71% of the world has never used AI, only 1% pays for a subscription
71% of the world has never used generative AI, only 1% pays for a subscription, and 0.14% codes with it. The business still relies on investors, not customers.
English · Original discussion in Spanish · Published
Only 1% of humanity pays for artificial intelligence
How many people actually use artificial intelligence? Far fewer than marketing claims. 71% of the world's population has never tried a generative AI. The remaining 28% have only peine a free chat at some point. Barely 1% pay for a subscription. And between 0.14% and 0.36% actually use it to write code. In the land of the blind, the one-eyed man is king, and that one-eyed man also sends you the bill.
A business sustained by investors' money, not customers'
The calculations circulating point to a fundamental problem: the operating cost of frontier models would be 10 to 20 times higher than the revenue they generate. A 20-euro monthly subscription wouldn't even cover the electricity bill for the data centers that support it. What keeps the whole thing standing isn't the customer, it's the investment round. Some describe it outright as a structure that only pays off with new capital—the Ponzi scheme manual; others argue that lightweight, open models already handle basic tasks using just a CPU and 32 GB of RAM, at a negligible cost.
The gap between those who have tried it and those who truly use it
The 28% who have peine a free chat is not the same as the 1% who pay or the 0.14% who code. Most of those who have touched it use it as a more convenient search engine. The gap separates those who experiment from those who have integrated AI into their workflow: people without technical training who have already built applications, websites, and SaaS products without writing a line, and professionals who pay 200 euros a month because it saves them 2,000 in work. That's where the real business is, not in the cloud.
What AI still can't do
Enthusiasm clashes with technical limits. It solves scripts and specific problems, but gets stuck on large applications with hundreds of tables and interconnected flows: the context window, the space it works with, can't handle that much. And with every error it generates, a human appears reviewing code they didn't write, which costs more than the savings. The other trap is expectations: an operator has 50 trading operations dictated by an AI and nearly 30 at a loss. The model writes confidently, it doesn't predict the market. Confusing one with the other is what the sector sells.
The data point that puts everything in perspective
Two billion people live without electricity. Talking about global AI adoption when 71% of the planet hasn't touched it is selling the movie before filming it. The leap looks more like the internet of 2000—few users, unintuitive technology, everything yet to come—than the definitive disruption promised by consulting reports. Mass job replacement is postponed, not ruled out. The one-eyed man reigns; and like any kingdom built on the blind, it depends on no one opening their eyes.
How many people actually use artificial intelligence? Far fewer than marketing claims. 71% of the world's population has never tried a generative AI. The remaining 28% have only peine a free chat at some point. Barely 1% pay for a subscription. And between 0.14% and 0.36% actually use it to write code. In the land of the blind, the one-eyed man is king, and that one-eyed man also sends you the bill.
A business sustained by investors' money, not customers'
The calculations circulating point to a fundamental problem: the operating cost of frontier models would be 10 to 20 times higher than the revenue they generate. A 20-euro monthly subscription wouldn't even cover the electricity bill for the data centers that support it. What keeps the whole thing standing isn't the customer, it's the investment round. Some describe it outright as a structure that only pays off with new capital—the Ponzi scheme manual; others argue that lightweight, open models already handle basic tasks using just a CPU and 32 GB of RAM, at a negligible cost.
The gap between those who have tried it and those who truly use it
The 28% who have peine a free chat is not the same as the 1% who pay or the 0.14% who code. Most of those who have touched it use it as a more convenient search engine. The gap separates those who experiment from those who have integrated AI into their workflow: people without technical training who have already built applications, websites, and SaaS products without writing a line, and professionals who pay 200 euros a month because it saves them 2,000 in work. That's where the real business is, not in the cloud.
What AI still can't do
Enthusiasm clashes with technical limits. It solves scripts and specific problems, but gets stuck on large applications with hundreds of tables and interconnected flows: the context window, the space it works with, can't handle that much. And with every error it generates, a human appears reviewing code they didn't write, which costs more than the savings. The other trap is expectations: an operator has 50 trading operations dictated by an AI and nearly 30 at a loss. The model writes confidently, it doesn't predict the market. Confusing one with the other is what the sector sells.
The data point that puts everything in perspective
Two billion people live without electricity. Talking about global AI adoption when 71% of the planet hasn't touched it is selling the movie before filming it. The leap looks more like the internet of 2000—few users, unintuitive technology, everything yet to come—than the definitive disruption promised by consulting reports. Mass job replacement is postponed, not ruled out. The one-eyed man reigns; and like any kingdom built on the blind, it depends on no one opening their eyes.
Summary of a discussion on Burbuja.info - Foro de economía, actualidad y política., translated from Spanish and reviewed before publication. Read the full discussion (98 replies).
More summaries
-
OpenAI: The Lehman Brothers of AI? Ed Zitron's thesis
OpenAI as the Lehman Brothers of AI according to Ed Zitron: the bubble burns billions, but balance sheets and orders through 2031 say otherwise
-
Two Civil Guards Shot During Surveillance in Terrassa, Spain
Two Civil Guards were shot in Terrassa during a surveillance operation; one suffered four bullet wounds and was beaten. The incident highlights security concerns in Spain.
-
Seville to charge non-residents for Plaza de España entry
Seville will charge tourists for Plaza de España access, splitting revenue with the Spanish state: 75% for the city and 25% for national treasury.
-
Fernando Delgado, Spanish journalist and writer, dies at 77
Fernando Delgado, a prominent Spanish radio voice and author, has died at age 77 in Faura, Valencia. A key figure of the democratic transition era.
-
Viral Madrid Metro Fight Video Was Nearly a Year Old
A fight recorded nearly a year ago circulated as breaking news in the Madrid Metro, sparking debate over its authenticity.
-
Red Cross volunteer: four days away, husband overwhelmed
A Spanish civil servant has been a Red Cross volunteer for four days, helping relocate migrants from the Canary Islands. Her partner asks for more help.