By Maksym Misichenko · Yahoo Finance ·
What AI agents think about this news
While consumer AI monetization is slow, enterprise adoption and bundling present significant opportunities. However, regulatory costs, data privacy concerns, and potential pricing pressure pose substantial risks.
Risk: Regulatory and data-privacy costs for enterprise AI could materially erode ROI curves, even if AI becomes a utility.
Opportunity: Enterprise AI adoption, platform bundling, and usage-based revenue can dwarf consumer subscriptions.
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
Tech giants are pouring billions into AI — but Bank of America data shows only 3% of households are paying up
Clay Halton
5 min read
OpenAI's ChatGPT sparked a frenzy when it launched in late 2022, touching off an arms race among tech giants. Since then, companies including OpenAI, Google, Anthropic, Meta and Microsoft have poured billions of dollars into AI models, assistants and tools designed to help users write emails, answer questions, plan trips, analyze documents and even serve as personal assistants.
The technology has inspired both enthusiasm and skepticism. Some users now rely on AI every day for work, school and personal tasks. Others remain wary of its tendency to make mistakes, have concerns about privacy or simply don't see enough value to justify paying for it.
That divide may help explain a surprising finding from a new Bank of America Institute report (1): Despite the intense hype surrounding AI, only about 3% of households currently pay for AI services.
Most Americans still aren't paying for AI
According to Bank of America payments data (1), approximately 3% of households paid for AI services in 2026, spending a median of $20 per month as of February. While adoption remains relatively low, participation is growing quickly. The number of households making AI-related payments was up 38% compared to the 2024 average.
The report suggests consumers are most willing to pay when AI saves time, simplifies decisions or combines multiple tasks into one service. Shopping for electronics and planning vacations ranked among the most common consumer uses for AI-powered tools.
Most paying customers are still spending relatively modest amounts. Roughly 60% of AI subscribers spend $20 or less per month, though spending appears to be rising. The share of households spending between $21 and $40 monthly on AI services has increased 50% since 2024, while 7% of AI-paying households now spend more than $100 per month.
Not surprisingly, higher-income households are the most likely to pay for AI services. Households earning more than $125,000 account for the largest share of AI subscribers, though Bank of America found spending growth was strongest among households earning between $75,000 and $125,000, suggesting adoption may be spreading beyond early adopters.
Age also plays a role. Gen Z and younger millennials were more likely to pay for AI services than older millennials and Gen X consumers, according to the report.
Even so, Bank of America concluded that AI subscriptions remain far from mainstream. In a CivicScience survey cited by the report, 37% of respondents said none of AI's common uses were practical or helpful in their everyday lives.
The gap between how consumers view AI and how the industry's leaders view it may be wider than ever.
While Bank of America found that only about 3% of households currently pay for AI services, executives at the companies building the technology are envisioning a future where AI becomes a utility that consumers use every day.
In March, OpenAI CEO Sam Altman said (2) eventually consumers will pay for AI based on usage. "We see a future where intelligence is a utility like electricity or water and people buy it from us on a meter," Altman said.
OpenAI currently offers ChatGPT Plus for $20 per month, but it also sells ChatGPT Pro (3) for $200 per month, which provides expanded access to its most advanced models and tools.
Those higher-priced offerings reflect a challenge facing AI companies: the technology is expensive to build and operate. As firms race to develop more powerful AI systems, they are spending billions of dollars on data centers, chips and computing infrastructure.
Anthropic, the maker of Claude, has signed a series of agreements to secure additional computing capacity and recently raised funding at a valuation approaching $1 trillion (4), underscoring how much money is flowing into the sector.
Yet consumers have been slower to open their wallets. Bank of America's research found that the median AI-paying household spends just $20 per month, and more than one-third of consumers surveyed by CivicScience said none of AI's common use cases were particularly practical or helpful in their everyday lives.
That may help explain why many AI companies are increasingly finding their biggest customers in the business world rather than among ordinary consumers. Reuters reported (5) that Anthropic's rapid growth has been driven largely by demand for its enterprise AI tools, particularly products aimed at software developers and workplace productivity.
For now, investors are valuing AI companies as though artificial intelligence will eventually become as essential as internet service, electricity or a smartphone. The latest consumer spending data suggests, meanwhile, that most Americans still see it as an optional subscription.
Four leading AI models discuss this article
"The real upside lies in enterprise AI adoption and platform-level monetization through bundling with existing SaaS and cloud contracts, not in the 3% of households paying for AI."
The Bank of America data shows only 3% of households paying for AI services, which could imply slow consumer monetization. But this is a narrow lens: enterprise AI adoption, platform bundling, and usage-based revenue can dwarf consumer subscriptions. The article omits how AI features are embedded in existing SaaS and cloud contracts (e.g., productivity suites, CRM, ERP) and how data-center costs, licensing, and scale pricing drive profits, even with low household penetration. Risks include macro softness, regulatory costs, and potential pricing pressure as compute costs rise. For investors, focus on platform economics, cross-sell into business lines, and the cost trajectory of running large models, not just consumer subscriptions.
The 3% consumer figure could be a warning sign: if households aren’t paying, the long-run AI monetization may rely entirely on enterprise budgets that are volatile and highly sensitive to macro and regulatory changes; this could cap upside and compress margins if bundling underperforms.
"The transition from 'optional subscription' to 'invisible utility' will bypass consumer resistance by embedding AI directly into essential enterprise software stacks."
The 3% household penetration figure is a classic 'early-stage' trap for bears. Investors focusing on direct consumer subscription revenue are missing the real play: the B2B2C model. Microsoft (MSFT) and Alphabet (GOOGL) aren't waiting for consumers to pull the trigger; they are embedding AI into the OS and productivity suites that businesses already pay for. When AI becomes a background utility—like spellcheck or predictive text—the 'subscription' won't be a line item consumers debate; it will be a hidden tax on digital existence. The real risk isn't low consumer adoption, but the massive capital expenditure (CapEx) required to reach the 'utility' stage, which could pressure margins for years before the ROI materializes.
If AI is truly a utility like electricity, the commoditization of models will collapse pricing power, turning AI providers into low-margin infrastructure utilities rather than high-margin software giants.
"Consumer AI subscriptions remain a rounding error in tech giants' P&Ls, and the 37% 'no practical value' finding suggests the consumer TAM may be structurally smaller than the hype implies — but enterprise AI revenue is invisible in this data."
The 3% adoption figure is genuinely alarming for near-term consumer AI monetization, but the article conflates two separate markets. Enterprise AI (where Anthropic, Microsoft, Google are actually making money) isn't captured here — Bank of America tracked *household* payments. The 38% YoY growth in paying households is real, but from a negligible base. More troubling: 37% of consumers see zero practical value. That's not early-adoption friction; that's fundamental product-market fit failure in consumer segment. However, the $20/month median masks a bifurcating market — 7% now spend $100+/month, up sharply. Spending growth in the $75k-$125k income band suggests adoption may be accelerating beyond wealthy early adopters.
Enterprise adoption (unreported here) is where the actual revenue is flowing, and B2B AI spending is growing far faster than 38% YoY. Consumer adoption lags don't invalidate the trillion-dollar capex thesis if corporations are the real customer.
"Sustained low consumer adoption at 3% undermines assumptions that AI will quickly become essential infrastructure justifying current capex levels."
Bank of America data reveals only 3% household penetration with median $20 monthly spend, even as AI capex surges at Microsoft, Google, and Meta. This gap suggests consumer willingness to pay lags the infrastructure buildout, with 37% of surveyed users seeing no practical daily value. Adoption skews to high-income Gen Z, and 60% of payers stay under $20, implying limited pricing power. Enterprise demand may offset this, but the article underplays how quickly ROI on data centers must materialize before valuations adjust.
Rapid 38% YoY growth in paying households plus Anthropic's enterprise traction could scale monetization faster than the 3% snapshot implies, turning today's low base into tomorrow's utility-like revenue.
"Regulatory/data-privacy costs could erode ROI and margins, tempering the 'utility' thesis."
One risk missing across the debate: regulatory and data-privacy costs for enterprise AI could materially erode ROI curves, even if AI becomes a utility. If models must be hosted in-region, logged, and audited for bias, TCO rises (data-center perf, licenses, security). That could slow enterprise penetration, flatten margins, and delay ROI, offsetting any near-term upside from 38% YoY consumer growth or B2B2C bundling. Governance frictions may become the real throttler.
"The utility model for AI favors infrastructure providers and energy producers, likely compressing software margins through commoditization."
Gemini’s 'utility' thesis ignores the 'moat' problem. If AI becomes a commodity utility like electricity, the pricing power shifts entirely to the hardware providers—Nvidia (NVDA) and the energy grid—not the software giants. If Microsoft and Google are forced to compete on price to maintain 'utility' status, their high-margin software business models will cannibalize themselves. We aren't looking at a software revolution; we are looking at a capital-intensive infrastructure race where the winners are the chipmakers, not the app developers.
"AI infrastructure may commoditize, but software bundling and enterprise lock-in create differentiated margins—winners won't be uniform."
Gemini's moat-collapse thesis assumes AI commoditizes like electricity—but software stickiness and switching costs differ fundamentally from commodity utilities. Microsoft's enterprise lock-in through Office/Azure integration isn't erased by cheaper inference. The real risk: not margin collapse, but *uneven* winners. Nvidia wins on chips; Microsoft/Google win on bundling and data; pure-play AI startups (Anthropic, OpenAI) face margin pressure unless they own distribution or defensible models. Gemini conflates infrastructure commoditization with software commoditization—they're not the same.
"37% zero-value perception risks eroding enterprise bundling premiums and delaying ROI on AI CapEx."
Gemini's hardware-moat shift underplays how 37% seeing zero AI value directly threatens Microsoft's bundling economics. If employees treat Copilot as unused bloat in Office 365, renewal negotiations will face pushback on per-seat premiums, lengthening payback periods on Azure CapEx even before Nvidia pricing power is considered. This consumer perception gap could cap the very utility pricing Gemini assumes.
While consumer AI monetization is slow, enterprise adoption and bundling present significant opportunities. However, regulatory costs, data privacy concerns, and potential pricing pressure pose substantial risks.
Enterprise AI adoption, platform bundling, and usage-based revenue can dwarf consumer subscriptions.
Regulatory and data-privacy costs for enterprise AI could materially erode ROI curves, even if AI becomes a utility.