Multiple reports now align on a picture of OpenAI under pre-IPO stress. Apptopia's analysis shows ChatGPT mobile-app download growth peaked after April 2025. OpenAI missed an internal goal of 1 billion weekly active ChatGPT users by end of 2025, reaching 900 million by February 2026. The Wall Street Journal reports OpenAI also missed monthly revenue targets in early 2026 and is losing ground to Google's Gemini in consumer markets and to Anthropic in coding and enterprise. CFO Sarah Friar has reportedly told colleagues OpenAI is not organizationally ready for the Q4 2026 IPO Sam Altman wants to pursue — Friar prefers a 2027 listing.
The bifurcation matters more than the topline. Gemini is outpacing ChatGPT specifically on three measurable axes per the Apptopia data: download growth, monthly-active-user growth, and time-spent-in-app. Those three together are the consumer-funnel metrics that drive future revenue, not just current revenue. On the enterprise/coding side, Anthropic is taking share — consistent with the MCP Atlas benchmark we covered earlier this week, where Claude Opus 4.7 leads protocol-tool-use and OpenAI left GPT-5.5's score blank in its own table. The picture is: consumer surface saturated and being out-iterated by Google's distribution; enterprise coding being out-quality-ed by Anthropic's MCP-native stack. Both are real, both have been visible for two quarters, and the IPO timeline is what is now forcing the public conversation.
Three patterns matter. First, the Oracle compute deal we covered yesterday gets harder to underwrite. Oracle's $300B compute commitment from OpenAI assumes OpenAI's revenue trajectory matures fast enough to pay; if user growth has peaked and revenue is missing targets, the Oracle CDS spread should widen, and watching it is the cleanest market read on this. Second, the Tumbler Ridge lawsuits we covered Tuesday are explicitly timed against the same pre-IPO disclosure window. Plaintiffs' attorneys generally maximize disclosure pressure in the weeks leading to a public offering; if Friar pushes the IPO to 2027, the lawsuits' tactical timing changes. Third, the OpenAI/Anthropic policy split that crystallized this week — Anthropic refusing Pentagon contracts, OpenAI publishing the cybersecurity-defender-democratization plan (TAC) — becomes a market-positioning bet. Anthropic gains enterprise/coding mindshare; OpenAI doubles down on consumer plus national security as the growth story. Both are coherent strategies; only one ages well into a delayed IPO.
For builders, three concrete things. First, your model-vendor risk profile changed this week. If your business depends on OpenAI's API economics, the company missing revenue targets while paying $300B to Oracle and expanding TAC defender-access programs means the cost base is rising while the consumer revenue that funds it is slowing. Diversify your inference suppliers; treat OpenAI exposure as financing-counterparty risk per yesterday's Oracle piece. Second, the consumer-vs-enterprise bifurcation is now scoped: Gemini wins consumer surface, Anthropic wins enterprise/coding. If you build a product that uses LLMs, your model choice should now match where your customers live, not which lab has the best general-purpose benchmarks. Third, watch the Friar-vs-Altman IPO timing dispute. If OpenAI announces a 2027 listing, Oracle's bet is the first thing to reprice in the credit markets — and that pricing will lead the public reporting by weeks. The CFO-vs-CEO split is the kind of pre-IPO friction that historically resolves through one of two routes: the CEO wins and the IPO crashes, or the CFO wins and the company tightens spending. Either outcome changes the rate at which OpenAI products get refreshed and priced.