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

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 

Natural gas faces a real but temporary 2027-2030 price spike as LNG and AI demand outrun buildout lead times, rather than Chronometer's structural, permanent rise. That mechanism is sound and little disputed on timing: processing takes 2-3 years and new frac capacity is unlikely before late 2028-2029. But Chronometer's export and depletion assumptions clash with 2025-26 data, as Haynesville output grew about 2.4 Bcf/d at roughly $4 gas. Rebalancing should follow once capacity catches up. x.com  … Show more

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.com 

Grid-connected AI data-center operators TeraWulf, Hut 8, Applied Digital and IREN look better placed than rivals building energy strategies solely around gas turbines, which will eventually backfire. The midterms are only the beginning; local opposition, resident protests and political pressure could then force leaders to restrict, delay or kill those projects. @MikaelXeeland counters that behind-the-meter power already exceeds 25% of planned US data-center capacity. x.com 

Natural gas is undervalued because the market hasn't connected AI infrastructure spending to the power that runs it. Everyone is buying AI infrastructure while nobody is buying the power behind it, and shale depletion is real. Even so, EQT fell nearly 4% this week, a sign the AI trade's power problem is not yet priced in. x.com 

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.com 

Global resistance to data centers could constrain December cloud growth despite the longer agentic AI buildout. Some estimates say roughly 120 US projects worth over $130B face delays; opposition spans the UK, EU, Australia and Malaysia too. Microsoft, Alphabet and Amazon remain favored, though guidance is at risk. CoreWeave, Nebius and IREN may bounce hardest short term but look weak longer term as funding tightens; Elon Musk expects extensive neo-cloud consolidation. x.com  x.com  x.com  x.com 

Natural gas is where the real squeeze lies if the strike on Iran's Pars region knocks out supply. South Pars/North Dome, shared with Qatar, is the world's largest gas reservoir, feeding Europe, Asia and global LNG. Europe is lean on storage into winter, Asia is fighting for every spare cargo, and the US becomes the only reliable alternative supplier. Henry Hub has been sleepy, but that changes fast with 40% of global LNG in question; the next 72 hours decide whether this is contained. x.com 

Antero, Expand Energy and Comstock sit on a timing gap as AI-driven gas demand arrives in a concentrated burst rather than a smooth ramp. Turbines, not gas, have been scarce: orders surged in 2024-2025 with three-to-five-year lead times, so the first large batch should arrive within about a year, often at behind-the-meter sites near gas fields. Buyers avoiding turbine waits are turning to mass-producible reciprocating engines that also burn gas. Storage, drilling and pipelines have grown slowly. … Show more