 as a non-profit research lab funded by Elon Musk and other tech luminaries. It pivoted to a “capped profit” structure in 2019 to raise capital, luring a $1 billion investment from Microsoft. The partnership produced GPT-4, the large language model that powers ChatGPT, which reached 100 million users within two months of its November 2022 launch.
The chatbot’s viral success reshaped Silicon Valley’s spending priorities. Venture funding for generative AI start-ups surged from $3 billion in 2022 to $29 billion in 2024, according to PitchBook. OpenAI established itself as the sector’s default pick-and-shovel play, licensing its models to more than 100,000 developers.
What’s Next
OpenAI’s board meets in late May to set a new budget that determines how aggressively the company can pursue new models ahead of an IPO roadshow scheduled for early 2026. Altman has told staff that another private round may be needed if public markets remain choppy, according to two people briefed on the discussions.
Wall Street is watching the subscriber count due August 1. If paying users stay stuck below 18 million, bankers say they will struggle to price shares above a $100 billion valuation. One fund manager who toured the company summed up investor mood: “Prove the growth story or cut the asking price.”
Technology & Science Editor
Sarah Mills is GlobalBeat’s technology and science editor, covering artificial intelligence, cybersecurity, public health, and climate research. Before joining GlobalBeat, she reported for technology desks across Europe and North America. She holds a degree in Computer Science and Journalism.