Free newsletter: The economics of AI don't make sense. Monthly subscriptions to AI services are scams that represent a product that is impossible for OpenAI and Anthropic to provide long term, and AI data centers are an economic disaster at scale.
wheresyoured.at/ais-economics-…
Github Copilot, as I reported last week, is moving towards token-based billing, charging the actual cost of token burn — a move that alienated many of its users who built entire workflows around a request-based product that no longer exists.
wheresyoured.at/ais-economics-…
This is the subprime AI crisis — and a crisis of deception. AI products tricked users by offering unsustainable monthly subscriptions built to separate the process of using AI from its real cost. Eventually everybody will go token-based.
wheresyoured.at/ais-economics-…
AI subscriptions are an intentional scam. When you're not paying the actual cost of using AI, you're willing to forgive AI's hallucinations and mistakes. The majority of LLM users have built workflows around all-you-can-eat cost structures.
wheresyoured.at/ais-economics-…
The AI industry tried to use a classic startup growth trick — subsidize, scale, and when you reach a critical mass, crank up the prices. Except the services are too expensive, and the costs too far removed from the cost of subscription.
wheresyoured.at/ais-economics-…
Subscription-based AI subsidies were essential to the generative AI growth story. It’s easy to write puff pieces about Claude Code when you’re paying $20 a month rather than, at minimum, $13 or $30 a day (based on Anthropic's own estimates).
wheresyoured.at/ais-economics-…
We'll find out exactly how "useful" AI is once companies and users are forced to pay for their actual token burn. When AI costs 10% of your headcount, and may reach 100%, flimsy claims of boosted productivity spouted by boosters and CTOs go quiet.
wheresyoured.at/ais-economics-…
The economics of AI data centers also don't make sense.
A 100MW data center costs over $4.5bn with $3bn+ in GPUs, with costs eating up 85% of your revenue, and the rest destroyed by debt. Oh, and the chips are obsolete in a few years.
wheresyoured.at/ais-economics-…
Things get worse with large data centers like Oracle's Stargate Abilene - a $52.8bn, 8-building campus that's years behind schedule, has billions a year in costs, and relies on a single tenant - OpenAI - who just missed its revenue targets.
wheresyoured.at/ais-economics-…
If OpenAI doesn't make $852 billion by the end of 2030, the Stargate Data Center project will destroy Oracle, as it needs to make $75 billion in annual revenue for its 7.1GW in data centers to keep up with the debt and its negative cash flow.
wheresyoured.at/ais-economics-…
Based on my analysis, for the 15.2GW of in-construction data centers due by the end of 2028 to make sense, they must generate over $156 billion in annual revenue. Right now, I'm struggling to find $2 billion a year outside of OpenAI and Anthropic.
wheresyoured.at/ais-economics-…
AI is a con, one rushed to market using unsustainable subscription products to convince people of its benefits while actively obfuscating the real costs that customers would eventually pay. That and rushing to make it seem "inevitable."
wheresyoured.at/ais-economics-…
OpenAI must make 10x its current revenue and raise at least $200 billion by 2030 or it will be unable to pay its bills, which will kill Oracle, which will in turn destroy Larry Ellison's wealth.
The warning signs have been there for years.
wheresyoured.at/ais-economics-…
Share this Scrolly Tale with your friends.
A Scrolly Tale is a new way to read Twitter threads with a more visually immersive experience.
Discover more beautiful Scrolly Tales like this.