AI's Affordability Crisis
Points and comments are a snapshot, not live.
AI platforms are raising prices as subsidies end, causing an affordability crisis.
The article argues that AI platforms subsidized usage to generate demand, but now face a revenue gap. OpenAI's 2025 losses hit $38.5 billion on $13.07 billion revenue. Anthropic and Microsoft shifted to token-based billing, spiking costs 7x for some users. Nvidia's VP noted AI compute costs exceed employee wages. A Financial Times analysis shows negative implied returns on AI investment for Microsoft (-9.2%), Alphabet (-15.7%), Meta (-28.8%), and Oracle (-35.6%), assuming zero costs.
To service an estimated $3 trillion in debt, the AI industry would need to displace 27-32.5 million US jobs. OpenAI considers "drastic" price cuts amid a price war with Anthropic, while Microsoft is moving internal teams from Claude Code to Copilot CLI to control costs.
What commenters are saying
Commenters debate whether AI costs are falling or rising. One camp cites API price drops of 50x over three years, suggesting efficiency gains will solve affordability. Another counters that hardware refresh cycles are accelerating, making both training and inference more expensive, and that price cuts don't help if companies need to recoup massive capital investments. A top comment notes that the US government may ban foreign models, creating a captive market for US labs, though another argues that constitutional free speech protections cover model weights and research papers. A lower-ranked comment dismisses the idea that frontier labs have lost the war for coding as an HN bubble view.