OpenAI's Q2 Numbers Reveal a New Problem: Anthropic Is Out-Earning It

For most of the modern AI era, the industry’s pecking order has been simple: OpenAI sells the most, spends the most, and loses the most — and everyone else races behind. The second quarter of 2026 has quietly broken that assumption. According to a Wall Street Journal report published August 18, OpenAI told investors that its revenue grew 18% quarter-over-quarter to $6.7 billion in the three months ended June, up from $5.7 billion in Q1. That growth rate, once considered explosive in any other software category, is now being read as a disappointment — because Anthropic, its closest rival, is growing several times faster.

The numbers behind the WSJ report

The Journal’s figures paint an uncomfortable picture for the company that defined the generative AI boom:

• Revenue: $6.7 billion in Q2 2026, up 18% from $5.7 billion in Q1.
• Operating loss: $12.3 billion, including stock-based compensation, widened from $9.3 billion in the first quarter.
• Losses are growing faster than revenue. OpenAI added roughly $1 billion of quarterly revenue sequentially while adding roughly $3 billion of quarterly losses.

Some shareholders described the results as tepid, per the report — a striking word for a company still posting double-digit sequential growth. The comparison that stings is Anthropic’s: preliminary figures first reported by CNBC put Anthropic’s Q2 revenue at more than $11.5 billion, up from $787 million in the same quarter a year ago — a roughly 14x year-over-year jump. Forbes later confirmed the quarter delivered Anthropic’s first-ever positive operating result, with an adjusted operating profit reported around $559 million against expectations set in May.

In other words, for the first time in the two companies’ histories, Anthropic out-earned OpenAI in a quarter — by a wide margin — and did so while covering its own costs, something no frontier AI lab had previously managed.

Why the gap opened

The divergence has been building for months, and the drivers are structural rather than cyclical.

First, Claude Code became the enterprise wedge nobody fully priced in. Anthropic’s terminal-based coding agent rode the agentic-coding wave of late 2025 and early 2026, converting developers into seat-holder evangelists inside enterprises. By the end of May, Anthropic had disclosed run-rate revenue of roughly $47 billion — already ahead of OpenAI’s then-trajectory — and analysts attributed the largest single share of that to Claude Code and adjacent API usage. OpenAI’s counter-offensive in coding tooling arrived, but later and with less enterprise lock-in.

Second, Anthropic’s revenue is disproportionately API- and enterprise-driven, which monetizes more directly per token of value delivered than consumer subscriptions. OpenAI’s revenue remains heavier in consumer ChatGPT subscriptions, a market showing early signs of saturation in mature economies — a factor behind the company’s recent teen-focused expansion pushes.

Third, the cost structures are diverging. OpenAI is carrying the financing burden of the largest infrastructure buildout in software history, with hundreds of billions in multi-year compute commitments. Its Q2 loss widening by $3 billion on a $1 billion revenue gain illustrates the math problem: every marginal revenue dollar currently arrives with several dollars of committed spend behind it. Anthropic, leaning heavily on partner cloud capacity and a leaner headcount story, reached a quarter where revenue exceeded operating costs.

The counter-argument: July changed the picture

OpenAI’s defenders — and the company itself — would push back on reading Q2 as a trend line. On July 29, CNBC reported that CFO Sarah Friar told employees that OpenAI’s annualized recurring revenue in July exceeded the entire second quarter’s total. “And Q2 was no slouch,” Friar said. Reddit and analyst estimates pegged July ARR somewhere in the $29–40 billion range, with figures around $34 billion commonly cited. For context, audited financials reported by Yahoo Finance put OpenAI’s 2025 net loss at $38.5 billion on $13.07 billion of revenue — meaning the company’s annualized revenue base has roughly tripled in about a year even as the growth rate decelerates.

The bull case is that OpenAI’s growth is lumpy, gated by capacity and product cycles rather than demand, and that the July ARR print shows the underlying curve remains steep. The bear case is that Anthropic’s curve is steeper still, and that in a market this young, relative growth momentum — not absolute scale — is what determines which platform’s ecosystem compounds.

What it means for the IPO race

Both companies are widely reported to be racing toward public markets, and the Q2 crossover lands at the worst possible moment for OpenAI’s narrative. Anthropic filed for an IPO in late spring with a growth story — 14x year-over-year, first operating profit, nearly $1 trillion in discussed valuations — that public-market investors can map onto classic software economics. OpenAI’s story is harder to underwrite: extraordinary scale, decelerating relative growth, and losses that widened to $12.3 billion in a single quarter against a backdrop of titanic, partially debt-financed compute commitments.

None of this makes OpenAI’s position precarious — $6.7 billion of quarterly revenue with July ARR reportedly topping the entire quarter is a business most companies would kill for. But the WSJ report crystallizes a shift that was already underway: the frontier AI market is no longer a one-company story, and on at least one quarter’s evidence, it isn’t even a one-leader story anymore. The second half of 2026 will show whether Q2 was OpenAI’s air pocket or Anthropic’s permanent breakout.

Figures in this article are drawn from investor disclosures reported by the Wall Street Journal, CNBC, Investing.com, and Forbes between May and August 2026. Anthropic’s Q2 figures are described as preliminary by the original outlets.