Stop Renting Your Cognitive Infrastructure - Thiyagarajan Maruthavanan, Kalmantic Labs
Enterprises are abandoning rented AI inference after runaway costs force infrastructure ownership.
“It's almost as if you're loading credits inside a casino. You put some and then you pull it and then you are so addicted to it then you end up doing more and more of it and by some time you realize that you've blown past the threshold that you had mentally kept in mind.”
A founder at Kalmantic Labs argues that rented AI inference from providers like Anthropic creates unpredictable, casino-like spending dynamics — citing a major retailer that spent $200M before building its own infrastructure and Uber burning through its annual token budget by month four. The talk highlights structural cost risks: agent loops multiplying inference calls, poor context compression, and API key theft exposing companies to uncapped liability. The core signal is a growing enterprise backlash against pay-per-token models driving a 'bring inference in-house' movement.