OpenAI Hits $25B Revenue—and the IPO That Will Expose Every AI Valuation Fantasy

OpenAI Hits $25B Revenue—and the IPO That Will Expose Every AI Valuation Fantasy

The $25B milestone is real. The $1T asking price is a bet. The internal consensus is not.

The Deal

OpenAI crossed $25 billion in annualized revenue in February 2026, up from $6 billion in 2024 and $20 billion at year-end 2025 — a growth arc that has no precedent in enterprise software history. Salesforce took 18 years to reach this number. Google took 17. OpenAI did it in 39 months from zero.

The company has 900 million weekly ChatGPT users. It closed a $110 billion funding round in March 2026 at an $852 billion post-money valuation, with commitments from Amazon ($50 billion), SoftBank ($30 billion), and Nvidia ($30 billion in GPU compute credits). The implied asking price for a public listing: up to $1 trillion.

And yet.

OpenAI is projected to lose $14 billion in 2026, with annual cash burn reaching $57 billion by 2027. Breakeven is not expected until 2030. The company has committed to over $500 billion in cloud compute capacity through the end of the decade. It is not remotely close to self-sustaining.

Two people inside the company are telling different stories. Sam Altman wants to list in Q4 2026. CFO Sarah Friar is urging caution. Amazon's $35 billion tranche — part of its $50 billion commitment — is reportedly conditional on an IPO milestone. The timeline is being set by investors with capital at risk, not by financial readiness.

This is not a typical pre-IPO tension. This is a structural conflict between ambitions that require public market validation and a financial reality that may not support the valuation.

What the Numbers Actually Mean for Boards

1. The Revenue Story Is Strong. The Margin Story Is Not.

$25 billion in annualized revenue — growing roughly 2x year-over-year for three consecutive years — is extraordinary. But the growth rate in absolute dollars is decelerating: from 3.4x in 2024-2025 to an expected 2.2x in 2026. The company is getting larger, not faster.

Meanwhile, compute costs scale with model capability. OpenAI has committed to approximately $600 billion in total compute spending through 2030. Every new model generation requires infrastructure investment that must be made before revenue can follow. The $14B loss in 2026 is not a sign of mismanagement. It is a structural feature of the business.

For boards evaluating AI investments: understand that leading AI companies are not just selling software. They are building and operating critical national infrastructure. The economics only work at extraordinary scale.

2. Anthropic Is the Story Nobody in the IPO Room Wants to Acknowledge

OpenAI's $25 billion revenue lead would be more comfortable if the distance to its nearest competitor were widening. Based on current trajectories, it is not.

Anthropic reached approximately $19 billion in annualized revenue as of early 2026 — nearly 10x its prior year level. Epoch AI has projected that at current growth rates (Anthropic growing at roughly 10x per year, OpenAI at 3.4x), Anthropic could overtake OpenAI in annualized revenue sometime between mid-2026 and 2027.

The enterprise API market tells the most revealing story. OpenAI's share of the combined OpenAI-Anthropic enterprise LLM API market has fallen from 50% in 2023 to 25% by mid-2025. Anthropic rose from 12% to 32% over the same period. Anthropic is now the enterprise API market leader — driven largely by Claude Code, which alone generates $2.5 billion in annualized revenue and is responsible for 4% of all public GitHub commits globally.

This matters for board due diligence on AI strategy. If Anthropic surpasses OpenAI in revenue before OpenAI lists, the IPO is no longer a validation of AI leadership. It is a snapshot of a competitive position that may already be in the past tense.

3. The Valuation Math Requires a Lot of Belief

At $1 trillion, OpenAI would be valued at approximately 38x projected 2026 revenue and 5x projected 2030 revenue. For context: Salesforce at peak traded at roughly 40x revenue. It was consistently profitable. OpenAI will not be profitable for four to five more years at current projections.

The bull case for 38x revenue without profits is the argument that AI is a winner-take-most market and OpenAI is the dominant player in it. The bull case requires three simultaneous assumptions: that AI revenue continues growing at 2-3x annually, that margins recover as compute costs stabilize, and that the competitive gap remains wide enough that Anthropic and others do not continue taking share.

The alternative: the IPO becomes the moment when public market investors apply normal software valuation discipline to an extraordinary growth story — and the correction ripples across every private AI company that has used OpenAI's valuation as a benchmark for its own.

4. The IPO Deadline Is Not Coming From the Inside

Sam Altman wants to go public in Q4 2026. CFO Sarah Friar is reportedly urging a more cautious timeline. The gap between them is not a personality conflict. It is a governance signal.

Altman's view: the narrative is strongest now. The competitive position is strong. The growth story is cleanest before Anthropic overtakes revenue. List before the story gets complicated.

Friar's view: the financial disclosures required in an S-1 will expose the $600 billion compute commitment, the $14 billion annual losses, and the 2030 breakeven timeline to every investor, analyst, and competitor simultaneously. IPO-readiness is not the same as IPO-desirability.

What neither is fully in control of: Amazon's $35 billion tranche is reportedly conditional on an IPO milestone. SoftBank's remaining $20 billion arrives in quarterly tranches — dependent on continued progress toward listing. The funding structure was built around the assumption that OpenAI goes public. The pressure to list is not internal ambition alone. It is contractual obligation to investors who structured their checks around a public market exit.

What This Means for Your Board

The OpenAI IPO is not a technology story. It is a capital markets stress test for the entire AI investment thesis.

If OpenAI lists at or near $1 trillion and the stock holds: AI valuations across private and public markets become permanently recalibrated upward. Every AI company — including your competitors, your vendors, and your portfolio — benefits from a legitimized benchmark.

If OpenAI lists and the stock struggles: the correction could exceed anything seen in the dot-com era. The $600 billion in compute commitments, the $57 billion burn rate, and the 2030 profit timeline all become public record. Investors who priced the company on growth narrative will be forced to price it on fundamentals.

For boards with AI exposure — whether through direct investments, API dependencies, or competitive positioning — the OpenAI IPO is the most consequential single corporate event in the technology sector since Google went public in 2004. The difference: Google was profitable when it listed. OpenAI is not.

Three questions every board should be able to answer:

1. What is our actual dependency on OpenAI's API pricing and availability? If OpenAI's margins compress post-IPO and pricing shifts to recover losses, what is our cost exposure?

2. Do we have Anthropic and/or open-source alternatives in our development pipeline? The competitive dynamic that led Anthropic to take enterprise API market share from 12% to 32% in two years could repeat rapidly in other verticals.

3. If OpenAI's IPO creates a market correction in AI valuations — as happened with dot-com pure-plays in 2000 — what is the impact on our own AI investment thesis? Are we building on the assumption that AI infrastructure costs will continue declining, or are we priced in to a trajectory that depends on continued access to cheap capital?

The $25 billion revenue number is the headline. The internal disagreement on IPO timing is the signal. The Anthropic trajectory is the context that determines whether the $1 trillion valuation is visionary or fictional.

The IPO will happen. The question is what it teaches us about the AI market once the cover story is removed.

Source: Multiple outlets including Reuters (March 5, 2026), The Information (April 5, 2026), Axios (March 18, 2026), Epoch AI, Sacra, InvestorPlace (April 24, 2026), and company disclosures.