The Last Two Times, OpenAI Announced It Itself. This Time Nobody Did.

On July 29 a story went up on The Information under the headline "OpenAI's ChatGPT Nears 1 Billion Weekly Active Users Seven Months After Target." Stephanie Palazzolo and Erin Woo wrote it, and the sourcing was internal OpenAI data. The claim: ChatGPT's weekly active users are approaching one billion.

Start with the verb. It's "nears," not "surpasses." And the party announcing it is a publication, not the company. That distinction carries more weight than it looks like it should.

Think about how the last two milestones landed. On October 6, 2025, Sam Altman stood on the DevDay stage and said more than 800 million people use ChatGPT every week. On February 27, 2026, OpenAI published a post titled "Scaling AI for everyone" that stated, in the company's own words, "more than 900M weekly active users." Those were numbers OpenAI wanted to shout about, and it shouted. This time there is nothing on OpenAI's blog and nothing from Altman. Reporters had to go get the number.

So as of right now, the last figure OpenAI has actually confirmed is 900 million. One billion is a reported estimate the company has never validated. Confuse those two and everything downstream falls apart.

And the real story lives in the second half of that headline: "Seven Months After Target." On March 31, in its own funding announcement, OpenAI wrote that it "was the fastest technology platform to reach 10 million users, the fastest to 100 million users, and soon the fastest to 1 billion weekly active users." That "soon" is now in its fourth month.

The Four Names Attached to This Number

Start with OpenAI. Since ChatGPT launched in November 2022, the company has rewritten the record book on how fast software reaches people, and that record has been structural to how it describes itself. In the March 31 post, OpenAI wrote that it is "now generating $2B in revenue per month" and that it is "growing revenue four times faster than the companies who defined the Internet and mobile eras, including Alphabet and Meta." Speed is the identity. Which is exactly why the phrase "seven months after target" lands differently at this company than it would at almost any other.

Second is The Information itself, a subscription publication that lives on Silicon Valley internals and has broken OpenAI financial and organizational stories repeatedly. This piece explicitly cites company data. It also sits behind a paywall, which means the full text cannot be cross-checked directly — only the portions that downstream outlets quote. Every number in this story therefore travels with a tag: as reported by The Information, not as confirmed by OpenAI.

Third is Google. One week earlier, on July 22, Alphabet reported second-quarter 2026 results, and Sundar Pichai wrote the numbers out on Google's own blog. The Gemini app: 950 million monthly active users. Daily active users tripled over the past year. Google's model APIs processing roughly 22 billion tokens per minute. And AI Overviews plus AI Mode combined have "surpassed 1 billion monthly active users." Alphabet's quarterly revenue was $119.8 billion, with Google Cloud at $24.8 billion, up 82%. While ChatGPT is still short of a billion, Google is already saying a billion out loud with numbers of its own.

Fourth is Anthropic and Meta, who are playing different games entirely. Anthropic has effectively conceded consumer scale and driven into enterprise and coding workloads, and has been reported to have passed OpenAI on revenue from enterprise customers. Meta reported Q2 2026 results on July 29 showing 3.60 billion people using at least one of its apps daily in June, with 2026 capital expenditure guidance of $130 billion to $145 billion. Mark Zuckerberg's framing was that "AI is accelerating our core business today." On the user-count scoreboard, Meta was never really playing by the same rules.

The Curve Bent, and the Table Shows Exactly When

The Information's reporting condenses into two claims. First, ChatGPT got to this scale in under four years and remains among the fastest-growing apps in internet history. Second, the road was rough — growth slowed last fall after user backlash to the GPT-5 model and amid intensifying competition.

That second claim is the actual story. Line up the milestones chronologically and the shape becomes obvious.

Date Weekly active users Who announced it Time since prior milestone
Nov 6, 2023 100 million OpenAI (DevDay) —
Feb 2025 400 million OpenAI —
Aug 4, 2025 approaching 700 million OpenAI ~6 months for +300M
Oct 6, 2025 800 million Sam Altman (DevDay stage) ~2 months for +100M
Feb 27, 2026 900 million OpenAI official blog ~4.7 months for +100M
Jul 29, 2026 nearing 1 billion (reported) The Information (internal OpenAI data) ~5 months for roughly +100M

From 400 million in February 2025 to 800 million in October 2025 is 400 million users in eight months — roughly 50 million net new weekly users per month. Getting from 800 to 900 took 4.7 months. Getting from 900 to somewhere near a billion has taken about five. That is a run rate in the low 20-millions per month. The absolute numbers are still staggering. The rate of change has been cut by more than half.

Layer OpenAI's own disclosures on top and the picture sharpens further. The February 27 post reported more than 50 million consumer subscribers and more than 9 million paying business users, and said January and February were "on track to be the largest months for new subscribers in our history." The March 31 post said enterprise is "more than 40% of our revenue" and "on track to reach parity with consumer by the end of 2026." It also said Codex had passed 2 million weekly users, up 5x in three months, that OpenAI's APIs process more than 15 billion tokens per minute, and that an advertising pilot "reached more than $100 million in ARR in under six weeks."

Here is how to read that combination. Free-user growth has clearly decelerated, while revenue per user and enterprise penetration have accelerated. It is probably not a coincidence that OpenAI's recent announcements lead with subscribers, business seats, and token throughput rather than with weekly actives. The metric the company most wants to put in a headline appears to be changing.

What Each Party Actually Gets Out of This Number

For OpenAI, one billion is still a trophy worth having, for three reasons. First, advertising. The March 31 post disclosed an ads pilot clearing $100 million in annualized revenue in under six weeks, and ad businesses are priced on reach. "A billion people a week" is not a vanity line, it is a media sales deck. Second, enterprise selling. OpenAI's repeated argument is that consumer scale is a distribution channel into the workplace — people who use it at home push to use it at work — and the wider the top of that funnel, the easier the sales motion. Third, capital markets. OpenAI closed its round on March 31 with $122 billion in committed capital at an $852 billion post-money valuation, and IPO chatter has followed the company ever since. In that narrative, "fastest ever to a billion weekly" is an expensive sentence to give up.

There is also a readable reason OpenAI might be sitting on the announcement. Either it has not actually crossed the line yet, or it has, and announcing it now guarantees the "seven months late" tag rides along in every headline. Waiting until the number is unambiguous and framing it in the company's own language is a defensible call. That said, this is inference rather than confirmed fact — OpenAI has not explained its silence.

What The Information gets is straightforward. Saying the number the company will not say is the whole product, and this is a clean example of it. The more interesting second-order effect is that when a company stops updating its own metrics for five months, the industry's reference number migrates from corporate disclosure to journalism. For the entire stretch between February and July, the figure the market reasoned with was one that reporters assembled.

What Google gets is the framing. Gemini's 950 million is monthly; ChatGPT's billion is weekly, and weekly is a considerably harder bar — someone who shows up every seven days is a different animal from someone who shows up once a month. But in a headline, 950 million and 1 billion sit next to each other, and the impression that forms is "basically caught up." Google has no incentive whatsoever to correct that impression. And the AI Overviews plus AI Mode figure of a billion monthly is a different animal again, because it requires no app install at all — the distribution is already there.

Advertisers and enterprise buyers get something real too. When two platforms are perceived as comparably large, negotiating leverage evens out. For anyone trying to avoid single-vendor dependence on their AI stack, that is a useful development.

Companies That Hit a Billion Before — Facebook and Twitter

Facebook got to a billion first. It announced one billion monthly active users on October 4, 2012, and the timing could hardly have been worse. The stock had IPO'd that May at $38 and had fallen into the $17 range by September, and the only question the market cared about was whether Facebook could make money on mobile. A billion users did not answer that question. What answered it was the next two years of News Feed advertising, and the stock recovered when mobile revenue share climbed, not when the user counter ticked over. The lesson is clean: reach is the starting line, not the finish.

The failure case is Twitter. Twitter made user count its headline metric, and the market obligingly valued the company on it — which was fine right up until the number stopped moving in the 300-millions. In its fourth-quarter 2015 results, reported in February 2016, monthly active users declined sequentially for the first time. From that moment the story about Twitter permanently changed from "what should the product be" to "why isn't it growing." The service's cultural influence was enormous and it did not matter. That is the bill a company pays for making a growth metric its identity.

And there is a third case whose shape matches the GPT-5 episode almost exactly: Snapchat. Snap pushed through a sweeping app redesign in early 2018, existing users revolted, and daily active users turned negative that year. The company insisted the product was better. Users experienced the loss of familiarity as a downgrade. When OpenAI shipped GPT-5 in August 2025 and retired GPT-4o, the same dynamic played out — Altman publicly conceded "I think we totally screwed up some things on the rollout," and GPT-4o was restored within days. The Information cites that backlash as one driver of last fall's slowdown.

Stack the three together and you get the frame. Facebook hit a billion and still had two years of proving left. Twitter's other strengths became irrelevant the moment growth stalled. Snap's improvement broke its own retention. Which of those three curves ChatGPT is on is not yet visible. What is visible, in the table above, is that its growth rate has roughly halved.

How Google, Anthropic and Meta Are Running the Numbers

Google's play is distribution. Gemini rides on Android, Chrome, Search, Gmail and YouTube — surfaces where the user never has to decide to install anything. What Pichai emphasized on the Q2 call was that combination, more than any single app number. The 950 million monthly actives on the Gemini app sit alongside AI Overviews and AI Mode passing a billion monthly, which means the largest traffic gateway on earth has itself been converted into an AI surface. OpenAI built its scale with a single app and a lot of demand creation. Google is building its scale by renovating habits people already have. The customer acquisition math is not remotely the same.

Anthropic's play is to use a different scale entirely. It gave up on matching OpenAI's consumer footprint and went at enterprise APIs and coding workloads, and has been reported to have passed OpenAI on enterprise-customer revenue. The advantage is that nobody evaluates Anthropic on weekly active users, so it is structurally immune to this particular headline cycle. The disadvantages are real too: weak consumer mindshare eventually costs you in recruiting and policy influence, and enterprise budgets move with the macro cycle in a way consumer subscriptions do not.

Meta is on a third road — not fighting for a standalone chatbot audience, but grafting AI onto an installed base of 3.6 billion daily users. Q2 revenue was $60.8 billion, but total expenses climbed 55% to $42.0 billion and net income fell to $15.8 billion, weighed down by $1.18 billion in severance tied to an 8,000-person headcount reduction and $2.4 billion in legal charges. Capital expenditure guidance for 2026 is $130 billion to $145 billion, against $72.2 billion in 2025 — close to a doubling. Meta's bet is that the user-acquisition contest is already settled and the only question left is what you put on top.

There is one more axis that is easy to miss in the competitive read: metric definitions. Weekly actives, monthly actives, app downloads, and AI-surfaced search impressions all count different things. OpenAI reporting WAU and Google reporting MAU is not an accident — each picked the scale that flatters it. Every time the press sets them side by side, the race looks either tighter or wider than it actually is. Checking the unit before reacting to the number is unusually practical advice in this market.

And OpenAI's answer to all of this is the superapp. In the March 31 post, the company said it will bring "ChatGPT, Codex, browsing, and our broader agentic capabilities into one agent-first experience," arguing that unifying surfaces lets model improvements translate directly into adoption and engagement. Read the other way, that is a declaration that the weight is shifting from adding more users toward getting more out of the users it already has.

So What Actually Changes for You

If you are a developer — nothing changes in your API pricing or rate limits this week. The directional signal is worth extracting, though. Multiple disclosures point the same way: enterprise is more than 40% of OpenAI's revenue and targeted to match consumer by year end, Codex weekly users grew 5x in three months to over 2 million, and OpenAI's APIs handle more than 15 billion tokens per minute — against Google's stated 22 billion. If you run a multi-model stack, vendor leverage on coding and agentic workloads is genuinely opening up right now, and contract renewal is the moment to use it.

If you are an investor — separate two things. First, the deceleration is real and visible in the table: from roughly 50 million net new weekly users per month down into the low 20-millions. Second, that is not automatically bad business. OpenAI stated $2 billion in monthly revenue as of March 31, disclosed 50 million consumer subscribers and 9 million paying business users, and has an advertising line posting early numbers. It is entirely plausible this is a phase where free-user growth slows while monetization accelerates. But an $852 billion valuation is priced on both engines running, so if either stalls, the repricing is not gentle. And every one of these figures is a private company's self-report, not an audited statement.

If you buy or run marketing at a company — this changes your channel planning. ChatGPT and Gemini now occupy a similar reach bracket, which raises the risk of optimizing for only one of them. That is especially true on the Google side, where AI Overviews and AI Mode are actively rewriting how search results appear and therefore how existing SEO assets get surfaced. It is also worth tracking when OpenAI's ad product graduates from pilot to general availability — $100 million annualized in six weeks suggests it will scale quickly once the gate opens.

If you just use the thing — practically speaking, nothing. Two items are worth filing away, though. One: OpenAI has publicly committed to building a unified superapp, so the ChatGPT interface is going to change meaningfully. As Snap in 2018 and GPT-5 in 2025 both demonstrated, redesigns that remove something familiar tend to generate real user anger. Two: the combination of slowing free-user growth and an arriving ads business is usually the setup for the free tier changing character. When and how, nobody knows — OpenAI has not published a plan.

🥄 Three Things You're Probably Wondering

— So did ChatGPT actually cross a billion or not? Officially, no. The last figure OpenAI has confirmed is the 900 million in its February 27 post, and the "nearing a billion" claim comes from The Information citing internal data. Until the company says it itself, "probably about to" is the accurate phrasing.

— Is being seven months late really a big deal? On the business itself, not especially. This is still among the fastest adoption curves in software history, and OpenAI says revenue is accelerating, not decelerating. But OpenAI has built its identity around being the fastest at everything, so missing a self-set target has narrative cost even when it has little operational cost. Whether the bent curve is a temporary dip or genuine saturation is too early to call from the data available.

— Does this mean Google is winning? Hard to say, because the units don't match. Gemini's 950 million is monthly and ChatGPT's billion is weekly, which is the stricter measure. What is clearly true is that Google's distribution through Search, Android and Chrome gives it a structurally cheaper path to users, and that advantage is durable. The more accurate read is not that one is beating the other, but that two companies reached comparable scale by completely different mechanisms — and those mechanisms will decide who holds it.

Further Reading

Numbers are as of announcement and may change.