Flexible Interconnection Promises Speed To Power For Util…

Hyperscalers racing to bring AI data centers online often encounter a common bottleneck when trying to connect to the U.S. power grid. In many cases, it’s not as simple as plug-and-play with the local utility. The electric infrastructure needed to support compute-intensive projects must be built or upgraded, which can take years. Many developers opt for on-site behind-the-meter generation solutions, at least in the interim- time is money, after all.

While some data center companies are comfortable running their own power plants, most prefer utility service, and as soon as possible. Which is exactly why Camus Energy has caught Google’s attention.

Recognizing what it deems a “generational opportunity for AI to redesign a sustainable and resilient global energy system,” Google recently selected 28 energy-focused, AI-driven startups for its 10-week Google for Startups Accelerator program, which runs from September 8 to November 12, 2026. Camus Energy was one of a dozen North American firms tapped (the others are listed at the bottom of this article). The California-based company’s FlexConnect software promises to shave three to five years off typical interconnection wait times utilizing flexible operating limits within existing infrastructure.

That’s what Google is driving at with its accelerator program: helping AI data centers connect to the grid faster using new energy transition techniques that leverage the poles and wires we’ve already got. Participants benefit from technical workshops, strategic consultations, and two in-person milestones with Google’s network of utility and investment partners. The curriculum focuses on grid modernization, demand flexibility, and clean tech infrastructure, empowering startups to develop cutting-edge solutions from electricity generation to grid operations.

As part of the program, Camus Energy will accelerate the development of FlexConnect, which it is demonstrating to large utilities including Duke Energy, Edison International, and PG&E. That demonstration will provide a practical blueprint for interconnecting data centers using flexible operating limits to protect grid reliability while utilities complete upgrades.

“The program gives hyperscalers a chance to move past the all-or-nothing approach to grid access, proving it is possible to reliably unlock massive new capacity while keeping the local grid stable, clean, and affordable,” said Astrid Atkinson, CEO and co-founder of Camus Energy.

David Bailey is a veteran software engineer with more than two decades of experience at Google and Amazon, where he built large-scale search and machine learning systems. He now applies that expertise to transforming complex grid data into reliable, scalable intelligence for Camus Energy. Bailey took some time to respond to questions from Factor This after Camus was selected by Google; the ing is a transcription of our exchange, edited lightly for grammar and clarity.

---

Paul: What can be gained via flexibility when we’re talking about data centers? A little goes a long way, right?

David: There’s quite a bit to be gained from flexibility, and the nice thing is that with it everybody wins. The data center gets connected more quickly because a roadblock has been removed. The utility gains a big customer.

As you say, a little goes a long way, and the data center gets almost all the power they wanted in the first place since the need to cut back is typically rare. And by solving this with software, costly grid upgrades are avoided, which could wind up making electricity more affordable—not less so—for the average citizen. Finally, by better utilizing our existing (and increasingly green) grid rather than building new, inefficient off-grid gas plants, flexibility helps fight climate change.

At a basic level, how does Camus’ FlexConnect software work? At a technical level?

It’s pretty simple, really. The data center connects to the grid with a promise of some firm power plus some conditional power that’s not guaranteed. Camus’ FlexConnect platform cooperates with the utility to monitor and analyze the grid and combines that with weather forecasts to predict those rare moments when conditional power won’t be available. FlexConnect then communicates an “operating envelope” to the data center that helps them comply with their commitment, with advance notice when possible. That lets them plan whether to solve the issue with battery backup or by deferring compute jobs or by other means.

What was your initial reaction to being selected for the Google for Startups Accelerator program? Are there any competing companies/ideas that were also selected that you’ve got an eye on?

I was thrilled when Camus was selected because it’s a great opportunity to deepen our relationship with Google. We are already working with Google to deploy FlexConnect in support of their data center growth needs. Meanwhile, FlexConnect itself is powered by Google Cloud, and we can always use help putting the cloud platform to best use, especially as it expands its AI features. What’s more, I’m a proud ex-Googler myself, having worked on the machine learning algorithms behind Google Search and their implementation across Google’s compute fleet for many years.

As for our fellow selected companies, I really view them through the lens of competition. We are a small community, and we need all hands on deck to solve these important societal problems. I hope we can build each other up. I’ve long admired Grid Status and their amazing and deep visualizations, but I’m also keen to be introduced to some of the companies new to me.

What are you fundamentally hoping to demonstrate to utilities Duke, PG&E, and Edison?

Flexible interconnection can work in real life. We have the tools and the technology to do it. Software-based orchestration of flexible assets can be counted on as a first-class solution to grid reliability.

Camus has been working in this space for a number of years. Previously, we focused quite deeply on the utility side, gaining quite a bit of expertise and trust in integrating utility data into dashboards and insights they deeply appreciate. Over the last year, we’ve added the data center focus, with promising early studies across multiple data center developers. While this aspect is newer, it plays to some of our past experiences, especially for Astrid Atkinson, our CEO, who pioneered a lot of the compute-side orchestration that drives the modern data center world. It’s still early days for proving out the connection between these two worlds, but we’re excited to make it work!

How does Camus Energy see itself positioned in the speed-to-power conversation? What will that conversation look in the future, if data centers can indeed be brought online flexibly?

I think what’s unique about our position is our experience with utilities. Years of hands-on analytics-building for operators and planners, and coordination of smaller flexible assets, leave us well situated to be a trustworthy partner in ensuring that flexible connections truly live up to their potential for improving grid reliability, not detracting from it.

It sure doesn’t hurt, though, that we also have an intuitive sense for what hyperscalers really do and don’t need from their data centers to support their own commitments to reliable compute services, and the ways in which flexible power can, in fact, be accommodated in doing so.

---

Here are the other 11 North American startups selected for the Google accelerator program:

• Anode (United States): an AI-orchestrated virtual utility mobilizing microgrids on demand
• Cosmic Robotics (United States): autonomous construction equipment for solar farms and data centers
• Dig Energy (United States): a modular drilling technology that enables geothermal heating, cooling, and storage
• Felix Fusion (United States): an AI-powered platform that helps data centers, mining operations, and industrial projects find energy and connect to the power grid more easily and flexibly
• Fram Energy (United States): a revenue collection layer for local, independent energy projects
• Grid Status (United States): a modern data and market intelligence platform for the energy industry
• Hestia Energy (United States): a solution that finds and unlocks reservoirs of geologic hydrogen
• Nyquis (United States): a solution that cuts the time to power for data centers and large new loads
• Optiwatt (United States): an electric vehicle–first energy platform connecting vehicle brands with programs that help balance energy usage and demand across North America
• Swish Solar (Canada): an operating system for the world’s solar infrastructure
• Zola Intelligence (United States): an AI-powered platform embedding automation and AI agents into finance and operational workflows alongside smart hardware for energy delivery