Building new power generation takes five to ten years. The AI industry needs power now. That gap is the entire business case for Soma Energy.
The company emerged from stealth this week with $7 million in seed and pre-seed funding, led by Category Ventures with participation from Haystack, RRE Ventures, TO VC, Uncork Capital, and pre-seed investors Panache Ventures and Walter Kortschak. The founding team is not theoretical about this problem - they are the people who built and operated Amazon Web Services' energy infrastructure at scale a decade ago.
What Soma Actually Does
Soma's platform uses AI to optimize across both supply and demand simultaneously in real time - something most energy software companies do not do, typically focusing on one side or the other.
For data centers, the platform connects on-site generation, storage, and load into a single control layer, turning large facilities into flexible grid assets. Rather than waiting years for new transmission and interconnection to be built, it unlocks capacity already present in existing infrastructure and delivers it in months. For power producers, the platform determines in real time when to generate, store, or trade across wind, solar, and battery assets to lower cost per megawatt-hour and reduce price volatility.
The company is currently optimizing two gigawatts of electricity for power-producing clients and is actively working with five data center customers. H5 Data Centers CEO Josh Simms confirmed the model works in practice, noting Soma helped them access capacity significantly sooner than expected and removed a critical constraint on expansion.
The Team Behind It
CEO Ath Caramanolis built AWS's renewable energy optimization program, scaling that portfolio to 10 gigawatts and negotiating over $1 billion in energy deals. CTO Mario Souto built the machine learning platform used to manage gigawatts of solar, wind, and storage assets at AWS. Chief AI Scientist Henrique Hoeltgebaum brings expertise in AI-driven forecasting and anomaly detection for complex energy systems. All three hold PhDs or equivalent deep technical backgrounds, and all three have operated at the scale they are now trying to serve.
Why the Timing Is Right
BloombergNEF projects US data center power demand could reach 106 gigawatts by 2035. McKinsey forecasts global data center capacity will grow 3.5x between 2025 and 2030. Those projections were made before the Iran war sent energy prices surging and added geopolitical complexity to an already constrained grid buildout.
In my four years advising executives on AI adoption, the power constraint is the one most companies underestimated most severely. The assumption was always that infrastructure would follow demand. What Soma is betting - correctly, I think - is that the bottleneck is orchestration, not generation. There is existing capacity in the grid that is underutilized because the intelligence to coordinate it at speed does not exist yet. That is the problem Soma is building to solve.
