David Senra × Zach Dell

US electricity demand could rise from roughly 2% to around 10% annual growth, with AI accelerating an already broad electrification trend.Base sells electricity rather than $20,000 batteries, owning distributed assets that monetize wholesale markets while avoiding interconnection and transmission constraints; scaling Base Core and long-term utility contracts remain key milestones.

Drew Baglino × Lukas Biewald

Baglino argues properly designed AI data centers can lower electricity rates by raising grid utilization: 800MW average demand versus 1-2kW for a home, with storage absorbing training ripple.Heron Link’s 5MW solid-state transformer halves grid-to-chip losses, adding about 35MW of useful compute per gigawatt; the thesis depends on data centers funding interconnection and utilities avoiding overbuild.

Peter Diamandis × Salim Ismail × Dave Blundin × Alexander Wissner-Gross × Ramez Naam

AI’s binding constraint is grid interconnection, not generation economics: roughly 100 hours of annual load flexibility could unlock about 100GW of existing capacity, representing roughly $5T of data-center capex.Texas’s interruptible-load fast track and FERC’s broader push could catalyze near-term capacity, while solar-plus-batteries offer a roughly 12-month build path; winter seasonality, transmission bottlenecks, and uncertain further cost declines remain key risks.

Jeremie Eliahou Ontiveros × Reyk Knuhtsen × Ellie Holbrook × Jordan Nanos

The claim that half of 2026 US data-center capacity was canceled fails a denominator check: SemiAnalysis says early-stage options were mistaken for committed builds, while its forecasts moved less than 5%.The sharper bottleneck is execution at financed projects such as Oracle’s Project Jupiter, supporting more than 40 GW of behind-the-meter additions by 2028 but leaving gas, permits, equipment, and schedules as catalysts and risks.

Sarah Guo × Isaiah Taylor

Valar Atomics treats nuclear as a manufacturing problem, targeting a simple, intrinsically safe “Toyota Camry” rather than a bespoke paper reactor.DOE testing authority enabled its 100 kW W-250 to reach criticality, while passive heat removal and a seven-month milestone interval provide operating evidence.Equity-funded deployment could precede project finance, with AI as an initial catalyst and cheaper energy as the longer-term demand thesis.

Harry Stebbings × KR Sridhar

AI factories make electricity the binding input, and KR Sridhar frames Bloom Energy’s opportunity against a $5.5 trillion power market versus $2 billion of revenue last year.Oracle’s 50+ MW contract in 90 days and delivery in 55 days support the architecture case, while scaling from 1 GW to over 2 gigawatts by year-end leaves permits, gas supply, and customer buildouts as constraints.

Jordan Nanos × Jeremie Eliahou Ontiveros × Nicolas Bontigui × Haroon Inam

AI rack density is pulling data centers toward 800 V DC, which can deliver almost triple the power through the same copper while reducing conductor weight, cost, and I²R losses.DG Matrix’s multi-port SST roadmap targets 400 kW units today, 6 MW in 2027, and 10 MW in 2028, but adoption timing, mixed hardware, protection requirements, grid bottlenecks, and cybersecurity remain material execution risks.

Chamath Palihapitiya × Aaron Cowen × Dan Dreyfus × Oleg Nodelman × Kyle Samani × Jason Calacanis × David Sacks × David Friedberg

MGM’s $48 bid offers a near-term floor while Osaka and possible Dubai legalization create asymmetric upside, whereas Talen monetizes a 2 GW nuclear and 6 GW gas portfolio amid a projected PJM power shortage.Aktis remains a binary 2027 clinical platform wager, and GEODNET links revenue to token purchases; portfolio sizing ultimately turns on liquidity, downside, regulatory intervention, and whether satellites displace terrestrial RTK.

Alex Silver

BWX Technologies combines a contracted, sole-sourced government business representing 75% of revenue with 400+ Navy nuclear cores built without a safety incident and scarce licensing and manufacturing capabilities.ANM’s thesis hinges on 4–5-year time-to-power, a potential mPower-based PWR white space, and more than 100% upside from about $200, but customer interest must de-risk any updated design.

Erin Price-Wright × Turner Caldwell × Drew Baglino

AI’s constraint is increasingly physical: Mariana Minerals targets mines and refineries, while Heron Power applies silicon and software to grid-scale power conversion.Mariana’s software-first operating model spans design through autonomous operation, but projects can take five years to build and another three to five years to reach operating rate; durable permitting, financing, and grid policy remain catalysts for scaling.