- SambaNova has completed the first close of a $1 billion Series F round at an $11 billion post-money valuation, led by General Atlantic
- The round follows its $350 million-plus Series E from February, when it launched its SN50 chip alongside an Intel collaboration
- JPMorganChase has selected SambaNova’s SN40 and SN50 systems to power secure, on-prem AI inference
For years, the AI infrastructure race centred on training ever-larger foundation models. Now investors are paying a premium for the companies that can run those models cheaply and fast in production, and SambaNova’s valuation just doubled on that bet alone.
SambaNova, the San Jose-based AI infrastructure startup founded in 2017 by Rodrigo Liang, Kunle Olukotun and Chris Ré, has completed the first close of a $1 billion Series F financing at an $11 billion post-money valuation.
The raise follows the $350 million-plus Series E it closed in February, when it launched its SN50 chip and struck a multi-year collaboration with Intel to build joint AI inference infrastructure.
“SambaNova’s $11 billion valuation highlights the central role that fast inference now plays in the enterprise AI stack,” Liang says.
Why investors are backing inference AI
General Atlantic led the new round, with significant participation from Seligman Ventures, T. Rowe Price Associates and Capital Group. Existing investors including funds managed by BlackRock, Intel Capital, Qatar Investment Authority, Vista Equity Partners and Battery Ventures also returned, while A&E Investment, Assam Ventures, Cambium Capital, Kabila Capital, QFO Capital and Volantis join as new backers.
As enterprises push AI agents into banking, healthcare, manufacturing and government, the cost centre is shifting from training models to running them. The AI inference market was valued at $97.24 billion in 2024 and is projected to reach $253.75 billion by 2030, according to Grand View Research, and that shift is exactly what SambaNova’s pitch leans on.
“SambaNova’s platform is differentiated, built for a market where inference has become foundational to enterprise and industry transformation,” Martín Escobari, co-president and head of global growth equity at General Atlantic, tells Tech Funding News.
“As AI moves into production, lowering cost per token while maintaining performance and efficiency will be critical for enterprise adoption. SambaNova’s purpose-built RDU architecture is uniquely positioned to address that challenge at scale,” adds Umesh Padval, managing partner at Seligman Ventures and a board observer at SambaNova.
SambaNova’s SN50 chip, launched alongside its Intel collaboration in February, is built to run agentic AI workloads at roughly three times lower cost than GPU-based systems, according to the company, with a single cluster able to connect up to 256 chips. That is the technical bet the new valuation now hinges on delivering at commercial scale.
JPMorganChase joins as customer
Alongside the funding, SambaNova revealed that JPMorganChase has selected its SN40 and SN50 systems as an inference infrastructure partner for secure, on-prem AI deployments.
“At JPMorganChase, AI infrastructure has to meet a very high bar for performance, control and reliability, Darrin Alves, chief information officer of infrastructure platforms at JPMorganChase.
A marquee bank as a named customer is a stronger signal than the funding figure alone. It gives SambaNova a commercial reference point that its wafer-scale and chip-only rivals cannot yet claim in the same regulated-industry category.
Competition intensifies as valuations climb
Competition in AI inference is heating up fast. Cerebras has continued expanding its wafer-scale AI systems after securing a $1 billion round at a $23 billion valuation, while Groq has emerged as another leading inference specialist, having raised a $640 million Series D in 2024 at a $2.8 billion valuation.
Unlike either rival, SambaNova sells a full stack, chips, systems, software and cloud, directly to enterprises, sovereign AI programmes and regulated industries, rather than competing purely on chip speed.
SambaNova says it will use the new capital to expand chip and systems manufacturing capacity, accelerate product development and scale deployments for enterprises, neo-clouds, sovereign AI customers and service providers.
The question now is not which company can train the largest model. It is which infrastructure providers can make inference cheap and reliable enough to make agentic AI commercially viable at scale, and whether a vertically integrated challenger like SambaNova can hold that ground against both GPU incumbents and the hyperscalers building their own silicon.
