The most expensive part of running an AI factory is no longer the silicon. It is the electricity required to keep it from melting, and the backup systems that ensure it never blinks. Vaquita Energy, founded in 2024 and based in San Francisco, is building “Energy Orchestration Infrastructure for AI Factory” [Vaquita Energy, 2024]. The company aims to sit above the proprietary controllers from equipment manufacturers and coordinate a heterogeneous mix of on-site power assets, battery storage, gas generators, solar, and wind, to serve the highly variable and immense loads of modern AI compute clusters [Vaquita Energy, 2024].
The Coordination Wedge
AI data centers are becoming their own microgrids. As power demands push beyond what local utilities can reliably deliver, operators are building behind-the-meter generation and storage. These systems,a battery from Tesla, a turbine from Mitsubishi, a solar array from a third installer,do not naturally talk to each other, or to the compute workload scheduler. Vaquita’s proposed platform would be the hardware-agnostic layer that does, attempting to reduce system stress, improve overall efficiency, and guarantee uptime [F4, 2024].
An Early and Quiet Start
Public information is sparse. The company is at the pre-seed stage, with no announced funding rounds or named investors. The founder is Ami Zou, whose background includes studies in algorithms and internet protocol. There are no listed customers, case studies, or partners. This places Vaquita in a category of early-stage ambition, sharing a conceptual space with industrial giants like Mitsubishi Power Americas and Hitachi Energy, but approaching from a software-integration angle rather than a hardware-sales one.
The company must prove its coordination logic works safely at scale, convince risk-averse data center operators to trust a new software layer with their most critical infrastructure, and out-execute the internal teams at cloud hyperscalers. For a sense of the stakes, consider a single, modest AI training cluster drawing 10 megawatts. At California’s high industrial electricity rates, that’s about $35,000 a day just in power costs. If Vaquita’s orchestration can smooth demand spikes and increase the utilization of cheaper on-site storage by 10%, it could save that cluster over $1.2 million a year. To succeed, Vaquita Energy must become for the on-site AI microgrid what a building management system is for a skyscraper: invisible, essential, and saving money with every decision it automates.