Hugging Face's Clément Delangue and investor Josh Wolfe argue rising token costs and lock-in fears drive open-source AI adoption

A huge moment for open source AI - as token usage skyrockets across orgs, so do concerns over cost & data/vendor lock-in & necessity of a multi-AI strategy

Keep an eye on @Lux_Capital portfolio cos @huggingface @togethercompute @SakanaAILabs & more! 🚀

@graceisford @Lux_Capital @huggingface @togethercompute @SakanaAILabs Have you heard about @poolsideai ? We recently launched two open source models and have processed 3 trillion tokens on OpenRouter in the last 3 months!

@graceisford @ClementDelangue @Lux_Capital @huggingface @togethercompute @SakanaAILabs Thanks for pointing to the right direction.

@graceisford @Lux_Capital @huggingface @togethercompute @SakanaAILabs The lock-in piece is probably bigger than most teams expect. Model quality shifts fast, so the painful part is rarely swapping one API call. It’s rebuilding evals and workflow logic around it. I keep seeing multi-AI become the default.

@graceisford @Lux_Capital @huggingface @togethercompute @SakanaAILabs enterprises aren't diversifying they're escaping the $70M/mo trap. open source just became the ultimate hedge."

@graceisford @ClementDelangue @Lux_Capital @huggingface @togethercompute @SakanaAILabs Costs be wild lol

@graceisford @Lux_Capital @huggingface @togethercompute @SakanaAILabs @graceisford The multi-model strategy is becoming the default — not because orgs planned it that way, but because their use cases grew faster than any single vendor roadmap. Open source gives you the optionality to route when that happens.

@graceisford @Lux_Capital @huggingface @togethercompute @SakanaAILabs Completely agree. I think this is what market maturity looks like: once token usage scales, cost discipline and avoiding vendor lock-in become core infrastructure decisions.