SambaNova Systems Raises Over $350 Million in Series E Funding to Scale AI Chip Production Amidst Competition with NVIDIA

The room did not feel loud when the number crossed. $350M+ tends to do that. It quiets conversations. It forces posture changes. It makes people stop scrolling and read slower. SambaNova Systems just closed an oversubscribed Series E, led by Vista Equity Partners with Cambium Capital in the mix, and Intel Corporation lining up roughly $100M with room to go higher. Not hype money. Decision money. The kind that shows up when infrastructure stops being theoretical and starts being unavoidable.

SambaNova Systems has always moved like a hardware company that understands software better than most software companies understand hardware. Founded in Palo Alto in 2017 by Rodrigo Liang, Kunle Olukotun, and Chris Ré, this was never about chasing the loudest benchmark. It was about building a full-stack AI platform where silicon, software, and services actually speak the same language. DataScale. Dataflow as a Service. A reconfigurable dataflow architecture that treats inference like a discipline, not a workaround.

Rodrigo Liang, Co-Founder & CEO, runs this thing with the calm of someone who has built processors before they were fashionable dinner conversation. Kunle Olukotun, Co-Founder & Chief Technologist, brings the long memory of multi-core computing, earned the hard way, in labs where theory meets heat and power curves. Chris Ré, Co-Founder, shows up with statistical rigor and a refusal to dumb anything down. Add Lip-Bu Tan, Executive Chairman since May 2024, and yes, simultaneously serving as CEO of Intel Corporation, and you get a boardroom that understands both the physics and the politics of compute.

Vista Equity Partners does not wander into silicon by accident. Their history lives in enterprise software where margins, discipline, and operational rigor are non-negotiable. That matters. Cambium Capital leaning in matters. Intel Corporation staying close after acquisition talks cooled matters. This round lands in a market desperate for inference that does not melt data centers or budgets, where NVIDIA is the reference point and alternatives have to prove they can scale without excuses.

SambaNova Systems is selling to enterprises that want answers now, not GPUs later. National labs, global consultancies, serious operators who care about tokens per watt and latency per dollar. The SN40L RDU exists because inference is no longer a side quest. It is the workload. Air-cooled racks, massive memory tiers, microsecond model switching. This is what happens when you build for production instead of demos.

The quiet part is the lesson. Capital flows fastest to teams who picked their lane early, ignored fashion cycles, and kept shipping while the market argued. Hardware is hard. Full stack is harder. Doing both while the industry debates shortcuts is harder still. SambaNova Systems kept its head down, its architecture tight, and its customers real, and the money followed, as it usually does when the work shows up first.