The most valuable resource in the American energy transition isn't lithium or sunlight. It's the time of a licensed engineer who can navigate a utility's interconnection queue. Vela Energy, a San Francisco startup, is betting that the right kind of AI can give that engineer a small army of assistants, turning a months-long slog of paperwork and procurement into a coordinated, automated workflow [Vela Energy, retrieved 2026]. The company, founded in 2025, just raised a $1.3 million pre-seed round led by a16z Speedrun and Z Fellows to build what it calls "AI execution agents" for large-load energy projects [YesPress, July 2026]. For data center developers staring down multi-year delays to get power, the pitch is simple: let software handle the repetitive tasks so humans can handle the hard decisions.
The wedge in the interconnection queue
Vela's software targets the operational "last mile" of power infrastructure projects. It doesn't just answer questions about a project; it is designed to execute specific tasks like preparing procurement documents, filing permit applications, and generating engineering studies. The company says its agents apply project-specific records and engineering context to produce actual work product, which is then presented to a human for approval [Vela Energy, retrieved 2026]. This human-in-the-loop model is central to the pitch, ensuring a licensed engineer remains the final authority on every conclusion or document. The intended buyers are the teams responsible for moving massive power projects through development: engineering, procurement, and construction (EPC) firms, owner’s engineers, energy developers, utilities, and data center operators [Vela Energy, retrieved 2026].
A team built at the intersection
Vela's co-founders bring a mix of hands-on industry experience and youthful ambition. CEO Tarun Batchu, a current software engineer at DoorDash, has a background steeped in climate technology and policy, having studied in UC Berkeley’s M.E.T. program and pitched projects to major energy companies as a teenager [YesPress, July 2026]. COO Tony Li brings direct experience from the front lines of energy hardware, having worked on bioenergy projects and, crucially, supported Megapack procurement at Tesla before resigning to co-found Vela [Vela Energy, retrieved 2026]. It’s a pairing that suggests an understanding of both the software execution layer and the gritty, physical reality of ordering and installing megawatt-scale equipment.
The company's early financial backing and accelerator pedigree signal investor belief in the timing of this bet.
Pre-seed 2026 | 1.3 | M USD
The risks of automating a regulated world
For all its promise, Vela is stepping into a domain defined by caution, regulation, and institutional inertia. The company has not yet disclosed any named customers or commercial deployments, demonstrating its product with representative projects instead [YesPress, July 2026]. Selling into utility procurement departments or major EPC firms is a famously long and relationship-driven process. Furthermore, the company must navigate potential brand confusion with an unrelated, older renewable energy platform also named Vela Energy. The core technical risk is whether an AI agent can reliably handle the nuance and variability of local permitting codes and utility-specific engineering standards without introducing costly errors. Vela's rebuttal is its human-in-the-loop design, but the unit economics of that model,how much time is truly saved versus supervised,remain unproven at scale.
What to watch in the next 12 months
The next year for Vela will be about moving from a compelling demo to a contracted tool. Key signals will be its first publicly disclosed utility or developer customer and the expansion of its technical team, for which it is already recruiting founding engineers [Y Combinator, retrieved 2026]. The real test is whether it can compress a timeline meaningfully. If a standard interconnection study takes a human team 300 hours, and Vela's agent can cut the human involvement to 60 hours of review and correction, that's a 240-hour saving. At an average fully-loaded engineering rate of $150 per hour, that's $36,000 of recovered productivity per study. That number needs to be larger than Vela's subscription fee to make the math work for a customer. To win, Vela doesn't need to beat a direct software competitor,there aren't any named yet. It needs to beat the incumbent: the sprawling, familiar, and deeply frustrating spreadsheet-and-email chain.
Sources
- [Vela Energy, retrieved 2026] Vela Energy | The AI workforce for power delivery | https://www.velaenergy.ai/
- [YesPress, July 2026] Vela Energy: The Startup Trying to Unstick America's Power Queue | https://yespress.io/vela-yc-w26
- [Y Combinator, retrieved 2026] Vela Energy Jobs Listings | https://www.ycombinator.com/companies/vela-energy/jobs