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. Watts Lindqvist reports on a new entrant trying to conduct that orchestra.
Vaquita Energy, founded in 2024 and based in San Francisco, has posted a single, declarative sentence on its public website: it is building “Energy Orchestration Infrastructure for AI Factory” [Vaquita Energy, 2024]. The ambition is clear, if the details are not. 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]. In essence, it is a traffic control system for electrons, where the destination is a rack of GPUs.
The Coordination Wedge
The bet is that 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. The problem is that these systems,a battery from Tesla, a turbine from Mitsubishi, a solar array from a third installer,don’t 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, crucially, guarantee uptime [F4, 2024]. The value isn’t in selling the batteries or panels, but in the software that makes the whole ensemble cheaper and more reliable to operate.
An Early and Quiet Start
Public information is exceptionally 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 [1, 2026]. There are no listed customers, case studies, or partners. This places Vaquita in a category of pure, 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 risks here are foundational, not incremental. 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 who are undoubtedly working on similar problems. The absence of any public traction or technical validation means the clock is ticking to move from a declarative website to a provable product.
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 even 10%, it could save that cluster over $1.2 million a year. That’s the unit economics they need to prove. 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.
Sources
- [Vaquita Energy, 2024] Vaquita, Energy Orchestration Infrastructure for AI Factory | https://www.vaquitaenergy.com/
- [F4, 2024] F4 | https://f4.fund/startups/vaquitaenergy
- [1, 2026] Background on Ami Zou | [Source details unspecified]