The case is for owning an energy basket around the AI data-center buildout, with BLDP, CEG, VST, and GE as the cleanest ways to capture 20-year hyperscaler power demand. Bloom is framed as the highest-upside but highest-risk name: it needs only a natural-gas pipeline, could avoid grid and siting friction, and is still early in monetizing a decade of R&D, but valuation, negative cash flow, and scale execution remain the key risks. CEG and VST are preferred for a more traditional mix of power generation, nuclear expansion, and long-term PPAs with Meta, AMZN, MSFT, and Facebook. The core thesis is that compute sits on top of energy, and the next years look like an energy glut only if the hyperscaler contracts and nuclear buildout fail to turn into durable revenue.
AI-driven data-center buildout over the next 5 years needs faster compute growth, making energy the missing piece and positioning Bloom as the likely biggest winner.
Bloom’s edge is a natural-gas pipeline requirement, not a natural-gas generator or turbines, plus lower noise pollution and less municipal/grid disruption because the data center produces its own energy.
Bloom is described as structured more like a technology company than a traditional energy company, with a decade of R&D still early in monetization and no product ramp at scale yet.
Bloom’s downside is high valuation and negative cash flow, even as revenue growth and raised guidance are cited as reasons to push execution forward.
Bloom’s premium energy pricing is said to get cheaper every year, with economics potentially reaching lockstep with equivalent-cost energy while still offering faster ramp-up.
Bloom still has to prove itself because half the world allegedly thinks it is not a real company; multiple short reports have claimed it lacks scandium about every nine months for the past 3 years.
Independent reports are cited as rebutting the scandium short thesis by arguing they miss recent efficiency gains in Bloom’s technology.
The main Bloom trigger risk is scaling failure: if deployments take much longer or big deployments break at wide scale, the whole position gets re-evaluated.
Proof of scale with one of the big hyperscalers is required; if Bloom does not deliver the entire data center, it is viewed as screwed, so the approach is baby steps before more deals.
CEG and VST are preferred because they combine traditional generation with AI-related work and long-term PPAs with large AI companies.
Meta, AMZN, MSFT, and Facebook are described as signing long-term power purchase agreements with CEG and VST to push cleaner energy goals.
CEG and VST are said to have government loans and commitments pushing them to fire up nuclear, while also keeping utility generation running on natural gas in a two-sided approach.
VST’s 2027 earnings outlook is described as 20% to 30% earnings growth, with 40% earnings growth this year.
PPAs are framed as hyperscaler power purchasing agreements covering the next 20 years, and the biggest tech companies in the world are said to be competing to lock in electricity for that span.
The energy-basket view is that the demand is either a passing bubble or the contracts play out; the stance taken is that PPAs will play out and power demand will not go away anytime soon.
GE is cited as the more boring name in the basket, adding turbines and selling out all the way to 2032, with backlog extending to 2032.