AI compute is the hardware and infrastructure used to train and run AI systems at scale. It includes GPUs or other accelerators, servers, storage, networking, power, cooling, and the cloud capacity that ties them together. In practice, AI compute is the fuel that makes large models possible.
In cyber security, AI compute matters because it is expensive, scarce, and highly concentrated. That makes it a target for theft, abuse, and disruption: attackers may hijack cloud accounts, steal API access, overrun quotas, or use compromised infrastructure for unauthorized model training and inference. Defenders protect AI compute with strong identity controls, workload isolation, inventory tracking, patching, and monitoring for unusual GPU usage or network traffic. When compute is tightly managed, organizations reduce both cost leakage and the risk of exposing sensitive data or critical AI services.