What happened
Inference neocloud General Compute raised a $15 million seed round at a $60 million post-money valuation, led by FUSE VC with participation from Carya Venture Partners and Village Global Ventures. The company plans to deploy $300 million worth of SambaNova's SN50 specialized inference chips, claiming 600 to 700 tokens per second for LLMs like MiniMax 2.7, significantly exceeding typical GPU performance of 250 tokens per second. These SN50 chips are air-cooled and consume less power, allowing installation in existing data centres without new infrastructure investments.
Why it matters
Specialised inference hardware is emerging as a critical factor for AI service providers, shifting performance and deployment economics. Procurement teams and data centre operators gain options for high-throughput, lower-power solutions that avoid costly water-cooling infrastructure upgrades. General Compute's adoption of SambaNova's SN50 chips, claiming 600-700 tokens per second, contrasts with general-purpose GPUs at 250 tokens per second, directly impacting operational costs and service delivery speed. This follows Cerebras' $57 billion IPO last week and Nvidia's $20 billion Groq transaction in December, highlighting increasing investment in dedicated AI compute.
