Futurist
Photonics, Space, AI Infrastructure & Robotics Commentary
AI's electricity shortfall creates a massive opportunity for companies solving the power bottleneck. Morgan Stanley estimates a 32 GW US data-center power shortfall by 2028. Constellation Energy and Vistra are nuclear names, Bloom Energy in fuel cells, GE Vernova in power generation and Eaton in grid infrastructure. x.comÂ
Power, not chips, is now AI's biggest constraint, per Jensen Huang, making AI energy one of the next decade's biggest investment themes. Big Tech is locking up nuclear: Microsoft at Three Mile Island, Meta 1.1 GW at Clinton, Amazon 690 MW at Calvert Cliffs and Google 890 MW of new nuclear. Beneficiaries span generators like Constellation and Vistra, grid builders like GE Vernova and Quanta, cooling from Vertiv, nuclear from Cameco to Oklo, on-site power from Bloom and storage from Fluence. x.com x.comÂ
A projected US power shortfall makes independent power producers Talen Energy and Vistra no-brainer investments. Morgan Stanley Research estimates a 38 GW net US power shortfall in 2026-28, up from 32 GW, after accounting for additional supply. x.com x.comÂ
Energy providers such as Bloom Energy, Vistra and Constellation Energy look most likely to win from the Musk-neocloud dispute because power is the main US AI bottleneck, as Musk has said. SpaceX might bring 20 GW online next year; that scale could bid up the energy sector even without Bloom supplying SpaceX. x.com x.comÂ
Constellation Energy, Talen, NRG and Modine look weak because of the data-centers-in-space narrative. Once the first big solar-powered rack of GPUs reaches orbit, the market will react and cooling names will crash too, whether or not many space GPUs arrive this decade. @InvestsCletus adds that regulatory hurdles to deals between existing power plants and data centers also weigh on these names, while distributed-generation builders like Bloom Energy stay strong. x.com x.com x.comÂ
Electricity is a clear bottleneck for AI, favoring AI-focused power and data-center infrastructure names over residential utilities. The US could be short 107 GW to power data centers through 2029, while China adds capacity year after year. Medium-term preferences are Cipher, Hut 8, Riot, TeraWulf, Galaxy, Equinix, Bloom Energy, Solaris, Vistra, GE Vernova, INIO, Williams, Liberty Energy, EROC and Cummins. x.com x.comÂ
AI data centers need cooling, electricity and transmission, sustaining the case for Vertiv, GE Vernova and Quanta Services. Vertiv's sales rose 26% in the past year; its systems cool AI chips. GE Vernova is the go-to power trade for its gas turbines. Quanta builds the lines to data centers and has a record $53B backlog. x.com x.com x.comÂ
Access to capital is the only variable that matters for the AI buildout behind Nebius, Talen Energy, CoreWeave, Constellation Energy and Micron, and it is faltering. The trade is extremely reflexive; as @YellowLabLife notes, there is not enough GDP to sustain spending at this rate. Multiples and estimates have to come down. x.comÂ
AI data centers' power needs support a portfolio spanning onsite generation, nuclear, grid hardware and fuel. Oklo offers onsite nuclear and Bloom fuel cells that bypass the grid; Constellation has nuclear baseload under a 20-year Microsoft PPA and Vistra a 20-year Meta deal. GE Vernova supplies grid hardware, Vertiv cooling and power systems, Eos long-duration storage and NextEra a renewable pipeline, while Centrus supplies HALEU for modular reactors and Energy Fuels domestic uranium. x.comÂ
Jan Zuazo Pfingsten
Thematic Investor, Physical Scarcity & AI Infrastructure, Primary-Source Research
Bloom Energy's fast on-site generation could earn a premium by shortening the wait for AI compute. Anthropic's S-1 says it may buy generation equipment above prevailing market rates if it can be deployed faster. Some planned Anthropic data centers rely on dedicated on-site gas generation, and it has committed to cover consumer power-price increases they cause. Bloom's pitch is power in months rather than years waiting for the grid. x.comÂ
Bloom Energy's demand may not become revenue when customers struggle to secure financing. Capital is difficult to raise in this climate; a reading of management's AI discussion also points to permitting, local opposition and customer credit as obstacles despite robust demand. Even after good earnings, the shares did not look very bullish. x.com x.comÂ
AI data centers are hitting a power wall, which favors builders of heavy power and cooling hardware over chipmakers alone. Co-packaged optics let racks pack far more compute, but optical clusters draw enormous current, so new campuses need gigawatts the grid cannot supply, and utility hookups take 5-10 years. Vertiv supplies liquid cooling, Eaton and Powell switchgear and transformers, GE Vernova gas turbines for on-site power, and Bloom Energy fuel cells that skip the grid. x.comÂ
