Archive — US — techcrunch
Has the hunt for AI compute uncovered the next Cerebras?
General Compute, a new AI inference neocloud, raised $15 million to meet the demand for specialized chips for running AI models. With a focus on SambaNova's advanced chips, General Compute aims to enhance processing speed and reduce costs in data centers. Their service claims to lead in running powerful open-source models efficiently, marking a pivotal shift in AI processing technology.
Key points
- General Compute raised $15 million in seed funding for its AI inference service.
- The company aims to address the high demand for AI processing power, focusing on the post-training phase of AI model execution.
- SambaNova's specialized chips outperform traditional GPUs, generating 600-700 tokens per second compared to 250 for GPUs.
- The chips are air-cooled and consume less power, easing installation in existing data centers.
- General Compute's offerings are drawing parallels to established relationships in the AI hardware and cloud sector, suggesting a promising outlook for their growth.
Why it matters
- The growth in AI has made specialized computing essential, indicating a trend towards dedicated hardware.
- Recent investments in AI chipmakers reflect a broader industry shift toward optimizing performance for inference tasks.
- General Compute's strategy illustrates a move towards more flexible data center solutions, similar to trends seen with cloud computing.
- Historical context includes the significant IPOs and partnerships of companies like Cerebras and Groq, showcasing the competitive landscape for AI processing.
General Compute's innovative approach to AI inference processing and its strategic partnership with SambaNova position it well in the rapidly evolving AI landscape, potentially reshaping how AI models are run in the cloud.