Point-Counterpoint: Consumers Will Never Pay for AI

Only 2.2% of US households had paid AI subscriptions as of April, according to PNC Research, leaving about 98% without one. The episode’s central question is whether that signals a lasting limit on consumer demand or simply an early stage of adoption. Adam at Repview argued that the market has room to grow because AI is still early. Ruben Hasid noted that US adoption took 19 years to rise from 2% to 50% for computers, 15 years for cars, and 6.5 years for smartphones. The co-founders contrasted those curves with household payment rates of 25% for SiriusXM, 55% for cloud storage, and 91% for at least one streaming service.

The spending figures suggest a large gap between occasional or unpaid use and intensive paid use. A16Z found that the top 1% of people already spending on AI pay an average of $903 a month, compared with a $25 median, and spend more than the bottom half combined. That statistic describes AI spenders, not all users. Among paid buyers, the top 1% were also much more likely than average to pay for particular tools: more than nine times as likely for Notion, 10 times for Runway, 15 times for Higgsfield, and 23 times for N8N. Nicholas Bustamante cautioned that use remains rare and shallow for many subscribers, making both broader adoption and deeper engagement challenges.

A16Z’s first consumer-app ranking by monthly revenue, based on Yippit data, differs from rankings by web visits and mobile users. Perplexity, for example, ranked 12th in monthly unique visits, 25th in monthly active users, and ninth in revenue. The revenue list had only 11 first-time entries, the fewest across seven lists; the host saw that as a possible sign of consolidation and network effects. The apps were concentrated in productivity, search and answers, creation, and photos and videos, with some in education and health, but none in several major categories such as social networks, travel, gaming, and dating. Companion-style apps featured more in traffic rankings than in the revenue chart, where professional-use tools were prominent.

The disagreement over personal AI agents was less about whether they can perform tasks than whether most people have enough tasks to justify using them. Basil Musharbash argued that people do not need agents to check into flights or handle email; Habibi Slopp noted that flight check-in takes only a few steps and about 55% of American adults take no flights in a year. Nick Carter pointed to more tedious possibilities, including triaging email, paying parking tickets, canceling subscriptions, and renegotiating insurance. The host rejected the idea that non-users are simply lazy or unimaginative, warning that technology enthusiasts may mistake their own workloads and ambitions for universal needs.

The host argued that consumer spending is not the only test of AI’s economics, and that AI may have less of a spending ceiling than traditional software when individuals and companies can justify thousands or tens of thousands of dollars per month through positive work returns. Consumer products can also sell entertainment rather than productivity: Suno serves musicians and other users, and the host said it had reached hundreds of millions of dollars in annual recurring revenue. But adoption at today’s most intensive level could create a separate constraint: Bustamante warned that if everyone used AI like the current top 1%, compute demand could become a shortage. Weak consumer adoption therefore matters, but does not by itself settle the industry’s financial prospects.

The subscription landscape also shows that companies can compete by converting users to more expensive plans or serving particular needs. During 2026, Claude caught up to Gemini in paid US subscribers, although both remained well behind ChatGPT; Claude converted users to higher-cost subscriptions more effectively and led ChatGPT in net-new subscriptions for much of the year before falling behind in July and August. A16Z highlighted startup strategies including differentiated models such as Suno and ElevenLabs, multi-model access such as Cursor and OpenRouter, and focus on specific audiences such as OpenEvidence and Venice. OpenRouter’s position on the revenue chart was presented as evidence that users seek cost optimization. Meanwhile, ChatGPT reached 1.2 billion weekly users, up 20% from the billion-user milestone announced at the end of July. Alex Imerman said its ads, launched in February, passed $100 million in run-rate revenue in six weeks and reached $1 billion by August, compared with $500 billion in combined ad revenue for Google and Meta; the host cautioned that consumer AI may look very different in a couple of years.