Has the hunt for AI compute uncovered the next Cerebras?

TL;DR

General Compute, an AI inference neocloud, raised $15 million in seed funding.
The company is specializing in renting out AI processing power for model inference, not training.
General Compute is deploying specialized chips from SambaNova, aiming to outperform GPUs.
SambaNova's chips are air-cooled, consuming less power and simplifying data center deployment.
The company is pursuing colocation deals to house its hardware in existing facilities, including those of crypto miners.
General Compute claims to be the fastest at running MiniMax 2.7, an open-source LLM.
The inference cloud model bets on a future with multiple AI models where speed and cost of inference are key competitive factors.

Continuar leyendo el artículo original