SambaNova Launches Fifth-Gen AI Chip SN50, Announces Multi-Year Collaboration with Intel

On February 24, SambaNova unveiled its fifth-generation AI chip SN50, touting it as the only chip capable of delivering the speed and throughput required for agentic AI. It achieves speeds up to five times faster than comparable chips, supports a single model with up to 10 trillion parameters, and offers a context window of 10 million tokens through multi-chip interconnect.

Compared to Blackwell B200 GPU, SN50 is five times faster at peak performance across a range of models, including Meta's Llama 3.3 70B, and delivers over three times the throughput for agentic inference. For inference tasks running models like GPT-OSS, SN50 is up to eight times more energy-efficient than the B200 GPU.

SN50 is scheduled to begin shipping in the second half of 2026. SoftBank Group will be the first customer to deploy SN50 in its next-generation AI data centers in Japan.

On the same day, SambaNova and Intel announced a multi-year strategic collaboration aimed at delivering high-performance, cost-effective AI inference solutions for AI-native enterprises, model providers, corporate clients, and institutions worldwide, built around Intel Xeon based infrastructure. Intel Capital is also participating in SambaNova's Series E financing round.