AI Got Cheaper on the Surface — The Real Costs Moved Elsewhere
Two years ago the fear was lock-in via token pricing. On the surface AI looks cheaper than ever. Underneath, the costs moved into power, water, throttling, and ecosystem lock-in.
It is wild to look back now. Just two years ago, everyone warned that AI companies would lock in users with token pricing — get you hooked, then squeeze. Look at where we actually are. On the surface, AI has never been cheaper. The irony is almost funny.
Where the bill actually went
The real cost is being buried in the fine print of the physical world: water, electricity, and the heat and residue stories that stay quiet. That side of the ledger is enormous.
Pricing games and ecosystem chains
Meanwhile the product side still plays for retention. Models get tweaked, access gets throttled just enough to keep you hungry for more tokens, and features migrate behind paywalls. Openness itself can reverse — vendors that were open can close earlier weights when strategy changes.
Prices climb in some places while other regions hold steadier. Capability gaps remain real; so do strategic ones.
Talent and tokens
For a long stretch, many companies slowed junior hiring and leaned on AI tools to fill gaps. The emerging pattern looks less like “AI replaces everyone” and more like “tokens become the preferred spend instead of growing early-career headcount.” That choice has cultural and technical consequences that will show up later.
What happens next is anyone’s guess. The ride is not over. The useful habit is the same as always: watch the surface price, and separately track who pays for power, water, lock-in, and the missing juniors.