The brief

Analyst view

• Raised $1B Series F at $11B valuation in July 2026; Series E ($350M, Vista-led) in February 2026 with Intel co-investing $35M.
• RDU approaches inference as graphs where data streams through operations like assembly lines, not GPUs.
• Intel cleared antitrust in May 2026. SN50 announced Feb 2026, shipping later; SambaCloud runs Llama 4 Maverick at 655 tokens/sec.

Technical approach

As reported

Dataflow architecture

Graph-based computation model

Models as graphs where data flows directly operation to operation. RDU keeps data in motion via on-chip memory.

Memory-compute grid

PCU and PMU codesign

RDU grid combines Programmable Compute Units (PCU) and Memory Units (PMU). Data stays local, cutting memory trips.

Scalability

Up to 256 RDUs in a system

Multiple RDUs can work together for larger models using the same SambaFlow programming model, enabling seamless horizontal scaling.