Key points:
The conventional wisdom among tech optimists held that software innovation would always outpace hardware constraints. That assumption is now collapsing under the weight of reality. AI models, particularly the large language models powering tools like ChatGPT and Google Gemini, consume electricity at staggering rates. A single query to an advanced AI system can require ten times the energy of a standard Google search. When multiplied across billions of users and millions of corporate applications, the electrical demand becomes astronomical.
Goldman Sachs research projects that global data center power demand will grow by 165 percent by the end of the decade compared to 2023 levels. This is not a gradual increase. This is a surge. And the existing grid infrastructure, already strained by aging equipment and political battles over transmission lines, simply cannot keep up. Utilities in the United States are now quoting two to four year wait times just to complete feasibility studies before any construction begins. For companies operating on quarterly earnings cycles, that delay is unacceptable.
Google learned this lesson the hard way. The company set aside $1 billion to build a data center in Indiana, believing it had found a suitable location with adequate power. But local planning commissions, wary of environmental impact and community opposition, threatened to veto the project entirely. Google reversed course. The money was spent. The power was not secured. This is not an isolated incident. It is a pattern repeating across the country as tech giants discover that permits, not innovation, are their biggest obstacle.
Long before AI forced this reckoning, another digital industry understood the physical limits of the virtual economy. Bitcoin miners, operating on a proof of work system that ties every unit of output directly to electricity consumption, faced the same crisis years earlier. The Bitcoin network consumes an estimated 175 terawatt hours per year in 2025. Producing a single Bitcoin today requires hundreds of megawatt hours of energy. Miners that relied on retail electricity rates or temporary discounts were priced out of the market. Those that controlled long term, low cost power sources survived and scaled.
This is where the investment strategy becomes clear. The companies that treated electricity as a product rather than a cost, that built their own infrastructure rather than renting grid capacity, are now positioned exactly where AI demand is exploding. Bitzero, trading on the NASDAQ under the ticker AIBZ, followed this blueprint. The company established itself in regions where power is abundant, underutilized, and structurally advantaged: hydro rich Norway, nuclear backed Finland, and grid secured rural North Dakota. Combined, this represents more than a gigawatt of potential clean power capacity tied directly to its own infrastructure. Most of that power is already permitted, connected, or under construction.
Amazon recognized the same imperative. The company signed a long term agreement with Talen Energy for up to 1,920 megawatts of carbon free nuclear power, one of the largest corporate electricity deals ever announced. Microsoft went even further, signing a 20 year agreement with Constellation Energy to enable the restart of Pennsylvania’s Three Mile Island Unit 1 nuclear plant, securing 835 megawatts of carbon free power for its expanding AI infrastructure. The symbolism is unmistakable. The site of America’s worst nuclear accident is being resurrected to power the AI revolution.
But for smaller investors, the question is which companies will benefit most. Constellation Energy has emerged as one of the biggest winners of the AI infrastructure boom, completing its $26.6 billion acquisition of Calpine to create the largest power producer in the United States with roughly 55 gigawatts of generating capacity. Yet the opportunity extends beyond utilities. Bitzero’s all in electricity cost lands around 3.5 cents per kilowatt hour, among the lowest on Earth. The company mines Bitcoin for roughly $50,000 per coin while the industry average exceeds $100,000. When Bitcoin trades at $80,000, most miners struggle. Bitzero still generates about $1 million in monthly EBITDA from its existing 40 megawatt Norway site.
Kevin O’Leary, the Shark Tank veteran who became a strategic investor in Bitzero, summed up the logic bluntly. Both AI data centers and bitcoin miners are going to be “fighting for power contracts.” Companies like Bitzero will end up acting as power companies, providing this critical resource to those who need it most. The market is already validating this thesis. As one 110 megawatt lease agreement in Norway demonstrated, the revenue potential is enormous. OneQode Networks signed a binding letter for a 15 year lease covering Bitzero’s Norway campus, a deal that could generate roughly $2.6 billion in contracted revenue over its lifetime.
The truth that Wall Street is only beginning to absorb is simple. Data may be the world’s most valuable resource, but data cannot exist without electricity. The companies that control the power will control the future. The rest will be left waiting for permits.
Sources include: