OpenAI President: H1 Was Quite Tough; Q3 Annualized Revenue Near $70 Billion as Enterprise Revenue Exceeds Consumer for First Time

On November 4, 2026, OpenAI President Greg Brockman publicly acknowledged at a Wall Street Journal tech event in Napa, California, that the company’s revenue performance in the first half of 2026 was “quite difficult.” He then gave Q3 figures: annualized revenue of nearly $70 billion, an annualized growth rate of about 70% in the quarter, enterprise B2B revenue rising from about 30% at the start of the year to surpass 50% for the first time, and enterprise revenue doubling in the quarter. On the same day, he also disclosed a previously undisclosed JPMorgan enterprise services contract.

To understand the weight of H1’s “difficulty,” one must first reconstruct the competitive landscape. According to Yahoo Finance, Anthropic’s quarterly revenue reached $11.6 billion in Q2 2026, while OpenAI’s was only $6.7 billion over the same period—a rare moment when Anthropic briefly surpassed OpenAI in quarterly revenue. By July 2026, Anthropic’s annualized revenue had approached $65 billion. This context shows that OpenAI’s H1 predicament was not simply a slowdown in growth, but being overtaken by a competitor by a clear margin in the most strategically valuable enterprise customer market.

Brockman attributed Q3’s turnaround to the launch of GPT-5.6 Sol and the ChatGPT Work enterprise service, describing the introduction of these two products as a “critical revenue inflection point.” The trajectory of enterprise revenue share shows the acceleration: about 30% at the start of the year, rising to slightly above 40% by March, surpassing half for the first time by the end of Q3, with revenue contributed by enterprise customers doubling in the quarter. This curve is not linear penetration; it looks more like a dense cluster of deals after a certain tipping point—large enterprise contracts moving from “in negotiation” to “signed” often follow this rhythm: a long early cycle, followed by concentrated conversion once the ice is broken.

The disclosure of the JPMorgan contract was the most commercially meaningful detail of the day. According to Brockman at the event, the contract covers three areas—Codex, cybersecurity models, and knowledge-work scenarios—which correspond exactly to the business lines where large financial institutions are most willing to spend to cut costs and improve efficiency. Meanwhile, Codex currently has more than 2 million weekly active users, up 70% month over month, and its scale has expanded fivefold within three months. A deal of JPMorgan’s size has a clear demonstration effect for similar institutions in highly regulated industries such as finance, law, and healthcare—it can significantly reduce the internal political resistance to procurement decisions.

What the Structural Shift in the Enterprise Market Means for Three Types of Participants

For enterprise IT buyers, the structural shift of enterprise revenue exceeding consumer revenue for the first time means that AI underlying platforms have moved from “tools in the exploration phase” to “infrastructure on the budget line.” At this stage, procurement decision cycles shorten, but requirements for vendor stability and compliance capabilities rise significantly, and bargaining power in contract negotiations will concentrate among vendors with mass-deployment capabilities and large reference customers.

For the developer ecosystem, Codex’s rapid growth is a double-edged sword. Enterprises’ willingness to pay proves that the tool’s value has been realized, moving past the early stage of “good to use but hard to sell”; but as OpenAI positions coding agents as a core selling point of B2B contracts, direct competition with products such as GitHub Copilot and Anthropic Claude Enterprise will intensify, price pressure will be transmitted downward across the entire market, and the room for independent tool vendors to survive will be further compressed.

For the capital markets, the tilt in the balance between consumer and enterprise revenue directly affects valuation logic. Consumer subscription revenue is characterized by high volatility, high customer acquisition costs, and retention that depends on product competitiveness; enterprise contract revenue has a better structure on key metrics such as contract stability, average contract value, and renewal rates, and typically commands a higher revenue multiple in the IPO narrative. This explains why Brockman, despite having abundant data, chose to emphasize “enterprise revenue surpassing half for the first time” rather than absolute revenue figures.

Historical Comparison: Annualized Revenue Expanded About Fivefold Within Five Years

Stretching out the timeline makes the magnitude of this shift clearer. According to Yahoo Finance, OpenAI’s full-year 2025 revenue was $13.07 billion, with a net loss of $38.5 billion that year; by Q3 2026, annualized revenue had approached $70 billion. On that basis, annualized revenue achieved roughly fivefold growth in less than two years, while the enterprise revenue share rose from below 30% to more than half—a qualitative change in growth structure occurring alongside a quantitative change in scale, which is uncommon.

From a competitive landscape perspective, in Q2 OpenAI’s quarterly revenue was overtaken by Anthropic by a ratio of $11.6 billion to $6.7 billion. This detail reflects not merely the failure of a particular product, but a lag in the overall enterprise market strategy at a certain stage. The core pressure behind H1’s “difficulty” came from Anthropic’s rapid penetration on the enterprise customer side, along with the resulting pricing pressure and sales-cycle competition. Comparing Q3’s annualized near $70 billion with Anthropic’s July annualized roughly $65 billion, the absolute gap between them is already quite limited—this is no longer a catch-up game with a huge disparity in scale, but substantive head-to-head competition.

Head-on Clash of IPO Narratives

The timing of Brockman’s public remarks at WSJ Tech is itself a capital-market signal. Just weeks earlier, Anthropic had already filed its S-1, with H1 revenue of $16.3 billion becoming a publicly verifiable reference point. Both companies are in a pre-IPO window, releasing revenue narratives in the same period, extending from product competition to a fight over the power to define the capital markets. Anthropic seized the information timeline by filing its S-1 earlier; OpenAI anchored in the opposite direction with the Q3 annualized figure of nearly $70 billion, emphasizing that its revenue structure has shifted toward enterprise. When asked whether OpenAI would be ready for an IPO in 2027, Brockman answered, “I think so”—in the capital markets, such an answer usually means the process is already underway, not merely a possibility.

The following is analytical judgment and does not constitute a statement of fact. The retention structure of enterprise contracts—whether Q3’s high growth came from multi-year agreements or short-term pilot contracts—will determine whether this curve can continue upward in Q4 and 2027; Anthropic’s IPO pricing range: if Anthropic completes its listing before the end of 2026, its valuation will immediately become an implicit benchmark for measuring OpenAI, and the capital markets will need to answer whether H1’s “difficulty” or Q3’s “reversal” better represents OpenAI’s fundamentals; the pace of OpenAI’s narrowing losses: against the backdrop of a $38.5 billion net loss in 2025, even with astonishing revenue growth, the question of its path to profitability will not disappear during the IPO roadshow.