SambaNova builds AI inference chips and systems for enterprises running large language models on-premises or in the cloud.
$1BSeries F
A $1B Series F for inference chips signals that enterprises are actually moving LLMs on-prem—not just talking about it—and that the market believes there's real margin in custom silicon vs. renting GPU capacity. SambaNova's getting this kind of capital because inference is becoming the recurring revenue play (models are trained once, inference runs forever), which means if you're building enterprise AI tooling, you should assume your customers will eventually want to run models locally for cost/latency/compliance reasons.