As reported by CNBC, from remarks OpenAI CFO Sarah Friar made to employees.
Update (Aug 20, 2026) — leadership is a snapshot, not a trend. The comparison above captures a second-quarter photo: Anthropic ahead on revenue level and valuation. Early third-quarter signals complicate it. By spend-data trackers — Ramp’s data, per its economist Ara Kharazian — OpenAI’s enterprisegrowth appears to be reaccelerating, and the mechanism is specific. OpenAI’s new GPT-5.6 “Sol” is winning developers on price and performance (roughly a third the cost of Anthropic’s flagship on large workloads), capturing about a quarter of OpenAI’s token volume in its early weeks. Anthropic’s newest frontier model, Fable 5, stumbled the other way: it drew only about 6% of Anthropic’s tokens in its first month, held back by roughly double the price and a mandatory 30-day data-retention policy with no enterprise zero-retention option — a direct collision with the compliance requirements many large customers operate under. The caveats matter both ways. Token share is not revenue, one model’s slow start is not a company’s trajectory, and specific Q3 quarter-to-date growth figures circulating (on the order of ~82% for OpenAI versus ~76% for Anthropic) are early spend-tracker reads, not audited results. And Anthropic already has a price answer: Opus 5 lands within about half a percent of Fable 5’s peak at roughly half the cost, which addresses the exact friction hurting Fable 5. The honest revision, then: Anthropic still leads on level and valuation, but OpenAI has re-accelerated on growth — so the crossover is a live contest, not a settled one, and the “goes public second” framing is about timing and readiness, not a durable revenue verdict.
The category-defining name in AI guides to a 2027 listing while Anthropic — now larger by revenue and valuation — briefs investors for fall 2026. The sequence is a readiness choice, and the framing explains why.
IPO SEQUENCING — AS OF AUG 19, 2026
March 2026
OpenAI closes ~$122 billion raise — the capital buffer Friar cites as the reason it can choose its moment to list. (Self-reported; not audited.)
Q2 2026 (reported mid-Aug)
Anthropic posts ~$11.5B in Q2 revenue vs. OpenAI’s ~$6.7B — the first quarter Anthropic out-earns OpenAI. Anthropic’s annualized run rate surpasses $65B; OpenAI’s sits at roughly $40B. Anthropic carries positive adjusted operating income and a valuation of ~$965B vs. OpenAI’s ~$852B.
Fall 2026 (targeted)
Anthropic briefs investors toward a fall 2026 IPO offering.
2027 — or sooner (stated intention)
Friar tells an all-hands OpenAI “will be a public company in 2027,” with flexibility to move sooner “if our business continues to inflect.” This is internal guidance reported secondhand by CNBC — not a filed timeline or a firm calendar date.
At a Wednesday all-hands, CNBC reports, OpenAI CFO Sarah Friar told employees that the company “will be a public company in 2027” — though potentially sooner “if our business continues to inflect.” Two caveats belong in the first sentence: this is an internal statement reported secondhand, not a registration filing, so “2027, or sooner” is a stated intention and a range, not a date on a calendar. The growth figures Friar showed employees — revenue run rate up roughly 35% quarter-to-date, enterprise run rate up roughly 50% quarter-to-date, and an AI coding product at approximately 20 million weekly active users — are OpenAI’s own self-reported internal metrics, not audited results.
The sentence that carries the most structural weight is not the year. It is the reframe: “The IPO is not a finish line, it is a milestone, another fundraise.” Friar pointed to the roughly $122 billion OpenAI raised in March as the reason it has the flexibility to wait. Earlier reporting had noted Friar’s preference for a 2027 timeline over a faster path, against a backdrop of approximately $600 billion in infrastructure commitments — including the Ohio campus and the compute deals behind it — that would need to be funded. OpenAI and Altman did not publicly dispute that framing, and this week’s all-hands language is consistent with it, though the company has not confirmed the specifics of any internal disagreement beyond what prior reporting established.
The competitive context sharpens the picture. As of mid-August 2026, Anthropic has passed OpenAI on the numbers that public markets would price first: an annualized revenue run rate above $65 billion against OpenAI’s roughly $40 billion; Q2 revenue of approximately $11.5 billion against OpenAI’s roughly $6.7 billion — the first quarter in which Anthropic out-earned OpenAI; a valuation of approximately $965 billion against OpenAI’s roughly $852 billion; and positive adjusted operating income. Anthropic is the lab briefing investors for a fall 2026 offering. OpenAI — the household name, the category-defining brand — is the one guiding to 2027. (Full Anthropic Q2 breakdown here.)
The key insight: OpenAI’s framing of its IPO as “another fundraise” is not rhetorical modesty — it is the precise description of what a public offering means for a company carrying ~$600 billion in infrastructure commitments against ~$40 billion in annualized revenue. The listing is a financing event. You time a financing event for when the terms are best, not when the brand is loudest. The $122 billion in the bank is what buys the right to make that choice.
The Structural Read
There are three things happening simultaneously in this story, and conflating them produces the wrong read in all three directions.
First: the IPO as fundraise, not finish line. OpenAI has committed to something on the order of $600 billion in compute infrastructure — the Ohio campus and the procurement deals layered behind it. (See the vendor-financing loop analysis.) At roughly $40 billion in annualized revenue, the gap between what OpenAI earns and what it has promised to spend is the defining number in its story. A private company can narrate that gap as a bridge — a bet on future revenue, a deliberate investment phase, a scale argument. A public company must report the same gap in a 10-Q every ninety days, repriced by markets in real time. (Full revenue-vs-cost-gap analysis here.) Friar’s language — “another fundraise” — is not modesty. It is an accurate description of what the listing is for and a signal that the company will not list until the narrative around that gap is one it can defend quarterly.
Second: mindshare and revenue have decoupled. This is the part worth stating without softening. OpenAI defined the category. ChatGPT is the most recognized AI product name in the world. By consumer mindshare, brand recall, and cultural presence, OpenAI leads. By the numbers public markets would price on day one — run-rate revenue, quarterly revenue, valuation, and profitability — Anthropic has passed it. That is not a competitive defeat; run-rate leads in a market growing this fast are volatile, and a fall-2026-versus-2027 IPO ordering reflects readiness and strategy, not a settled scoreboard. But the intuition that the famous name goes first because it’s larger is wrong. The famous name goes second because it isn’t — not yet, on these metrics — and because it has the cash to wait until it is. (On the enterprise crossover that may close that gap.)
Third: the option value of $122 billion. The “sooner if our business inflects” clause is not a hedge — it is the actual strategy. The enterprise run rate growing roughly 50% quarter-to-date and roughly 20 million weekly active users on the AI coding product are the leading indicators OpenAI is watching. If those lines move fast enough that the revenue-versus-commitment gap begins to look like a bridge rather than a chasm, the company has both the financial runway and the stated intention to move the timeline forward. The $122 billion raise is not a comfort number; it is the option that makes the wait rational rather than forced. From the outside, strength and not-yet-ready look identical. Only the trajectory separates them, and that is exactly what Friar is promising to show before listing.
The Fifth AI Bottleneck — Applied
Capital is the constraint that makes timing matter
The fifth AI bottleneck is not compute, talent, data, or energy — it is the capital structure that funds them at scale. OpenAI’s IPO calculus is a live example: the company that needs the most infrastructure has the most to lose from going public before its revenue makes the commitment look fundable. Friar is not scheduling an event; she is managing a capital-structure risk. The listing happens when the enterprise inflection makes ~$600B of commitments look like a phased buildout rather than an open-ended liability. Full framework: The Fifth AI Bottleneck. Related: Beyond NVIDIA’s Moat.
Three Implications
IMPLICATION 1 — FOR ANTHROPIC’S IPO
Anthropic goes to public markets first, now carrying the higher revenue run rate and the higher valuation. That changes what its S-1 narrative looks like: it is no longer the challenger story but the current-numbers story. The risk is that public scrutiny arrives before the lead is consolidated; the upside is that it prices into a window before OpenAI’s listing resets the comparison set. Neither outcome is guaranteed, and run-rate leads in a market at this growth rate can compress quickly.
IMPLICATION 2 — FOR OPENAI’S ENTERPRISE BET
The ~50% quarter-to-date enterprisegrowth and ~20 million coding users are not incidental metrics — they are the specific signals OpenAI needs to move its IPO timeline forward. Enterprise revenue is more durable, more auditable, and more legible to institutional investors than consumer subscriptiongrowth. The faster the enterprise crossover accelerates, the shorter the window between “another fundraise” and a filed registration. The coding product’s weekly active user count is a proxy for that shift: developer adoption precedes enterprise procurement, and 20 million WAUs is a meaningful leading indicator if it converts. (Context on the enterprise crossover.)
IMPLICATION 3 — FOR HOW INVESTORS PRICE AI INFRASTRUCTURE BETS
When the most recognized name in AI frames its own IPO as a financing event for a ~$600 billion infrastructure commitment, it signals to every infrastructure investor and compute supplier what the actual constraint is. The question public markets will be asked to answer — for OpenAI in 2027 and for every hyperscaler-adjacent company behind it — is whether the revenue trajectory can grow into commitments made at the peak of a buildout cycle. That is a capital-allocation question dressed up as an equity story, and both Friar’s language and the delayed timeline reflect how difficult that question is to answer convincingly before the enterprise inflection is fully visible. (On the vendor-financing loop.)
Business Engineer Framework
The Fifth AI Bottleneck: Capital Structure as Competitive Constraint
The compute wars, talent markets, and data moats get most of the analysis. The fifth bottleneck — the capital structure that funds all four — shapes which companies can afford to wait and which are forced to the market on someone else’s timeline. OpenAI’s IPO framing is a direct application: $122 billion in the bank converts a public offering from an obligation into an option. The Map of AI places every major lab, infrastructure provider, and enterprise buyer inside the same capital-constraint system.
OpenAI’s CFO told employees the IPO is “another fundraise” — and that framing is more informative than the year. The most recognized name in AI now trails Anthropic on run-rate revenue (~$40B vs. >$65B), Q2 revenue (~$6.7B vs. ~$11.5B), and valuation (~$852B vs. ~$965B), and goes
91,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
Gennaro is the creator of FourWeekMBA, which reached about four million business people, comprising C-level executives, investors, analysts, product managers, and aspiring digital entrepreneurs in 2022 alone | He is also Director of Sales for a high-tech scaleup in the AI Industry | In 2012, Gennaro earned an International MBA with emphasis on Corporate Finance and Business Strategy.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
