The modern power grid is a math problem that has outgrown its human solvers. Every new rooftop solar array, every electric vehicle charger, and every battery storage unit adds a variable to an equation that must balance supply and demand across thousands of miles of copper and steel. For grid planners, a single year of hourly power-flow analysis for one circuit can take weeks. ThinkLabs AI, a 2024 spinout from GE Vernova, thinks the answer is not more engineers, but a better digital brain.
Its product, an AI-powered digital assistant called Copilot, aims to automate the complex planning and operational workflows that keep the lights on. The company just secured a $28 million Series A, led by Energy Impact Partners and backed by strategic players like NVentures and Edison International, to build it out [GlobeNewswire, Mar 2026]. The bet is that utilities, facing unprecedented complexity, will pay for software that speaks the language of physics, not just data.
The physics-informed wedge
ThinkLabs is not building another generic machine learning model. Its core differentiator is the phrase "physics-informed AI." This means its algorithms are trained on and constrained by the actual laws of electrical engineering,Ohm's law, Kirchhoff's laws, power flow equations. In a sector where a software hallucination can lead to a blackout, this is the only kind of AI that gets a seat at the control room table.
The platform works by first creating a digital twin of a utility's grid, converting physical assets into a high-fidelity, physics-based network model [thinklabs.ai, retrieved 2024]. On top of this foundation, AI agents automate specific, high-value tasks for planners and operators.
- Interconnection automation. Streamlining the process for connecting new solar farms or industrial facilities to the grid.
- Capacity and flexibility planning. Identifying where the grid is constrained and where distributed energy resources can provide relief.
- Real-time operational support. Performing tasks like state estimation and contingency analysis to keep the system stable.
The promise is speed at a scale humans cannot match. In a pilot with Southern California Edison, ThinkLabs reportedly reduced the time for detailed annual circuit analyses from weeks to under three minutes, generating engineering reports in about 90 seconds [citybiz, Unknown]. For an industry racing to integrate renewables, that kind of acceleration is not a nice-to-have; it's a prerequisite.
A team forged inside the incumbent
ThinkLabs did not emerge from a hackathon. It was incubated inside GE Vernova, one of the world's largest grid technology suppliers. This origin story is its most potent commercial asset. Founder and CEO Josh Wong was previously a General Manager for Grid Orchestration at GE Vernova, and key technical leaders like Head of Machine Learning & Data Engineering Surendranath Vallabhajosyula also hail from the industrial giant [Crunchbase, Unknown] [Environment+Energy Leader, Unknown].
This grants the startup two critical advantages: deep domain credibility and pre-existing relationships with the very utilities it now sells to. The leadership team it has built reflects a blend of big-grid experience and modern software execution.
| Metric | Value |
|---|---|
| Seed (Incubation) | 0 M USD (Spinout) |
| Series A (2026) | 28 M USD |
Strategic capital and the path to utility CIOs
The Series A investor list reads like a who's who of strategic energy capital. Energy Impact Partners is a specialist climate fund with deep utility ties. NVentures is Nvidia's venture arm, signaling a belief in the computational demands of the task. Edison International is a major regulated utility holding company, a potential lighthouse customer and distribution channel rolled into one [GlobeNewswire, Mar 2026].
This capital is a tool to land and expand within utility organizations. The initial wedge is likely a specific, painful workflow like interconnection studies or hosting capacity analysis. Once integrated, the platform can expand to adjacent use cases, aiming to become the central AI layer for grid management. The collaboration with Southern California Edison, part of the Edison International family, is the first public proof point of this strategy in action [LA Times, Jan 2026].
Where the electrons could stop flowing
For all its tailwinds, ThinkLabs faces a market defined by caution. Selling mission-critical software to electric utilities is a marathon of compliance, security reviews, and long sales cycles. The company's direct competitors, like Pearl Street Technologies and Urbio, are also chasing the same automation dream with different technical approaches.
The larger risk is cultural. Grid engineers are, rightly, skeptical of black-box solutions. ThinkLabs's answer is its physics-informed foundation,it must prove its AI is a trustworthy colleague, not an inscrutable oracle. Its success hinges on demonstrably improving an engineer's work without asking them to blindly trust the output.
Financially, the $28 million war chest is substantial but will be consumed quickly by the high-cost enterprise sales motion and the computational expense of training and running massive, physics-constrained models. The next twelve months will be about converting pilot projects like the one with SCE into multi-year, seven-figure enterprise SaaS contracts.
The incumbent it must beat
On paper, the back-of-the-envelope calculation is simple. If a utility spends 10,000 engineer-hours annually on manual grid studies at a blended rate of $150 per hour, that's $1.5 million in labor cost exposed to automation. A ThinkLabs subscription that cuts that time by 80% could command a significant portion of that saved cost while making the grid more nimble.
The real competitor for ThinkLabs, however, is not another startup. It's the internal spreadsheet, the legacy planning software suite, and the institutional inertia of the utility itself. It must beat the ingrained habit of doing things the slow, known way. If its Copilot can become as indispensable to a grid engineer as the calculator once was, the $28 million bet will look very small indeed.
Sources
- [GE Vernova, May 2024] GE Vernova launches startup ThinkLabs AI, Inc. | https://www.gevernova.com/news/press-releases/ge-vernova-launches-startup-thinklabs-ai-inc
- [GlobeNewswire, Mar 2026] ThinkLabs AI Closes $28 M Series A Led by Energy Impact Partners | https://www.globenewswire.com/news-release/2026/03/31/3265239/0/en/ThinkLabs-AI-Closes-28-M-Series-A-Led-by-Energy-Impact-Partners-Backed-by-NVentures-and-Edison-International.html
- [thinklabs.ai, retrieved 2024] ThinksLabs - AI that Powers the World | https://thinklabs.ai/
- [citybiz, Unknown] ThinkLabs AI pilot with Southern California Edison | https://www.citybiz.co/article/310704/thinklabs-ai-pilots-its-ai-platform-with-southern-california-edison/
- [Crunchbase, Unknown] Josh Wong - Founder & CEO @ ThinkLabs AI | https://www.crunchbase.com/person/josh-wong-828c
- [Environment+Energy Leader, Unknown] ThinkLabs AI appoints Surendranath Vallabhajosyula | https://www.environmentalleader.com/2025/10/thinklabs-ai-appoints-surendranath-vallabhajosyula-as-head-of-machine-learning-data-engineering/
- [LA Times, Jan 2026] ThinkLabs AI collaborates with Southern California Edison | https://www.latimes.com/business/technology/story/2026-01-15/thinklabs-ai-southern-california-edison-grid-solutions