In May 2026, US data analytics firm Indagari analyzed the credit card transactions of roughly 28 million American consumers. The goal was singular: track where the people actually paying for AI subscriptions were moving. The results defied expectations. ChatGPT still held onto its overwhelming majority of paid subscribers — that much hadn't changed. But the gap wasn't widening. It was narrowing. Claude's paid-subscription transactions had risen roughly 75% since January 2026.

Around the same time, online learning platform DataCamp released data covering 20 million users. Learners voluntarily searching for and enrolling in Claude courses outpaced ChatGPT-related courses by more than 3 to 1, and over the most recent 30 days, demand for Claude courses was up 18-fold. In DataCamp's site-wide search rankings, "Claude" became the single most-searched term — outranking even the generic query "AI."

Both companies are heading toward public offerings. OpenAI is preparing to go public, and on June 1, 2026, Anthropic filed its own IPO application. Paid-subscriber trends are no longer just a curiosity about user taste — they're now read as a growth metric investors are watching closely. That's exactly why Indagari's analysis landed at this particular moment.

Why Paying Users Move Faster

There's a behavioral difference between people who use AI for free and people who pay for it. Free users tend to stick with the name they already know. With no switching cost, there's little incentive to actively hunt for alternatives. Paying subscribers are different: the moment they commit to that first monthly charge, they set a bar. An expectation forms that the service needs to earn back that cost. When that expectation goes unmet, paying users switch faster than free users do.

In this context, it's hard to overlook the event that directly triggered Claude's subscriber surge in early 2026. Shortly after Anthropic publicly refused requests to support the Trump administration's mass-surveillance program and its use of autonomous weapons, reports emerged of a noticeable jump in paid subscribers. This was a case where the deciding factor in a subscription choice wasn't technical performance but the company's public stance. Among consumers, a mood took hold in which which service you pay for each month started to read as a signal of which direction you support.

A similar conversation reignited in mid-June 2026, when the US government imposed restrictions limiting a specific AI model's service to non-US users. Once again, it became clear that an AI provider's sustainability and its policy responses have a real effect on subscription decisions.

What 75% Growth Shows — and What It Doesn't

Here, the skeptical view deserves honest consideration. According to Sensor Tower's app analytics data, ChatGPT still holds far more paid users than Claude across every platform. That "75% growth" figure can be an optical illusion that looks dramatic simply because it started from a low base. Going from 100 to 175 and going from 1 million to 1.75 million are both 75% growth, but they mean very different things at scale. In absolute terms, the distance between the two services remains substantial.

If the trigger for the subscription surge was a political event rather than technical differentiation, how durable that growth proves to be is also an open question. There's no data yet on whether subscriptions born from an ideological reaction convert into long-term retention. DataCamp's 18x figure also doesn't specify its baseline — if the starting point was low, the number could create a stronger impression than the underlying reality warrants.

Even with those caveats on the table, there's a reason this trend is hard to dismiss as noise: three independent data sources — credit card transaction analysis, education-platform search data, and course-enrollment patterns — are all pointing in the same direction. When independent measurements align this consistently, it's difficult to explain away as a single measurement error.

There's another factor behind the growth. Claude tends to score well with users on long-context document work — handling lengthy contracts or reports, or refining output across extended back-and-forth conversations. DataCamp's demand spike reads as this specific usage pattern converting directly into learning demand.

One Question to Ask Before Your Next Monthly Charge

What Korean solo founders and planners can practically take from this data is that AI subscriptions have already claimed a concrete line item in the cost structure.

Financial thinking starts with asking where and how a spent resource comes back to you. Treat a monthly AI subscription as just another software line item, and it becomes hard to tell whether it's actually cutting your work time or improving the quality of your output. Plenty of subscriptions get charged without being used, or get kept out of inertia without anyone checking whether a better option exists. To call something an investment, you first have to confirm what it's returning, and how much.

Real performance differences show up depending on the type of work. Some services are stronger at maintaining long document context and running analysis; others excel at real-time web search integration and a broad ecosystem of external tools. Reviewing contracts, drafting reports, and structuring ideas call for strong long-context handling — a different skill set from tasks that need up-to-the-minute market information or integration with outside services. Once you identify which category your current workload mostly falls into, you can check whether your current subscription choices are actually allocated to match it.

One more thing the DataCamp numbers suggest: mastering the service you already subscribe to can deliver faster real gains than adding a new subscription. Claude course demand outpacing ChatGPT's by more than 3 to 1 signals rising demand to actually learn how to use a specific service well. Before adding another subscription, raising how fully you use the one you already have may be the better use of resources.

Two facts emerge from this analysis of 28 million credit card transactions: ChatGPT still leads the market, and a portion of paying users have started making a different choice. What stands out to me in this data isn't the competitive dynamic so much as a single question: is the AI subscription you're paying for every month actually earning its cost in your work — and is that judgment resting on verification, rather than habit?