SambaNova secures $1B Series F to scale AI inference chips

SambaNova’s $1 billion Series F is one of the largest private financings for an AI hardware player this year and sends a clear signal that the bottleneck has shifted from model training to large-scale inference. By positioning its RDU-based systems as “premium inference” for multi-trillion-parameter models, SambaNova is staking out the space right between hyperscaler GPUs and smaller edge accelerators. The JPMorganChase win matters because it validates the thesis that top banks will build sovereign, on-prem AI stacks instead of relying exclusively on cloud APIs.

Strategically, this round shows that capital is still willing to underwrite alternative compute architectures if they promise better cost-per-token at frontier scale. It also deepens Intel’s ecosystem play around heterogeneous inference, which could loosen Nvidia’s grip on the high-end stack over time. For the race to AGI, the key implication is that specialized inference hardware for agentic workloads is now a funded category, not just a slide in conference decks. If SambaNova can reliably run cutting-edge models faster and cheaper, it will make it easier for enterprises and governments to put truly large models into production, accelerating real-world experimentation with near-frontier systems.