OpenAI's $6.7B quarter, $12.3B loss and the case against a single-vendor model bet

OpenAI told investors it generated $6.7 billion in the three months to June, up from $5.7 billion in the first quarter, an 18% sequential increase, while its operating loss widened to $12.3 billion from $9.3 billion, according to a Wall Street Journal report sourced to undisclosed people familiar with the company's finances ^{[1][2][3][4][8]}. One day earlier, Anthropic disclosed second-quarter revenue up more than 50% sequentially to $11.6 billion and its first operating profit, reportedly $559 million ^[5][6][7]. That divergence is the substance of the story, and it lands while an OpenAI initial public offering is widely expected as soon as later this year ^[10].

Work the arithmetic before the narrative. Revenue rose about $1.0 billion quarter on quarter; the operating loss rose about $3.0 billion, roughly three times as much ^[1]. The loss grew about 32% against 18% revenue growth ^[2], which is the mechanical version of the source's observation that losses are expanding faster than revenue ^[19]. That puts OpenAI's second-quarter operating margin near negative 184% ^[3] and OpenAI has said its margin moved further into the red ^[9]. Anthropic's reported margin on the same line is roughly positive 4.8% ^[4], on revenue about 1.7 times OpenAI's ^[5]. The spread between the two companies' operating results in a single quarter is on the order of $12.9 billion ^[6].

The composition matters more than the totals for buyers. ChatGPT's growth has stalled, while Anthropic's coding tool Claude Code has done well specifically with enterprise customers ^[11]. OpenAI's response has been organisational: a string of senior departures including Chief Revenue Officer Denise Dresser after less than a year, former Chief Operating Officer Brad Lightcap and Fidji Simo, alongside a pivot toward agents that automate business work ^[12]. President Greg Brockman has taken a larger role in product and business development ^[18].

OpenAI's counterargument is that the quarter is stale. It told investors privately that growth accelerated after a new model generation launched in July, without providing numbers ^[16], and says a new combined application bundling ChatGPT with its Codex coding tool and an AI-native browser is growing quickly ^[17]. Neither claim is currently checkable. What is checkable is that OpenAI's sequential growth trailed other AI-cycle beneficiaries including CoreWeave, Micron and Palantir ^[15].

The structural risk is not OpenAI's balance sheet alone. The company has raised roughly $180 billion, spent heavily on data centres and cloud contracts, and those commitments assume it eventually reaches hundreds of billions in annual revenue; a shortfall would be felt in the share prices of Nvidia, Oracle and other suppliers ^[13][14]. For a procurement team, that is a vendor-continuity variable, not a market-commentary one.

Two caveats keep this from being a decision on its own. The OpenAI figures come from unnamed sources rather than audited disclosure ^[8], and one quarter of Anthropic operating profit is a data point, not a durable cost advantage ^[6][7].

What to watch: whether the third quarter substantiates OpenAI's claim of post-July acceleration with actual figures ^[16]; whether Anthropic's operating profit repeats or was a one-quarter artefact ^[6]; whether OpenAI's agent and bundled-app pivot converts into enterprise seats rather than consumer usage ^[17][11]; and how the loss trajectory is framed in IPO documentation, since that is the first version of these numbers anyone will be able to hold the company to ^[10][4].