Five Maps and a Warning: Energy Geopolitics and the Realignment of Power

In The New Map (2020), Daniel Yergin argues that physical territory, pipeline routes, and embedded infrastructure constrain energy transition more than policy ambition can accelerate it. Five years later, his framework explains why Microsoft is restarting Three Mile Island, why DeepSeek trained frontier models on export-restricted chips, and why the U.S.-China AI race is fundamentally an energy and supply chain story.

---

This is Part 4 of a five-part AI × Energy series. Energy and Civilization established the 50-year rule for energy transitions. Power Density quantified the spatial constraints. The 84% Gap calculated the capital shortfall. This instalment adds the geopolitical layer: how nations compete for energy resources and why that competition now shapes the AI race. Part 5 closes the series with “Jevons Meets Jensen”: can AI realistically be powered at projected scale?

---

TL;DR

Daniel Yergin, Pulitzer Prize-winning author of The Prize and vice chairman of S&P Global, structures his 2020 analysis around five distinct “maps”: America, Russia, China, the Middle East, and Climate. Each traces how physical geography, resource deposits, and installed infrastructure create constraints that persist across administrations and ideologies. The core contrarian claim: the current energy system, over 80% dependent on hydrocarbons with trillions in embedded infrastructure, will take decades to reshape regardless of stated policy ambition.

Yergin wrote The New Map in 2020. His framework has since proven more relevant than he anticipated, but not primarily for oil. The AI race is fundamentally an energy race. Global datacentre electricity consumption reached 415 TWh in 2024, roughly equivalent to France’s total annual demand, and is projected to double to 945 TWh by 2030, roughly Japan’s¹. U.S. datacentres consumed 183 TWh in 2024, more than 4% of national electricity. American datacentres now draw more power than Pakistan’s 240 million people. The same geopolitical constraints Yergin identified for oil (supply chain dependencies, infrastructure lock-in, territorial control) now apply to chips, power, and the critical minerals that enable both.

In 2019, wind and solar constituted roughly 7% of global electricity generation. A coal plant operates for forty years. A car stays on the road for twelve. An airplane for thirty. These asset lives create what Yergin calls “path dependence”: yesterday’s capital decisions constrain tomorrow’s options. The “energy transition” has so far been an “energy addition”, with renewables layering atop fossil fuels rather than displacing them.

Smil showed us the 50-year rule (Part 1) and the power density constraints (Part 2). Subran calculated the 84% investment gap (Part 3). Yergin adds the geopolitical layer.

Here, Yergin’s insight meets a contemporary reality he did not address: the intertwining of state power and corporate power. The hyperscalers building AI infrastructure (Microsoft, Google, Amazon, Meta) operate at scales that rival national energy systems. Microsoft’s 2024 power purchase agreements exceeded the electricity consumption of several European countries. Yet these companies remain embedded in their home jurisdictions, subject to export controls, reliant on state-backed grid expansion, and increasingly entangled in national security frameworks. Nations do not optimise for carbon reduction. They optimise for energy security, economic growth, and strategic advantage. The hyperscalers that matter are those whose interests align with state objectives, or whose scale makes them impossible to ignore. When these objectives align with decarbonisation, transitions accelerate. When they conflict, transitions stall.

---

The Cartographic Method

Yergin treats geopolitics as literally geographic. This is not a new idea (Halford Mackinder’s heartland thesis dates to 1904), but Yergin updates it for an era where energy flows and digital dependencies intertwine. Each section traces how physical territory, pipeline routes, shipping lanes, and resource deposits create constraints that persist across administrations and ideologies.

Consider the raw positioning in 2020. The United States had become the world’s largest producer of both oil and natural gas, surpassing Saudi Arabia and Russia. Russia remained what Putin calls an “energy superstate” whose budget depended on hydrocarbon exports. China, the world’s largest energy importer, fortified the South China Sea because 80% of its oil transits the Strait of Malacca. Saudi Arabia needed oil at $80 per barrel to balance its budget; Russia could manage at $42. These numbers shaped foreign policy more than any climate communiqué.

---

The Shale Gale and America’s New Map

The American map centres on what Yergin calls “the most dynamic element in the world oil market in recent years”: shale.

Shale is sedimentary rock containing oil and gas trapped in tiny pores, inaccessible through conventional drilling. For decades, energy companies knew it was there but could not extract it economically. The breakthrough came from an unlikely source: George Mitchell, the son of a Greek immigrant goatherd who became a Texas wildcatter. Mitchell spent two decades perfecting hydraulic fracturing (fracking), a technique that injects high-pressure water and sand into shale formations to crack the rock and release hydrocarbons. Combined with horizontal drilling, which allows a single well to access rock formations stretching miles in every direction, Mitchell’s innovation unlocked reserves that had seemed permanently out of reach.

The consequences were immediate and global. By February 2020, U.S. output had reached thirteen million barrels per day, enough to fill 5,700 Olympic swimming pools daily. This abundance changed not just supply curves but America’s geopolitical posture. Energy independence, long a slogan, became something closer to operational reality. Washington could impose sanctions on Russian pipelines because it no longer feared reciprocal energy leverage.

Yet Yergin is clear-eyed about shale’s fragility. The industry’s economics depend on constant drilling to offset rapid decline rates in individual wells. Unlike Saudi or Russian conventional fields, shale wells deplete 70% in their first year. A shale producer must drill continuously just to maintain output, let alone grow it. When COVID collapsed demand in 2020, American producers became casualties alongside OPEC members. The shale revolution provided flexibility, not immunity.

---

Russia’s Pivot and the Malacca Dilemma

Russia’s map is defined by Putin’s drive to restore Great Power status through energy exports. Oil and gas constitute the foundation of Russian influence over Europe and the mechanism for its pivot toward China. The Nord Stream 2 saga, the annexation of Crimea, and the conflict in eastern Ukraine are all, in Yergin’s telling, fundamentally energy stories.

The U.S.-Russian relationship had sunk, Yergin wrote in 2020, “to a level of hostility not seen since Soviet days in the early 1980s.” Yet Russia and China were converging. They shared opposition to American “hegemony” and complementary needs: China required energy; Russia required markets. The Power of Siberia pipeline, delivering Russian gas to China, represents infrastructure that will operate for decades. Once built, it creates its own path dependence.

China’s map presents the mirror image: a rising power whose energy insecurity drives expansion. Yergin introduces a concept that explains much of Chinese foreign policy: the Malacca Dilemma. China imports 75% of its oil, and most of it passes through the Strait of Malacca, a narrow waterway between Malaysia and Indonesia that could be blocked during a conflict. The Belt and Road Initiative, the fortification of artificial islands in the South China Sea, the purchase of African mining assets: these are elements of what Yergin interprets as a systematic effort to reduce this vulnerability. The South China Sea matters to Beijing not primarily because of historical claims or fishing rights, but because one-third of the world’s natural gas shipments pass through it.

By 2019, China produced almost 70% of the world’s solar panels and 70% of photovoltaic cells. For solar wafers, China’s share exceeded 95%. The “Made in China 2025” goal of dominant positions in new industries had already been achieved in green energy. The Malacca Dilemma, ironically, helped drive China’s renewable investments: solar panels do not need to transit chokepoints.

This matters profoundly for European investors: decarbonisation does not eliminate supply chain dependencies. It relocates them.

---

The Transition Paradox

Yergin accepts the scientific consensus on warming without reservation: the IPCC’s findings, he writes, leave “no room for doubt.” But he separates the science of climate change from predictions about the pace of energy transition, arguing that physical, economic, and political constraints make rapid decarbonisation unlikely regardless of stated ambitions.

The world’s energy system cannot be transformed overnight because it consists of massive, long-lived physical infrastructure: power plants, refineries, pipelines, ships, vehicles. A coal plant operates for forty years. A car stays on the road for twelve. An airplane for thirty. These asset lives create path dependence. As we established in Part 1, Smil’s observation holds: “even if we were given free renewable energy, it would be economically unthinkable for nations to abandon the enormous investments already made.”

Consider the global auto fleet. In 2019, EVs constituted less than 3% of new car sales worldwide. In Yergin’s “Rivalry” scenario, the world’s auto fleet grows from 1.4 billion to over 2 billion by 2050. Of that 2 billion, about 610 million are electric vehicles, almost a third. But annual new-car sales represent only 6 to 7% of the total fleet. The world would end up with almost the same number of oil-powered cars on the road in 2050 as today.

“Electric cars are not the end of the oil era,” Fatih Birol, executive director of the IEA, tells Yergin. Cars and light trucks constitute 35% of world oil demand. The rest: heavy trucks, ships, trains, airplanes, petrochemicals. Jet fuel has no obvious substitute at scale. Plastics are embedded in everything from hospital operating rooms to solar panels to the N95 masks that became the emblem of the pandemic.

Yergin coins a phrase that captures his core insight: we are living through “energy addition,” not energy transition. Renewables are growing rapidly, but they layer atop fossil fuels rather than displacing them. Global primary energy consumption continues to rise. Coal use reached a new high in 2023. Oil demand, briefly collapsed by COVID, has recovered. The “transition” is real, but it is slower and messier than the word implies.

---

What Happened Since 2020: The AI Dimension

Energy Constraints Become Compute Constraints

Yergin published The New Map in September 2020. In the five years since, his framework has proven more predictive than even he might have anticipated, not primarily for oil, but for a technology he barely mentions: artificial intelligence.

The IEA’s April 2025 Energy and AI report projects global datacentre electricity consumption will reach 945 TWh by 2030, roughly 3% of total global electricity. U.S. datacentres consumed 183 TWh in 2024, more than 4% of national electricity, projected to grow 133% to 426 TWh by 2030¹. Grid Strategies estimates 60 gigawatts of additional data centre capacity by 2030. 60 GW is roughly Italy’s peak hourly power demand, or the output of 60 large nuclear reactors.

This is not an abstraction for grid planners. In the PJM electricity market stretching from Illinois to North Carolina, data centres accounted for an estimated $9.3 billion price increase in the 2025-26 capacity market. The average residential bill is expected to rise by $18 per month in western Maryland and $16 per month in Ohio. A Carnegie Mellon study estimates data centres could lead to an 8% increase in the average U.S. electricity bill by 2030, potentially exceeding 25% in Northern Virginia.

The power density implications connect directly to Part 2. Data centres operate at 200 to 500 W/m² of building footprint. AI-optimised facilities push toward 1,000 W/m² as GPU density increases. A solar farm at 10 W/m² needs 50 times the land to match a data centre’s footprint consumption. A 500 MW AI campus would require a solar farm covering roughly 50 km², the area of San Marino. This is why Microsoft signed to restart Three Mile Island, Amazon purchased a nuclear-powered campus, and Google announced SMR procurement. Nuclear returns, pulled by the physics of power density.

The New “Big Three”: Chips, Not Barrels

Yergin’s description of the April 2020 oil deal established what he called the “Big Three” configuration: the United States, Russia, and Saudi Arabia coordinating global oil supply. A parallel configuration has emerged for AI compute, but the players differ: the United States controls chip design, Taiwan controls fabrication, and China represents the largest market and most determined competitor.

The dynamics Yergin identified for oil (supply chain vulnerability, weaponised interdependence, the race for self-sufficiency) now apply with equal force to semiconductors. The United States directly controls the design stage: three American companies account for over 75% of advanced chip design. TSMC manufactures 80 to 90% of sub-7nm chips, mainly in Taiwan. High-bandwidth memory (HBM), a critical input for AI chips, comes from two Korean companies (Samsung and SK Hynix) and one American (Micron).

China’s position mirrors its oil dependence: the world’s largest consumer, acutely aware of its vulnerability. Just as the Malacca Dilemma drove Chinese investment in solar panels and overland pipelines, chip dependency is now driving massive investment in domestic semiconductor fabrication. The October 2022 export controls, expanded in December 2024 and again in January 2025, represent the most significant use of technology export restrictions since the Cold War.

Then came the efficiency breakthrough. In January 2025, DeepSeek released R1, an open-source reasoning model that matched OpenAI’s o1 on multiple benchmarks at a fraction of the cost². The immediate consequence: a record 17% drop in Nvidia’s share price on January 27, 2025, approximately $600 billion in lost market value. That single-day loss exceeded the GDP of Sweden. As one former DeepSeek employee explained: “Rather than weakening China’s AI capabilities, the sanctions appear to be driving startups like DeepSeek to innovate in ways that prioritise efficiency.”

Constraint bred innovation. This is Yergin’s framework in action: export controls created pressure, pressure created adaptation, adaptation potentially accelerated the very capabilities the controls were designed to prevent. By December 2025, DeepSeek released V3.2, which the company claimed performed comparably to GPT-5 while cutting inference costs in half.

Hui Huang, writing in Noema in October 2025, frames this competition through an older lens: the Warring States period of ancient China (475-221 BCE), when seven rival kingdoms competed through military innovation, institutional reform, and shifting alliances before Qin’s eventual unification³. In Huang’s telling, the U.S. and China are “essentially accusing each other of being Qin: the hard, efficient, norm-breaking state.” Both are gravitating toward models that prioritise internal control, technological dominance, and strategic advantage over international consensus. The Warring States analogy illuminates what Yergin’s framework implies but does not state explicitly: we may be entering a period where the rules of engagement themselves are contested, where export controls and counter-controls escalate in cycles of pressure and adaptation, and where the eventual equilibrium looks nothing like the cooperative frameworks of the 1990s.

The escalation continues. On June 12, 2026, the U.S. government issued an export control directive ordering Anthropic to suspend all access to its two most capable models, Fable 5 and Mythos 5, for any foreign national, whether inside or outside the United States. The Mythos model family is reportedly capable of detecting software vulnerabilities that have remained undiscovered for decades, a capability deemed too dangerous for foreign access. Because Anthropic cannot reliably distinguish foreign nationals from domestic users in real time, the company disabled both models for all customers worldwide, three days after Fable 5’s public launch. Anthropic disputes the rationale, arguing that the jailbreak cited by officials is narrow and that equivalent capabilities are available from other publicly deployed models. The incident illustrates Yergin’s path dependence in a new register: once export controls are applied to chips, they migrate to the models those chips train, and eventually to the APIs that serve those models. The control surface expands with each iteration.

The Talent Geography

Yergin’s cartographic method (treating geography as determinant of strategic possibility) applies to AI talent with unexpected precision. The San Francisco Bay Area has attracted three-quarters of U.S. AI venture capital funding since 2019. AI-related tech talent in the region grew 24% year-over-year to 76,079 workers in 2025, nearly double New York’s 47,245. In 2025, seven of the top ten largest venture funding rounds went to San Francisco startups, totalling $96 billion, 28% of all venture dollars invested nationwide. That is more than France’s entire annual defence budget flowing into a single metropolitan area’s AI ecosystem.

The concentration is self-reinforcing. Y Combinator’s Garry Tan tells founders they “have to be in San Francisco.” OpenAI, Anthropic, and xAI are headquartered there. AI engineers command $300K+ base salaries; top researchers receive multi-million dollar packages.

For European investors, this geography matters. The talent is not coming to Paris, Berlin, or London at equivalent density. The EU’s AI Continent Action Plan of April 2025 addresses compute infrastructure but overlooks compute efficiency, precisely where DeepSeek has demonstrated that constrained players can compete.

---

Conceptual Toolbox

Concept	Definition	Evidence from Yergin	Implication
Energy addition	Renewables layer atop fossil fuels rather than displacing them	”So far, the energy transition has actually been… ‘the phase of energy addition‘“	AI power demand adds to, not substitutes for, existing load
Path dependence	Yesterday’s infrastructure constrains tomorrow’s options	”A coal plant operates for forty years. A car stays on the road for twelve”	Data centre infrastructure locks in power sources for decades
The Malacca Dilemma	China’s vulnerability to energy imports passing through a single chokepoint	”80% of China’s oil transits the Strait of Malacca”	Explains both South China Sea fortification and renewable investment
The Big Three	Trilateral coordination of strategic resources	”The deal itself was historic… with the United States at the center”	U.S.-Taiwan-China forms the new triangle for chips
Mineral security	Decarbonisation creates new resource dependencies	”China now produces almost 70% of the world’s solar panels”	Chip supply chains create new chokepoints

The interrelations among these concepts form Yergin’s core argument, and now apply to AI with equal force. Energy addition explains why AI power demand adds to existing load. Path dependence explains why data centres pursue nuclear and natural gas rather than waiting for renewables. The Malacca Dilemma explains China’s drive for self-sufficiency in both energy and chips. The Big Three configuration has its parallel in chip geopolitics.

---

Orthogonal Insights

The Developing World Asymmetry

Yergin devotes a chapter to what “energy transition” means for nations where three billion people lack access to clean cooking fuels. The WHO calls indoor air pollution from burning wood, charcoal, and dung “the greatest environmental health risk in the world today.” Close to four million people die annually from its effects, more than malaria, tuberculosis, and HIV/AIDS combined.

For India, where 85% of oil is imported and hundreds of millions live on $1.25 per day, “energy transition” means delivering propane cylinders to 80 million rural households. Nigeria’s petroleum minister put it bluntly: “We’re told we have to move on beyond natural gas. The reality is that Africa is not there yet on renewables.”

This asymmetry now extends to AI. As Part 3 established, the EUR 480 billion annual climate investment gap competes with the capital demands of AI infrastructure. Africa’s ten largest economies need USD 120 billion annually in energy investment by 2030, sixfold the 2020 baseline. That capital cannot simultaneously build AI data centres.

Technology Breakthroughs Required

Yergin collaborated with former U.S. Energy Secretary Ernest Moniz on a study identifying twenty-three technologies with “highest breakthrough potential”: utility-scale batteries, advanced nuclear, hydrogen, carbon capture. Moniz’s assessment was blunt: “We don’t have the technologies for advancing the energy transition to net zero carbon.”

Five years later, some progress is visible. Small modular reactors are moving from concept to procurement. Battery storage has scaled from under 1 GWh to over 100 GWh globally. But the technologies that would allow AI data centres to run entirely on renewable power (storage at multi-week timescales, transmission spanning continents) remain at early stages. The “dark calm” periods that German energy planners fear, extended stretches of low wind and minimal sunlight coinciding with peak demand, apply equally to data centres requiring 24/7 uptime.

---

Closing Note

Yergin’s maps were drawn for oil, gas, and climate. They now describe chips, watts, and the geography of talent. The same logic holds: chokepoints, path dependence, and state-corporate alignment set the pace of change more reliably than policy communiqués.

Energy sets the boundaries. What we build within them remains our own affair.

---

This is the fourth article in a five-part AI × Energy series on the essential books for understanding energy, technology, and investment:

1. Vaclav Smil, Energy and Civilization: A History — The 50-year rule and why energy transitions cannot be rushed

2. Vaclav Smil, Power Density — Why land, not cost, constrains the energy transition

3. Ludovic Subran & Markus Zimmer, The 84% Gap — What climate investment actually requires

4. Daniel Yergin, The New Map (this article) — Energy geopolitics and the realignment of power

5. The Synthesis: Jevons Meets Jensen — The rebound effect and the nuclear revival in the age of AI compute

Footnotes

1. [Source] IEA (2025). Energy and AI. International Energy Agency. https://www.iea.org/reports/energy-and-ai ↩ ↩²

2. [Source] MIT Technology Review (2025). How DeepSeek released a top AI reasoning model despite U.S. sanctions. See also DeepSeek R1 technical report (January 2025). ↩

3. [Source] Huang, H. (2025). Welcome to the New Warring States. Noema. https://www.noemamag.com/welcome-to-the-new-warring-states/ ↩