Energy & Power Generation

7GW undisclosed turbines available plus mobile units enable immediate power; SpaceX and Microsoft (Chevron 2.7GW) shifting to behind-the-meter generation to avoid 18-month grid delays, accepting medium-voltage inefficiencies for speed.

Starship's low-cost launch enables space-based solar power stations that beam infinite clean energy to orbital AI data centers, solving the terrestrial power constraint that limits AI scaling and turning energy abundance into a strategic moat for SpaceX.

Renewable energy adoption is accelerating faster than appreciated, with California at 50%+ solar and New Mexico shifting from gas to wind/solar/batteries; Tesla's plan to scale solar production to 100+ GW/year vertically integrated will crash energy costs, creating a massive productivity tailwind not priced into markets.

In retail choice markets (Texas ERCOT, Illinois), Base Power sells electricity directly to homeowners and monetizes battery fleet for grid services. In regulated markets (California, El Paso), utilities control batteries and pay for access. This dual-model approach unlocks value across US market structures and aligns incentives with grid needs.

Space-based solar power beaming loses 95% energy in transmission. StarCloud's pivot to orbital compute keeps energy in space, using DC power directly from solar arrays (no AC/DC conversion loss) to run GPUs. This makes space compute energetically superior to both terrestrial and beamed-power models.

Unprecedented private investment in fusion energy; DOE national strategy updated with more private capital than ever before, suggesting fusion timeline may accelerate beyond perennial '50 years away' narrative.

Renewables are rapidly displacing fossil generation (California 50%+ solar, New Mexico 70%+ wind/solar/battery), Tesla plans order-of-magnitude solar scaling, and China is industrializing fusion, collectively making abundant cheap energy a near-term reality.

In deregulated markets like Texas (Houston, Dallas, parts of Austin), Base Power acts as a retail electricity provider, selling power to homeowners while using aggregated home batteries to lower wholesale costs. In regulated markets (California, El Paso), they partner with utilities who control the batteries and pay for capacity. This dual-model unlocks a massive TAM across US grid architectures.

Historical data shows gold rises before inflation but underperforms during actual inflationary periods (1980s); equities/S&P 500 outperform gold over full cycles because they own productive cash-flowing assets.

Exxon Mobil has risen 189% over five years to a 22x P/E even though net income is lower than 20 years ago, reflecting a safety-seeking premium that ignores commodity cyclicality and long-term energy transition headwinds.

Energy availability for new data centers is emerging as a bigger constraint than even chip supply; China's state-directed buildout of power generation and robotics gives it structural advantage in the physical infrastructure layer of AI.

Short-term auto-industry protection delays combustion bans, but China's green-tech dominance proves early movers capture markets; ignoring climate sacrifices 20-year GDP for quarterly optics.

Alan argues that energy consumption per capita, not financial GDP, measures real economic productivity. UK consumption down 25% in 25 years, US flat, China surging. Advocates all-of-the-above energy strategy (coal, gas, solar, wind, nuclear) with carbon as engineering constraint not objective.

The bottleneck for AI scaling has moved from chip supply to energy availability, making investments in energy storage, geothermal, virtual power plants, fusion, and fission critical for AI infrastructure and climate resilience simultaneously.

Planet A's core thesis expects energy costs to decline toward zero over the next two decades, fundamentally changing manufacturing, compute, and resource economics in Europe's favor.

Boom Supersonic repurposes jet engines for data center power; Helion pursues fusion; Zephyr Fusion targets space-based fusion; 2-3 year lead times for power equipment create structural energy bottleneck for AI buildout.

Bavarian government provided €400M non-dilutive capital alongside Max Planck Institute (1,200 people) and RWE, creating a de-risked ecosystem that catalyzed a €411M private round led by European investors (Google, KfW, family offices). Model demonstrates how sovereign deep-tech champions can be built without early US capital dependence.

For the first time in history, the market dynamics of AI compute create sufficient ROI to fund sustainable energy and grid upgrades purely on commercial terms, making energy stocks winners and enabling a self-funding green transition.

Real-time AI control of industrial processes like aluminum smelting cell baths and gas turbine operations can materially reduce energy consumption and improve yield, linking industrial AI directly to power generation efficiency.

1KOMMA5° has aggregated 1 GW of flexible capacity from home batteries, EVs, and heat pumps, qualifying for frequency containment reserves in Sweden. Each subscriber adds 10-30 kW, and the flywheel between hardware sales and Heartbeat AI subscriptions accelerates re-purchase cycles from years to months with zero CAC.

With nuclear 5-10 years out, gas turbines (days to deploy, 6-12 month life) and solid oxide fuel cells (4-7 year life) are the only viable near-term solutions; companies controlling this equipment and permitting (GEV, BE) capture massive pricing power.

Ireland hosts outsized cloud capacity but grid investment hasn't kept pace with AI-driven demand; fixing energy infrastructure is a prerequisite for Europe to capture AI supply chain value beyond model training.

Using the Stockholm Resilience Centre's planetary boundaries framework, Transition identifies power availability and critical mineral supply as the binding constraints for AI and industrial transformation, directing capital to energy generation, grid infrastructure, and resource efficiency.

When AI builds AI and robots build robots (10-15 years out), the only binding constraints are energy cost and access to iron, copper, coal. China and Russia hold structural advantages here; Germany has neither cheap energy nor resources. This makes manufacturing and robotics unwinnable for Germany, reinforcing the case for energy-light sectors like biotech and software.

Massive data center power demand is driving up electricity prices, which in turn accelerates innovation in energy generation including fusion and other sources.

Hyperscaler capex surging 62% and data center electricity demand projected to nearly triple by 2030 creates structural demand for gas turbines (Siemens Energy) and drives nuclear renaissance, as the grid cannot expand fast enough to meet new AI load.

The EU Battery Regulation mandates >80% lithium recovery by 2031, but tozero demonstrates the unit economics are already viable today — customers are pre-signing contracts for local supply to hedge against raw material price swings and supply chain disruptions.

E-methane and synthetic natural gas can decarbonize chemical, cement, steel, long-haul aviation, and rocketry overnight as a chemically identical drop-in replacement, addressing sectors where batteries and direct electrification are physically impractical.

Companies like Bloom Energy are temporary beneficiaries of desperate AI power needs; as grid infrastructure and cheaper generation come online, their pricing power and high multiples will erode.

Solar assets with existing PPAs mark-to-market at higher power prices driven by gas scarcity, generating margin expansion without capex; residential solar becomes economic without subsidies as peak daytime prices surge 2028-2030.

Manufacturers (Caterpillar, Bloom) are expanding gas turbine/fuel cell capacity assuming cheap gas, but structural deficits post-2028 will make gas too expensive for these high-heat-rate assets, leading to order cancellations and impaired investments.

First Solar's case reveals that domestic thin-film solar manufacturing cannot compete with Chinese crystalline silicon on cost without ~$2B/year in Section 45X production tax credits. With backlog declining despite the subsidy advantage, the structural competitiveness of US-made solar modules remains unproven, creating a policy-dependent investment thesis for the entire domestic solar supply chain.

Massive data center buildouts assume linear power scaling, but algorithmic efficiency gains (1Gb to 100Gb fiber analogy) could drastically reduce per-token energy needs, stranding power infrastructure and turning data centers into 'pickleball courts'.

Renewable energy represented 93% of all new electricity generation installed worldwide last year, with IEA projecting >95% ahead. Private capital supplied 85% of the investment, proving the transition is now market-driven rather than policy-dependent.

Fortescue's commitment to zero fossil fuel by 2030 through electrification and green hydrogen will eliminate $1B/year in diesel costs, proving renewable energy can outcompete fossil fuels in hard-to-abate sectors without subsidies.

With gas prices elevated, coal use is ticking up in China, India, Indonesia, the US, and even parts of Europe as a short-term energy security measure, reversing the expected plateau and creating a temporary tailwind for coal producers.

Holcim gets paid to take waste materials, processes them into fuel substitutes, reducing both fossil fuel use and CO2 emissions while lowering costs — demonstrating sustainability as a profit driver rather than a cost burden in heavy industry.

Nigeria's gas resources (largely discovered accidentally while drilling for oil) are far larger than oil; deliberate gas exploration plus 5,600km pipeline to Spain via West African coast could position Africa as Europe's primary LNG supplier, replacing Russian gas.

The primary risk from AI is not stretched valuations but its energy intensity pulling the world off its decarbonization trajectory; climate change remains real and requires both mitigation and adaptation, with IMF already integrating climate shock risks into country programs like Jamaica's resilience planning.

Intelligence applied to energy systems yields efficiency gains (Google's 40% data center savings) that offset AI's electricity demand; AI-enabled grid management and green compute scheduling create net positive climate impact.

ECB research finds meeting solar and wind targets could reduce electricity prices by more than a quarter by 2030 and weaken the gas-electricity price link; battery pack costs have fallen to roughly one-tenth of 15 years ago, making storage increasingly viable for managing intermittency.

Tesla's energy business (generation + storage) is scaling alongside AI and manufacturing, providing the power foundation for the autonomy platform; the speaker expects energy to eventually be queried via API like autonomy tokens once the business reaches sufficient scale.

Texas has doubled utility-scale solar in 5 years, 8x'd wind, and doubled storage, leveraging permissive permitting and federalist competition to become the 'balance sheet of AI'. Data centers follow energy; Texas model shows renewables + fossil backup can scale faster than California's regulatory approach.

Historical nuclear achieved ~3 cents/kWh (2024$) with massive bespoke construction; factory-produced SMRs and microreactors with learning-curve economics should structurally undercut that, enabling energy-intensive manufacturing (aluminum, steel) to reshore.

Solar/wind have doubled every 22 months for 40 years; IEA consistently underestimates by decades (predicted 2030s gas crossover, happened 2026). China 10x US in solar deployment. Perovskite next S-curve. Winning nations connect cheapest electrons to compute. Investment implication: energy-compute nexus, solar supply chain automation (SpaceX/Tesla), grid infrastructure, Dyson swarm orbital solar as ultimate scale.

When consensus saw oil at $12-15, Kaplan's analysis said $100. He built Leor Exploration, made a monster Texas gas discovery, sold to Encana for $2.55B in Nov 2007 — timing the top before the financial crisis and shale revolution changed the calculus.

2.5-3 year waits for power/step-up transformers bottleneck AI buildout. 50% of 9GW planned 2026 compute delayed by protests. Nuclear parallel: US zero new reactors in 30 years vs China 100. Lunar solar advantage (cheap panels, easy cooling). Protests 'noise in hindsight' but cause near-term delays.

The world is fundamentally short watts (power) and wafers (TSMC capacity); this dual shortage may prevent AI capex overbuild even if demand exists, making power generation and grid infrastructure critical enablers.

IREN's design target of 130-200 kW per rack reflects the rapid escalation of power density in AI clusters. This creates a structural demand for high-voltage grid connections, on-site generation, and advanced cooling — a tailwind for power infrastructure providers and a gating factor for data center deployment speed.

Data center infrastructure stocks (energy, cooling, power) returned 419% past year. Markley bought 1M valves for liquid cooling (10 leaks/day). Generators sold out. 100 GW US target requires massive power generation, transmission, cooling innovation. Energy is co-equal with chips in the singularity loop.

US energy dominance (largest oil/gas producer) enables an investment continuum: export LNG to allied ports, build regasification, power critical mineral mining (power-intensive), and extend consumer power to villages and businesses — replicating the Marshall Plan's market-hardwiring effect at scale.

Joby's hydrogen-electric VTOL achieved 500+ mile flight leveraging hydrogen's 3x specific energy over jet fuel and 2x fuel cell efficiency. This breakthrough could redefine all aircraft classes for commercial and defense, solving the energy density bottleneck for electric aviation.

The compute demands of AI factories will drive a trillion dollars per year into energy infrastructure — nuclear, renewables, hydrogen, and grid equipment — representing the largest energy investment cycle in a century.

Current AI trajectory consumes gigawatts for inference and training; with humanity's total brain power at only 160 GW vs 9,000 GW global electrical capacity, the von Neumann compute paradigm's inefficiency will exhaust available energy within years, making fundamental substrate innovation an existential necessity for AI scaling.

The human brain's 20W power consumption for 86B neurons versus GPU racks requiring hundreds of megawatts suggests that energy constraints on AI scaling may be solved by mimicking biological efficiency rather than building more power plants.

Emanuel agrees with tech leaders that the physical grid — transmission, distribution, and permitting — may delay full AGI deployment more than compute or algorithms, buying society 'a little bit of time' to adapt.

The merged entity aims to scale space-based solar power to 100 terawatts (100x US current generation), solving AI's energy bottleneck and enabling civilizational-scale compute.

American Turbines leverages uniquely cheap US natural gas to build pickup-truck-sized 1MW generators for rapid deployment. Founder argues turbine design is simple (<40 parts); the real moat is automated mass manufacturing — a Ford Model T moment for distributed energy.

California's aggressive refinery regulation eliminated in-state capacity (including Chevron's West Coast largest refinery), forcing imports of dirtier fuel from thousands of miles away with larger carbon footprint, while losing high-paying jobs and tax base. Texas delivers cheaper, cleaner power through innovation and grid infrastructure. Smart policy — like incentivizing midday EV charging when solar is abundant — can align climate and affordability goals, benefiting utilities, grid tech, and EV infrastructure.

If Leopold's hedge fund is positioning in solar developer T-1 Energy, it suggests solar power can deploy on timelines relevant to AI data center buildout (years not decades), challenging the assumption that only fossil fuels or nuclear can meet near-term compute power demand.

Placing storage behind the meter at the load side (homes) smooths both generation intermittency and demand peaks, raising utilization of existing distribution lines from <50% toward 100%. This avoids costly new transmission build-out — charging in one city and discharging in another creates 'virtual transmission' without physical wires. Texas's real-time wholesale market (5-min pricing) enables immediate monetization.

Texas pays generators only for energy produced (price changes every 5 min), creating efficient price signals that reward flexibility. Capacity markets (PJM, ISO-NE) pay for availability, locking in less efficient generation. This market design lets distributed batteries arbitrage price spikes ($5/kWh vs 9¢ retail) and sell ancillary services — a structural advantage for storage deployment that other states may replicate.

IREN locates at stranded renewable sites (West Texas 45-50 GW vs 12 GW transmission) to convert cheap green electrons into high-value compute tokens; grid connection guarantees 24/7 reliability.

IREN sees nuclear as structural correlation between human progress and energy consumption; SMRs will unlock distributed clean power for compute but require decade-long lead time for deployment.

The cost to generate electrons has fallen sharply due to solar, but total electricity prices have risen because aging grid infrastructure makes moving electrons expensive. Distributed batteries deployed at load centers move power through time instead of space, capturing cheap generation while avoiding costly transmission and distribution upgrades — a structural tailwind for storage deployment.

Texas' competitive wholesale market exposes participants to real-time price signals, creating strong incentives for developers to build supply when prices rise. This has made Texas the national leader in solar, wind, and now data center growth — a model other regions may emulate as electricity demand surges.

While logic, CoWoS, HBM, and EUV scale in 2-3 years, energy infrastructure (power plants, grid, permitting) takes far longer — making energy policy the critical constraint for US AI leadership and reindustrialization.

A pro-growth economy requires generating more energy alongside jobs and opportunities, particularly for rural and forgotten communities.

Iran war and Ukraine conflict prove energy dependence creates geopolitical vulnerability. Europe is reversing course: Germany restarting nuclear, Italy and Spain reintroducing investment tax credits for energy. Energy independence preserves optionality for Western allies.

Helion's direct electricity extraction (95% efficiency) enables mass-producible 50MW modules. Microsoft PPA de-risks demand. Advance Act regulatory pathway (state-level, ~1 year) vs NRC decade. Manufacturing scale-up (60x) mirrors data center deployment model. Multiple private companies (Helion, Commonwealth, TAE) converging on late 2020s timeline.

AI is the business of turning electricity into intelligence. Energy may be the ultimate bottleneck. US power infrastructure is a patchwork of 1950s technology; China's is extraordinary. Building multi-gigawatt data centers requires solving grid, permitting, and local opposition — all slow, incremental processes.

30% federal ITC until 2033 plus decade of double-digit annual cost reductions (Moore's Law-style) make fuel cells cost-competitive with gas turbines per Rystad Energy analysis, while delivering superior efficiency and emissions profile.

Technology de-risked like early shale revolution; needs time to prove geology, techniques, and management to reach project finance markets for low-cost capital; 5-year horizon for industry inflection.

Panel deflation (mostly China) now a shrinking share of total system cost; land, labor, transmission, and capital costs are inflationary and growing — same dynamic may hit batteries as lithium prices rise 50% recently.

Local opposition and regulatory veto points make US energy infrastructure projects take 10+ years vs 5 planned; bipartisan federal permitting reform is likely this year and is the key policy lever to unlock supply for data center demand.

US natural gas prices fell 20% while Asia/Europe doubled/tripled after Strait of Hormuz closure risk. US is now largest oil/gas producer and exporter, with dramatically less energy-intensive economy than 1970s. This structural energy cost advantage supports AI data center buildout and aligns with administration's manufacturing goals.

After food, energy is the most vital sovereignty; distributed generation using local renewables plus on-site hydrogen storage eliminates dependence on adversarial energy suppliers and prevents resource wars, fundamentally altering geopolitics.

The US, Australia, Canada, UAE, and Qatar can supply affordable, low-carbon natural gas globally; using Bloom's high-efficiency fuel cells avoids coal while enabling rapid deployment, making 'good the enemy of the best' a strategic error.