Vistra's shift toward long-term hyperscaler contracts may warrant a higher multiple than its roughly 14x forward earnings, versus about 21x for Constellation. Its 3.8 GW of nuclear commitments with Meta and Amazon could reduce exposure to volatile wholesale prices; Lotus added 2.6 GW of gas capacity and the pending Cogentrix deal could add 5.5 GW. Further contracts and controlled leverage remain important to the upside case. x.com 

Power is now the binding constraint on AI compute, making power suppliers like Vistra, Constellation, Talen and NextEra the clean AI infrastructure trade. About 20 earnings calls showed Vistra turning Meta's nuclear demand into contracted revenue, Exelon warning on PJM reliability and NRG sizing ERCOT's large-load pipeline at 36 GW. AI infrastructure is becoming investment-grade, with Hut 8 issuing the first single-sponsor IG construction bond for a data center. x.com 

Vistra's story is easy to understand: AI data centers need far more power, and Vistra already owns the power plants. AWS and Meta have signed long-term nuclear deals for about 3.8 GW, some of whose upside isn't even in 2027 guidance yet, so demand is obvious before earnings fully catch up. x.com 

2G Energy's growing orders and long service tail support its unusually cheap AI data-center power valuation. Nine-month 2026 orders exceed €880m, more than twice 2025 revenue; backlog runs into 2028 and service contracts last 15-20 years. Management first guided to €750-850m of revenue in 2028 and says units already sold could support over $100m in annual service revenue once running from mid-2027. At about 17x expected 2027 earnings, it looks cheapest in this space and is preferred to INNIO. … Show more

Vistra is growing exponentially on the AI buildout yet trades at 17 times forward earnings and a 0.5 five-year PEG on a $56B market cap. It has a deeply diversified energy footprint with one of the largest nuclear fleets in the US, and Meta and Amazon are among its biggest customers. The chart is one of the best setups around, and almost no one is paying attention. x.com 

Constellation Energy owns the scarcest asset in the one commodity hyperscalers absolutely need: 24/7 firm clean power from the largest US nuclear fleet, 21 reactors and about 33 GW that cannot be replicated this decade. US power demand is growing fastest since WWII, and about 75% of carbon-free output is uncontracted, so each Microsoft- or Meta-style PPA re-rates earnings. Management guides 20% EPS CAGR through 2029, at 23x forward, with the PTC floor limiting downside. x.com 

New Era Energy & Digital is lining up power and permits for a 1.4GW AI campus, Texas Critical Data Centers, on 493 acres in Ector County. It signed a 20-year deal to buy up to 207MW from Vistra's Odessa gas plant, with power expected in Q3 2027, and Vistra gets a 5% non-voting stake in the part it powers. Construction permits are in hand, but the data center business is pre-revenue, making it the highest-risk name on the list. x.com 

Power demand will keep growing exponentially as AI models and data centers get larger, even after the AI hardware rotation slows, creating an AI power supercycle across four layers. Fuel: EQT, Kinder Morgan, Williams, Energy Transfer. Onsite power, the near-term bottleneck fix: Bloom Energy, GE Vernova, Caterpillar, Cummins, Kodiak Gas Services, T1 Energy. Grid: Eaton, Quanta, Vertiv, Hubbell, nVent, Emerson. Nuclear baseload: Oklo, NuScale, Nano Nuclear, Constellation, Vistra, BWX Technologies. … Show more

Oklo is a high-conviction play because advanced nuclear is the answer to the AI energy crisis and the company is executing. It reported its first revenue, $1.2M, in Q2 2026, agreed with Meta to develop a 1.2 GW clean-energy campus in Ohio for AI data centers, and signed a letter of intent with Centrus Energy for domestic HALEU fuel for up to five Aurora reactors. The chart shows a higher low after a dip to $36.61, and options gamma clusters at the $50 strike. x.com  x.com 

Deliverable electricity is by far the most important AI bottleneck, the backdrop for owning energy infrastructure such as Williams. BloombergNEF now expects 118GW of US data-center capacity by 2030, 52% above its December 2025 forecast, and 194GW by 2035, about 20% of US energy use. Expected AI-chip sales imply 207GW of US capacity, 63GW more than BNEF's realistic forecast, so chips are outrunning power and something has to give. Data centers are also faster to build than grid expansions. x.com