OpenAI's third-quarter revenue passed $10 billion for the first time, according to reporting on its investor call, up about 50% from the prior quarter. The previous quarter grew roughly 17%, which makes the acceleration the more notable part of the story. The same call disclosed an adjusted operating loss of about $10.9 billion, wider than the $10.1 billion before it, and roughly 100 million paying users.

The company also closed an initial $11 billion tranche of a round valuing it at $1.25 trillion. Its president said publicly last week that OpenAI will be ready to go public in 2027.

Where the growth came from

Consumer subscriptions do not explain a $3 billion-plus sequential jump on a base of 100 million paying users. Enterprise contracts and API volume fit better: higher contract values, longer terms, more predictable renewals.

The widening loss is consistent with that mix. Inference cost scales with volume, and enterprise deals carry volume discounts. Revenue quality improved; gross margin per dollar is still being eaten by compute.

Businesses in ChatGPT is the more strategic move

The beta lets companies claim a brand profile inside ChatGPT, maintain their listed information, sync product feeds through Shopify Catalog, and see how users engage. Eligible US ecommerce and online-service businesses can deploy a Business Agent that answers questions and closes sales inside the conversation.

This is the same playbook search engines ran with local business listings two decades ago: once people ask an assistant "where can I get this fixed nearby," the platform needs a structured merchant database to answer, and a structured database is where ads and commissions live.

Three implications for merchants:

Accuracy becomes a traffic channel. Wrong opening hours, stale inventory or an outdated returns policy will be repeated back to a customer at the last step before purchase.

Support economics shift. An agent absorbs the standard questions, but only if you define escalation clearly: what must go to a human, and what context travels with it.

Do not wire it into everything yet. The beta is US-only, and rules, fees and eligible categories will move. Run a narrow pilot first.

The 2027 timeline matters more than the quarter

Put OpenAI's IPO-readiness comment next to Anthropic's live roadshow and the shape of the next two years becomes visible. Model companies are entering the phase where capital markets grade them on revenue slope and loss trajectory, and that grade determines the price of long-term money.

For buyers, this competition is mostly good: price pressure, faster iteration, more API compatibility. The real risk is platform lock-in. Wrapping business logic behind a replaceable interface layer is worth more than picking a winner.

One caveat on the numbers themselves. Revenue and loss figures reported through investor calls are not audited financials, and comparisons between the two companies use different adjustments. Treat them as directional, not as a scoreboard.