As of May, only 2.2% of consumers were paying for AI services, spending an average of $31 monthly, according to Andreessen Horowitz's semiannual State of Markets report released this week, which drew figures from a PNC research report published this summer. The analysis reveals that despite enormous improvements in AI model performance, neither the share of paying customers nor their average spending has grown substantially. The report warns that consumer AI faces fundamental economic challenges that won't disappear as the technology gets better, forcing even popular products to seek enterprise revenue or face unsustainable losses.
The data shows growth in both metrics has been surprisingly linear, with massive performance leaps barely registering on adoption charts. The jump from GPT-5.2 to Astra, for instance, produced almost no visible change in consumer willingness to pay for AI services, the report notes. Using Netflix's 325 million subscribers as a benchmark for saturated online services, revenue of $34 per customer would generate only $11 billion annually—less than one-third of OpenAI's operating costs. Bank of America found roughly 3% of U.S. consumers paid for AI in March, up 40% from the prior year, while a September survey from Menlo reported that a quarter of adults use AI daily, with half of those users paying for it.
Andreessen Horowitz frames the low adoption as an opportunity, stating "it's still so early when it comes to mature AI adoption and utilization." The report emphasizes that the core problem isn't revenue but cost, noting AI is exceptionally expensive to operate compared to lightweight predecessors like social networking or cloud computing. Even hundreds of millions of paying customers doesn't guarantee breaking even, the analysis concludes.
The report explains this dynamic by pointing to AI's operational costs, which dwarf those of earlier internet services. While recent releases like Meta's Muse assistant and OpenAI's Dots suggest consumer AI may be staging a comeback, the underlying economics haven't improved, forcing companies to eventually confront monetization realities regardless of technical capability. OpenAI has adapted by pivoting toward enterprise contracts, with business bookings reportedly doubling since July, while even its Dots launch highlighted uses for software engineers and agency creatives. The report notes this mirrors a broader industry shift toward the Anthropic model of focusing on enterprise deals and vertical-by-vertical expansion, since selling popular-but-cheap consumer services to businesses at a markup remains one of the few proven paths to profitability.
Products like Muse and the $10 billion-valued Instinct assistant can delay these questions—Meta through its advertising infrastructure and enterprise exploration, Instinct through taking a cut of purchases and avoiding frontier model training costs—but the hard cap on consumer revenue growth remains, the report concludes. It's one of the few certainties in an industry otherwise defined by rapid change. The shift to enterprise isn't a temporary detour but a structural response to economics that better models alone won't fix. Companies betting on consumer enthusiasm will eventually need to explain how they'll overcome cost structures that have already redirected the frontier labs, or accept that mass adoption and profitability may remain incompatible goals in this technology category.