---
title: "The Substitution Wave in AI"
description: "Frontier model prices keep rising while open-source crosses the good enough line. Coinbase, Lindy, Harvey \u0026 Cursor are substituting — \u0026 the savings go straight back into more tokens."
categories: ["AI","startups"]
keywords: ["AI cost structure","open source AI models","frontier model pricing","Kimi K2.5","Cursor Composer","model substitution","AI buyers","Anthropic alternatives","DeepSeek v4","Jevons paradox AI"]
ai_summary: "AI buyers across enterprise, app, and seller layers are substituting cheaper open-source models for frontier closed models. The savings don't shrink the AI bill — they get reinvested in exponentially more tokens per task."
date: 2026-06-07
lastmod: 2026-07-23
canonical_url: https://www.tomtunguz.com/inflation-deflation-ai/
author: "Tomasz Tunguz"
---


Three forces are reshaping the AI cost structure :

1. Foundation labs are moving up the stack into applications,[^2] [^3]
2. Frontier model prices keep rising for the smartest models,[^1]
3. Open-source models have crossed the good enough threshold for most use cases.[^4] [^5]

The natural response from AI buyers is substitution.

Coinbase[^6] :

> At Coinbase we're working hot on routing prompts to cheaper models where appropriate, & in some cases have been able to keep costs roughly flat, while token usage continues to grow exponentially.

Lindy[^7] :

> Pulled the trigger today & switched 100% of Lindy traffic to DeepSeek v4, churning from Anthropic models. Saves us millions of $ & we're actually seeing an _increase_ in performance on many core use cases. Transformative for the business.

Harvey[^8] :

> On a 100-task slice of our Legal Agent Benchmark (LAB), SFT moved Kimi 2.6's all-pass rate from 11% to 15%, beating Opus' 14%. But the cost gap was even more striking : $84 vs $954 across the same 100 tasks, or ~11x cheaper.

Cursor went further. They post-trained Kimi K2.5 into their own production model, Composer.[^9]

> Composer 2.5 is exceptionally intelligent & up to 10x more efficient than similarly capable models.

Coinbase's quote shows where the savings go : costs flat, tokens exponential. Buyers don't pocket the discount — they spend it on more intelligence.

Closed models are getting more expensive at the frontier; open models are getting cheaper at parity. The choice is which slope you want under your unit economics.

{{< email_image src="zoiothfpwaee5fqfrjq5" alt="Ramp cost curve framing for AI buyers and app purveyors" width="540" height="412" >}}

[^1]: https://tomtunguz.com/ai-model-inflation/

[^2]: https://x.com/Law360/status/2062263047578673481

[^3]: https://theoryvc.com/blog-posts/are-foundation-models-and-application-companies-friends-or-foes

[^4]: https://tomtunguz.com/using-local-ai-to-work-faster/

[^5]: https://tomtunguz.com/the-thriving-ecosystem-of-open-models/

[^6]: https://x.com/brian_armstrong

[^7]: https://x.com/Altimor/status/2062389885437366342

[^8]: https://x.com/harvey/status/2062218656420167785

[^9]: https://x.com/cursor_ai/status/2056415414977187904