Felix Prehn Names 4 Power Stocks as the Real AI Trade; Bitcoin (BTC) Faces the Same Grid

Prehn’s AI Thesis Puts Electricity Ahead of Chips

Former banker and analyst Felix Prehn argued in an August 24 X thread that the next major artificial-intelligence trade is electricity, not chips, and named Constellation Energy, Talen Energy, Vistra and GE Vernova as the four companies closest to that bottleneck. Prehn opened the thread by saying investors who missed Palantir at $20, Intel at $45 or Seagate at $95 were not alone — each stock ran 500% to 1,000% in the next 12 months — and that he went hunting for similar candidates. His list is built on contracted demand rather than speculative exposure: Big Tech is signing long-term nuclear and power-generation agreements even as hundreds of billions of dollars flow into chips and data-center construction, and without reliable clean power that spending cannot fully become operating capacity. The four names, he says, sit directly in the path of that demand, each tied to concrete contracts with major AI infrastructure buyers. Constellation owns the largest US nuclear fleet, including the restarted Three Mile Island unit now branded Crane Clean Energy Center, and signed 920 megawatts of new long-term nuclear power purchase agreements in Q2 2026, including a Walmart deal. Talen holds a long-term Amazon Web Services contract covering up to 1,920 megawatts. GE Vernova’s AI data-center orders more than doubled in the first half of 2026 versus all of 2025. The stocks still trade far below their peaks: Constellation is down roughly 34% from its 52-week all-time high, Talen is correcting from an all-time high near $451, and Vistra trades near $135, well off its 52-week high of $219.82. Prehn recommends having a clear exit strategy before entering any of these positions, noting that recent gains already discount a large part of the expected growth.

The structural case rests on a shortage of reliable clean power rather than a speculative narrative. Prehn describes the four as sharing a single catalyst: growing, contracted demand for clean and reliable power, which separates them from speculative AI plays tied to chip demand or software hype alone. Constellation’s second-quarter disclosure showed the 920 megawatts of new agreements average 18.5 years in duration, and management raised adjusted operating earnings guidance to $11.50 to $12.50 per share while describing existing plants as the bedrock for powering data centers in this early phase. Talen, which owns the Susquehanna nuclear plant, raised its 2026 adjusted EBITDA guidance to $2.025 billion to $2.225 billion and free cash flow guidance to $1.2 billion to $1.35 billion after closing the Cornerstone acquisition, with a pipeline of roughly 4 gigawatts in data-center options. Vistra, backed by long-term contracts with Meta and Amazon, launched Helix Digital Infrastructure alongside NVIDIA, KKR and the Kuwait Investment Authority at an initial commitment of up to $1 billion; it also posted more than 30% growth in ongoing operations adjusted EBITDA in Q2 and received FERC approval for its Cogentrix purchase. GE Vernova’s gas turbine backlog reached 116 gigawatts, management expects to surpass 125 gigawatts by year-end, the company’s overall backlog reached $176 billion, launched a new MV-UPS system built for AI factories, signed battery storage contracts in Australia, and shares trade near an all-time high of roughly $1,196 reached in July 2026. Prehn’s warning is central to the trade: if AI spending slows or expected profits fail to materialize on schedule, this group could correct sharply, and he advises diversification and position sizing alongside the thesis.

The list contains no crypto assets, but the electricity bottleneck Prehn describes is the same variable that shapes Bitcoin (BTC) mining economics, since proof-of-work is effectively an electricity bill. While his four picks are equities rather than altcoins, the underlying tension is familiar to digital-asset markets: secure power supply and pricing determine whether Bitcoin mining capacity can grow. That symmetry is why Bitcoin’s hashrate tends to concentrate in regions with low-cost power, and why grid access is a competitive moat for miners. In his thread, Prehn is explicit that utilities holding contracted AI demand are the beneficiaries, and equally explicit that the trade reverses if the buildout disappoints. For crypto investors, the implication is a shared grid — tight power markets pressure miner margins even as they hand pricing power to generators that have already signed long-term agreements.