SambaNova Systems has always moved to its own rhythm, and January 2026 is another reminder that this company does not dance on anyone else’s tempo. Founded in November 2017 in Palo Alto, SambaNova came out of Stanford with a full stack ambition, not a chip demo. Rodrigo Liang, Kunle Olukotun, and Christopher Ré did not set out to win a benchmark contest. They set out to build infrastructure that could carry enterprise and sovereign AI workloads without collapsing under power draw, memory limits, or regulatory gravity.
The name was never an accident. Samba is motion, syncopation, pressure applied in the right places. Nova is the burst. New energy. That blend mirrors Rodrigo Liang’s own path from Taipei to Brazil to Silicon Valley, and it shows up in how the company builds. Custom silicon. A compiler that actually understands dataflow. Managed services that assume customers care where their data sleeps at night. This is not AI as spectacle. This is AI as plumbing, and the pipes matter.
Now the moment. On January 21, 2026, SambaNova Systems went out to market seeking a $300 million to $500 million Series E. This comes after a proposed Intel acquisition, valued around $1.6 billion including debt, stalled late last year. That number sat well below the $5.1 billion valuation stamped in April 2021, and the market noticed. So did the employees. So did the customers. The easy narrative would be retreat. That narrative does not survive contact with the floor.
One day later, Rodrigo Liang told the team that the conversation around SambaNova Systems had shifted. Fiscal year 2026 sales targets were exceeded. Sovereign AI infrastructure contracts landed across three continents. Australia, the United Kingdom, and Germany moved from talking about AI sovereignty to deploying it, with energy profiles that do not melt the grid and data residency that does not make regulators twitch.
This is where the product earns its keep. Reconfigurable Dataflow Architecture is not a slogan. It is the reason SambaNova racks pull 8 to 15 kilowatts while GPU-heavy alternatives gulp ten times that. It is why model switching happens in microseconds, not the length of a coffee order. It is why national labs, financial institutions, and governments with long memories are willing to bet production workloads on a platform that does not require them to contort their data or their laws.
The boardroom story adds its own bassline. Lip-Bu Tan serves as Executive Chairman while running Intel, a relationship born from deep semiconductor history and recent complexity. Intel Capital has been in since Series B. SoftBank Vision Fund 2, BlackRock, Temasek, GIC, and GV all have skin in this. Capital has been patient. Capital has also marked things down. That tension is real, and it is earned.
SambaNova Systems is not asking the market for applause. It is asking for conviction. Inference is where AI lives day to day, and sovereign, energy-aware infrastructure is no longer theoretical. The beat is already playing. The only open question is who steps onto the floor next and decides this rhythm fits their future.
