Solar CEO: As AI Grows, Energy Becomes the Limiting Factor

For years, software bragged about moving fast while power picked up the tab. Now, data centers, electric cars, and heat pumps all pull from the same worn grid. Deep Patel, founder and CEO of GigaWatt, the parent company of DIY Solar companies Unbound Solar and Real Goods, believes the story of AI is really an electricity tale.

That’s because every mass scale energy model eventually lands on someone’s power bill. Patel shared the below predictions and thoughts about the future of AI, data center, and solar's role in it all.

AI runs on power, and it wants it constantly. Data centers pull electricity in thick, steady streams that make old demand curves look quaint.

Patel warns that “energy becomes the real bottleneck for AI. Even the founder of OpenAI has said the limiting factor on GPU expansion is energy, not hardware—a dynamic that I think will accelerate in 2026.”

Patel doesn’t expect that squeeze to stay inside server rooms. “AI has created a two-part energy problem. Data centers need fast, scalable power, and homeowners need protection from the cost pressures AI will put on the grid,” he says. The same megawatts that keep a chatbot running also shape residential bills in a big way.

Electricity was an afterthought of many household budgets until recent increases pushed it to the front. As utilities race to keep up with new demand, costs may rise while patience for outages stays thin. He expects that “electricity bills will become a national affordability crisis, even as customers still face outages and reliability issues.”

That pressure sends households looking for exits close to home. Solar panels and battery storage can signal a lifestyle choice or a neighbor trying to win Earth Day, and more buyers treat them as financial tools and backup plans that may soften rate spikes and keep the lights on when the wider system struggles.

Centralized grids were built for a different era of use. AI, constant connectivity, and electrified homes introduced new spikes into that picture and shortened the path between a server cluster and someone’s kitchen.

One prediction from Patel highlights how “Natural gas may temporarily support data center expansion, but new turbine infrastructure takes six to 10 years to commission and does nothing to shield households from rising electricity rates.” He adds, “Solar and storage already work today, can be deployed immediately, and serve both sides of the equation.”

Distributed solar and batteries answer only part of the problem, giving households and small businesses more control over when, where, and how they consume power. For investors, that shift may signal a move away from pure AI enthusiasm and toward the hardware that keeps those models online when grids start to creak.

AI will keep demanding more power, even if the hype around it cycle cools. Electric vehicles, heat pumps, and server farms share the same system that already groans on hot afternoons during peak hours. In another prediction, Patel says, “2026 kicks off a global AI-energy arms race, with countries competing for electrons the way they once competed for oil.”

For households, the question increasingly centers on timing. Bills rise, outages sting, and ignoring rate notices gets harder. The next era of energy may be decided as much on rooftops and in breaker boxes as in boardrooms and trade shows. In that sense, the most important part of AI’s story may turn out to be the part that happens off-screen, wherever the next kilowatt is coming from.