What happened
In an interview at the Wall Street Journal’s Tech conference, OpenAI co-founder and president Greg Brockman described Q1 and Q2 of 2026 as “pretty rough.” He said OpenAI appeared not to be part of the conversation in coding, despite having Codex, and that customers were not talking about it. The account is Brockman’s characterization of the company’s position, not a quantified performance report.
Brockman said the picture changed in Q3, when OpenAI’s new tools for agentic software coding found traction. He cited the release of a new frontier model, GPT-5.6 Sol, and ChatGPT Work as setting the stage for a comeback. He described the period as the result of “painful behind-the-scenes work” and said enterprise sales became OpenAI’s biggest revenue driver, surpassing consumer sales. The article does not give revenue figures for either business or break down the contribution of the model, ChatGPT Work, or coding tools.
The Journal points to third-party evidence of a competitive shift: a September analysis from OpenRouter covering roughly 120,000 companies that use Anthropic and OpenAI tools found that their spending share was approximately even. At the beginning of the year, Anthropic had commanded about three-quarters of that total. The supplied account does not provide the underlying methodology or absolute spending amounts, so the figures describe relative share within that dataset rather than the entire enterprise market.
Brockman said OpenAI had recently formed a strategic partnership to provide AI services to JPMorgan Chase, including access to cybersecurity-focused models through its Daybreak program. JPMorgan is also a partner in Anthropic’s Project Glasswing, which provides early access to Anthropic cybersecurity models. On corporate financing, the Journal reports that OpenAI filed IPO paperwork in June and now plans to go public in 2027. The article also says the company had raised less than half of the $30 billion initially targeted in its latest private funding round, now seeking $20 billion at a $1.25 trillion valuation rather than the initially sought $1.4 trillion.
What the archive adds
This is a new development in a broader OpenAI business story, not the same event as OpenAI’s reported revenue projection gap. That earlier note focused on a measurement-boundary problem; Brockman’s comments and the OpenRouter comparison add a claim about a third-quarter inflection and a shift in relative enterprise spending.
Trend signal
This supports, weakly, the pattern that AI-native organizations are moving toward review and orchestration rather than individual production speed: enterprise buying appears to be responding to a coding-agent product suite, not just a model announcement. The account does not establish how teams deploy or verify these tools, so it is evidence of commercial traction, not proof of an organizational redesign. AI-native organizations move toward review and orchestration.
Why now
The comments arrive as Anthropic prepares for a possible public listing as soon as this month, according to the Journal, while OpenAI has moved its own expected listing to 2027. Brockman’s comparison of a rough first half with a stronger third quarter is also set against fundraising pressure and competition for enterprise customers.
Connections to other events
The archive has also covered OpenAI’s reported revenue outlook gap. The two stories concern different evidence: the earlier report questioned how a revenue figure was measured, while this one describes relative spending among a sample of companies and Brockman’s account of enterprise sales becoming the company’s largest revenue source. Neither alone establishes the company’s overall financial trajectory.
Why it matters
For teams choosing coding assistants or enterprise AI suppliers, the story is a reminder that product adoption and vendor economics can change quickly. A company that appeared to lag in coding earlier in the year may, according to its president, gain traction after a coordinated set of model and product releases. The OpenRouter comparison offers a useful directional signal, but its sample and relative-share framing mean procurement teams should not treat it as a market-wide ranking.
The JPMorgan example also shows that large buyers may work with competing providers at once: the bank is linked to OpenAI’s Daybreak and Anthropic’s Glasswing. For builders, that favors evaluating tools on task fit, security, integration, and governance rather than assuming a single vendor relationship. The source gives no comparative product performance data, so it cannot support a conclusion about which coding assistant teams should adopt.
Limits and uncertainty
Most of the account’s interpretation comes from Brockman, who has an obvious interest in presenting OpenAI’s recent performance positively. The article supplies no audited revenue figures, customer counts for ChatGPT Work or Codex, usage data, retention measures, or causal analysis separating the effects of the model and product launches. OpenRouter’s roughly even spending share applies to the companies in its analysis, not necessarily all enterprise AI purchases. The story also does not describe product availability, pricing, technical capabilities, or the terms of the JPMorgan partnership.
Reader takeaway
OpenAI’s president says coding tools helped turn a difficult first half into a stronger third quarter, and a cited company-spending sample shows OpenAI closing a gap with Anthropic. Treat this as a meaningful signal of competitive momentum—not a product benchmark or proof that the same results will transfer to your team.