There's a peculiar irony in the way tech companies talk about infrastructure these days. They'll pay billions for real estate, erect warehouse-sized buildings to house servers, then discover—sometimes months into construction—that nobody can actually turn the lights on. Not enough power on the grid.
Enter Soma Energy, which thinks it has found a workaround. The Vancouver startup came out of stealth April 2, 2026, with $7 million in its pocket and a pitch that sounds almost too convenient: use artificial intelligence to squeeze more capacity out of the grid that already exists, helping data centers get online in months instead of years.
Whether that's innovation or just expensive improvisation depends on who you ask.
Category Ventures led the seed round. Haystack, RRE Ventures, TO VC, and Uncork Capital joined in, along with earlier pre-seed backers Panache Ventures and Walter Kortschak. There's a minor discrepancy in the telling—Uncork called itself a co-lead in a blog post, while Soma's own press release credits Category alone. These things happen in venture-backed startupland, where everyone wants a piece of the narrative.
The Team That Knew Amazon's Grid
The founders came from AWS, which lends a certain credibility. CEO Athanasios Caramanolis led the company's renewable energy optimization efforts, scaling a portfolio to 10 gigawatts and negotiating deals worth over a billion dollars. His co-founder and CTO, Mario Souto, has a PhD in electrical engineering and built machine learning systems to manage solar, wind, and battery storage at similar scale.
Then there's Henrique Hoeltgebaum, the chief AI scientist, whose background centers on forecasting and anomaly detection for energy grids. Together, the trio claims fluency across the alphabet soup of North American grid operators: CAISO, SPP, MISO, PJM, ERCOT, WECC. If you don't know what those acronyms mean, suffice it to say they represent the people who decide whether your lights stay on.
Still, pedigree only gets you so far.
A Market That Can't Wait
Soma says it's already working with five data center clients and optimizing two gigawatts of power for producers. One of those customers, H5 Data Centers, went on record. CEO Josh Simms said Soma helped his company access capacity "significantly sooner than expected." Sooner than expected is a relative term, of course, but in an industry where timelines stretch like taffy, maybe it counts for something.
The context matters here. BloombergNEF projected in December that U.S. data center power demand would hit 106 gigawatts by 2035—a number they revised upward by 36 percent from a forecast issued just seven months prior. And the day before Soma's announcement, Axios noted that data center spending is now closing in on the investment levels typically seen in the energy sector itself, per Rystad Energy.
Translation: there's a lot of money chasing a very finite resource.
What the Software Actually Does
Soma's platform breaks into three pieces. AI Trader handles procurement and hedging. BESS Dispatcher coordinates batteries, deciding when to store or release power. Load Optimizer tries to match computing workloads with available energy in real time.
The theory goes like this: by creating a coordination layer between power generation, storage, and consumption, Soma can help data centers extract more capacity from grid connections that already exist. For independent power producers, the software ostensibly delivers smarter dispatch decisions—when to generate, when to store, when to sell—lowering costs and smoothing out price swings.
It's an appealing idea, particularly for an industry that has spent the better part of two years scrambling to secure megawatts. The question is whether optimizing around bottlenecks is a sustainable business or merely a clever Band-Aid while the grid catches up.
The Road Ahead
Soma plans to deploy the fresh capital into hiring—engineers, salespeople—and expanding across North America. The company is making a bet that the mismatch between data center demand and grid supply won't resolve itself anytime soon. That bet might be right. Utilities move slowly. Permitting takes forever. New transmission lines can require a decade.
But software has limits too. You can optimize dispatch algorithms and shift workloads all you want; at some point, you still need actual electrons. Soma's pitch is that the gap between now and then is wide enough to build a business.
Whether that gap stays open long enough—or whether it even needs to—is the gamble every investor in the round is taking. For now, the founders have runway. The market has urgency.
And somewhere, a data center is still waiting for someone to flip the switch.
