App services equipped with artificial intelligence (AI) technology are struggling to retain users after initial market expansion, fueling debate over the AI industry's ability to generate practical profitability. As astronomical capital expenditures continue, user churn is appearing faster than for general apps, raising questions about sustainable profit models for AI businesses.
According to the "2026 subscription app ecosystem report" by subscription management platform RevenueCat on the 12th, the annual retention rate of AI-based apps over the past year was 21.1%, which was found to be 9.6 percentage points lower than non-AI apps (30.7%). Monthly retention was also lower for AI apps at 6.1% compared with 9.5% for non-AI apps.
In particular, paying subscribers are churning faster on AI apps. RevenueCat analyzed that the annual cancellation speed for AI app subscriptions is about 30% faster than for non-AI apps. Refund rates were also about 20% higher for AI apps, and some services showed refund rates as high as 15.6%.
RevenueCat said, "AI apps show a high conversion rate from free trials to paid subscriptions but are highly volatile in terms of long-term user value and experience," adding, "Revenue volatility is high, and they may face difficulties maintaining long-term service quality."
Although AI apps are growing rapidly, securing actual profitability is not easy, analysts say. Running Generative AI models requires hyperscale data centers and high-performance graphics processing units (GPUs), pushing corporations to invest in infrastructure worth tens of billions of dollars.
Concerns are also being raised about the financial soundness of large AI corporations. Credit rater Moody's warned in a recent report that major IT corporations may be managing large lease expense for data centers off their balance sheets. By using special purpose vehicles (SPVs) or not disclosing lease renewal expense, the actual financial burden may not be fully reflected in financial statements, the analysis said. Moody's noted, "We may reassess tech corporations' credit ratings to reflect these potential liability," adding, "If hidden debt materializes as AI infrastructure investment continues to expand, financial risk could increase."
Debate also continues over valuations of AI corporations. OpenAI, the developer of ChatGPT, is preparing for an initial public offering (IPO), with a valuation discussed at about $850 billion (about 1,253 trillion won). That is about 28 times this year's expected sales, which some view as rich compared with Nvidia, which trades at about 12 times sales. Short seller Jim Chanos said, "Nvidia has a dominant position, high margins and stable cash flow, but OpenAI must shoulder massive data center investment expense," adding, "It's unclear why it should command a higher valuation."
Academia is also issuing warnings about the AI investment boom. Joseph Stiglitz, a Nobel economics laureate and professor at Columbia University, said in a recent interview that "about one-third of U.S. economic growth last year was sustained by AI investment," labeling the current situation an "AI bubble." Stiglitz said, "Even if the technology succeeds, intensified competition between global big tech and Chinese corporations could make it hard to reap the expected profits," adding, "If the AI investment bubble bursts, a short-term shock could ripple across the macroeconomy."
Still, there is pushback that the AI industry's growth potential remains strong. Nvidia, which leads the AI Semiconductor market, posted record results with quarterly revenue of $68.13 billion, up 73% from a year earlier. In a survey Nvidia conducted of corporations worldwide, 88% of respondents said AI adoption contributed to sales growth, and 86% said they plan to increase AI-related budgets. Jensen Huang, Nvidia's chief executive officer (CEO), said, "Computing demand is growing exponentially, and the Agentic AI inflection point has arrived," adding, "Demand for AI infrastructure will continue to expand."