SambaNova Raises $350M+ for Agentic AI Chip
SambaNova Systems unveiled its latest chip designed for agentic AI workloads and announced it has secured over $350 million in Series E financing. The funding round included investors such as Vista Equity, Intel Capital, and T. Rowe Price. The company is developing hardware specifically optimized for multi-agent coordination and real-time inference, signaling a broader industry shift toward building the AI stack around workflow-driven use cases.
- This new financing brings SambaNova's total capital raised to over $1.48 billion; its previous Series D round in April 2021 valued the company at $5.1 billion, though a new valuation was not disclosed with the Series E round. - The company was founded in 2017 by Stanford professors Kunle Olukotun, known as the "father of the multi-core processor," and Christopher Ré, a MacArthur Genius Award recipient, along with former Oracle executive Rodrigo Liang. - The new chip, the SN40L, is built on what SambaNova calls a Reconfigurable Dataflow Unit (RDU) architecture, which creates custom processing pipelines to minimize data movement, contrasting with the architecture of traditional GPUs. - The SN40L chip utilizes a three-tier memory system that combines on-chip SRAM, on-package High-Bandwidth Memory (HBM), and off-package DDR DRAM, allowing it to handle models with over 10 trillion parameters. - As part of the announcement, SambaNova revealed a multi-year strategic collaboration with Intel to scale manufacturing and create integrated, cost-effective AI infrastructure as an alternative to GPU-based systems. - A single SambaNova system rack combines 16 of the new SN40L chips and can be integrated into existing air-cooled data centers. - SoftBank Corp. is slated to be the first customer to deploy the new chip in its AI data centers in Japan to power inference services for enterprise clients. - Prior to this, SambaNova's technology has been adopted by U.S. national labs, including Lawrence Livermore and Los Alamos, to accelerate workloads for scientific research.
