AI Compute: Covina Startup Secures $100 Million for Inference Chips

General Compute Inc. announced on Tuesday that it closed a $100 million Series A round, positioning itself against the dominant cloud providers that typically manage storage and databases. The Covina-based firm is betting on a specific slice of the artificial intelligence market: inference. Unlike competitors that focus on training models with massive datasets, General Compute builds infrastructure for deploying trained models in real-time environments.

The startup emerged as the AI industry shifted focus from model creation to practical application. While giants like Amazon, Microsoft, and Alphabet control the bulk of the cloud market, General Compute targets the high-performance needs of inference workloads. This niche allows smaller firms to offer no-frills services tailored specifically for running AI models rather than just storing data.

The new funding was co-led by Blackbird Ventures and Abstract Ventures. Santa Monica-based BOLD Capital Partners, Covina-based Village Global, and Fuse Venture Partners also participated in the round. The capital will support the construction of infrastructure designed for chips made specifically for inference.

General Compute has assembled a suite of inference-focused hardware from manufacturers including Positron, Cerebras, and SambaNova. Finn Puklowski, co-founder and chief executive officer, stated that the era of one-size-fits-all AI hardware is over. He noted that the future belongs to specialized silicon as inference becomes the largest computing workload.

Puklowski said the company’s mission is to bridge the gap between advanced alternative chips and companies that need them. The first deployment from General Compute is scheduled to go live in the first quarter of 2027.