SambaNova Systems

SambaNova Systems is an enterprise AI accelerator company headquartered in Palo Alto, California, founded in 2017 by Stanford professors Kunle Olukotun and Chris Ré alongside Google veteran Rodrigo Liang. The company raised $350 million in a Series E round in February 2026 led by Vista Equity Partners and Cambium Capital at a $2.2 billion valuation, bringing total funding to approximately $1.5 billion from investors including Intel Capital, QIA, GV (Google Ventures), Battery Ventures, T.

Rowe Price, and BlackRock. SambaNova built the Reconfigurable Dataflow Unit (RDU), a chip architecture fundamentally different from GPUs: where GPUs process data through fixed pipelines that must be reconfigured for each operation, RDUs implement AI computations as dynamic dataflow graphs that adapt to model structure in real time.

The flagship SN50 chip, unveiled in February 2026, runs AI models up to five times faster than competing processors with a three-tier memory architecture supporting models with up to 10 trillion parameters and 10 million token context lengths — capabilities that exceed what standard GPU clusters can efficiently handle. SoftBank Corp.

is the flagship SN50 customer, deploying the chip across next-generation AI data centers in Japan; SambaNova reports record bookings and revenue as of end-2025 across financial services, telecommunications, energy, and sovereign deployments. Intel holds approximately 9% of SambaNova following a strategic investment and entered a multiyear collaboration to deliver high-performance, cost-efficient AI inference solutions optimised for leading open-source models with predictable throughput and total cost of ownership.

The SambaStudio software platform enables enterprises to deploy, fine-tune, and manage large language models and multimodal models on SambaNova hardware with full data privacy — a key differentiator from cloud-only inference providers. Enterprise customers use SambaNova systems for AI applications requiring maximum model capability with on-premise control, including regulated industries where data cannot leave corporate infrastructure.