The ugly economics of consumer AI
TechCrunch Russell Brandom ● Covered by 31 sources
Consumer AI is getting hot again, but the money still looks thin. More people will pay, but the bill and the costs still don’t match up.
Based on reporting by TechCrunch, Russell Brandom — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Consumer AI is having a very good week, at least on the surface. Meta’s personal assistant Muse, with its plush mascot Jolly, is getting attention. OpenAI’s new Dots app is leaning into the same cartoonish helper idea. And Instinct has already hit a $10 billion valuation by pitching itself as an agent that books trips, makes restaurant reservations, and cancels subscriptions.
That makes the bull case feel obvious. The products are getting good enough to do real errands, and some people are clearly willing to pay for that convenience. To investors, it looks a lot like the early ChatGPT moment in 2022, when raw model capability suddenly made a new product category feel possible. The trouble is that the consumer side has been flirting with this story for a while, and the economics keep refusing to cooperate.
Andreessen Horowitz’s latest State of Markets report, based on PNC research, puts some numbers on the problem. As of May, 2.2% of consumers were paying for AI, at an average of $31 a month. Andreessen tries to read that as early adoption with lots of room to grow. But the lines on the charts are still painfully straight. Better models haven’t translated into a big jump in either the share of paying users or what they spend.
The gap gets ugly fast when you compare it with the kind of revenue needed to support frontier AI. The report points to Netflix as a saturated consumer subscription benchmark: at 325 million subscribers, $34 a customer only gets you to $11 billion a year, which is still less than a third of OpenAI’s operating costs. Bank of America’s estimate from March, that roughly 3% of U.S. consumers pay for AI, and a Menlo survey from September, which found that a quarter of adults use AI daily and half of those users pay for it, point in the same direction. There are users. There is usage. There just isn’t much pricing power.
So the industry has been moving toward the safer money: enterprise contracts and vertical-specific products. OpenAI’s push into business sales has reportedly been paying off, with enterprise bookings said to have doubled since July. Dots even has an enterprise story, aimed at software engineers and agency creatives. Meta has the extra cushion of ad targeting, and Instinct says it can take a cut of purchases made through its agent. But the core lesson hasn’t changed. Consumer AI may be popular, even beloved. It still has to answer to brutal unit economics.
My take — AI-written commentary, not fact-checked reporting
This is the part of AI that keeps getting dressed up as a consumer revolution and then quietly walks back through the enterprise door. Fancy mascots and agentic errands make for a nicer demo than procurement software, but they don’t change the math. The industry keeps learning the same lesson because the industry keeps hoping this time the bill will be smaller.
Read more about this at: TechCrunch