Another company in the NeoClouds category has secured significant funding: General Compute has just raised $100 million. General Compute is a neocloud (dedicated AI cloud) company focused on ultra-high-speed AI inference. - It breaks down the inference process by purchasing, deploying, and operating heterogeneous hardware: using NVIDIA/AMD GPUs for prefill (prompt preprocessing) and dedicated inference chips (ASICs) such as Cerebras, SambaNova, and Positron AI for decoding (token generation), achieving speeds and latency far exceeding pure GPU clouds. - Customers don't need to purchase and manage expensive dedicated chips themselves; they can directly obtain bare-metal servers or guaranteed tokens-per-second (TPS) SLAs and use open-source/open-weighted models (such as MiniMax and DeepSeek) through OpenAI-compatible APIs. This is particularly suitable for speed-sensitive scenarios such as AI agents, code agents, and real-time speech. The company's core strategy is to bet that "demand for fast inference far exceeds demand for slow inference and will grow exponentially." It is deploying non-Nvidia chips on a large scale through debt financing (such as the $400 million Upper90 facility) and providing services in managed data centers, with the first large-scale capacity planned for Q1 2027. Website: https://www.(generalcompute.com)/