The world has invested trillions of dollars in renewable energy over the past twenty years, yet the share of coal, oil, and gas in the global energy mix has not declined significantly. Daniel Yergin, an energy historian and author of The Prize, published an essay in Foreign Affairs online on February 25, 2025, and in the March-April 2025 issue that names this reality without hesitation: what we call “transition” is actually an “addition.” New sources are piling on top of old ones without replacing them.
The Essential Points
- The energy transition has so far functioned as addition rather than substitution, because physical and administrative infrastructure timelines are often long, but can be shortened through reforms, technologies, and better planning.
- Agreements totaling several dozen GW were announced principally in 2024, and network interconnection timelines reach approximately five years according to FERC data for facilities built in 2022.
- Electricity demand from data centers is projected to reach 1,300 TWh by 2035, compared to 430 TWh globally in 2024, according to IEA projections.
- Yergin argues for embraced pragmatism: acknowledging that the transition will be long and nonlinear, and financing adaptation of existing grids in parallel with renewable deployment.
- Recalibrating climate objectives to match actual physical timelines while preserving their mobilizing force remains an open challenge.
Yergin, the Historian Who Measures Before Promising
Daniel Yergin has spent forty years documenting energy systems as they function, not as they should function. The Prize (1991), his history of oil, remains a canonical reference in economics faculties and energy ministries worldwide. The New Map (2020) mapped the geopolitical rivalries born from hydrocarbons. His 2025 essay in Foreign Affairs is shorter, more concentrated, more urgent: it is a warning addressed to decision-makers who confuse political commitments with physical realities.
Yergin writes from Cambridge Energy Research Associates, the firm he leads and which advises both governments and companies. This position sometimes leads to accusations of complacency toward the fossil fuel industry. The accusation deserves to be raised, then set aside: his argument in this essay rests on flow data and infrastructure timelines, not on a defense of oil companies. Readers can verify the figures independently.
The Thesis: Physics Before Politics
Yergin’s argument hinges on a simple observation. When a new energy technology arrives, it does not eliminate the previous one. Coal complemented wood, oil complemented coal, gas complemented oil. Renewables complement the whole. Global energy demand grows fast enough that each new source finds its place without forcing others to disappear.
This addition mechanism is common, but EIA historical data also shows major substitutions between sources, notably the replacement of wood by fossil fuels and then a sharp decline in coal in favor of gas and renewables. And Yergin adds a physical constraint that political calendars systematically ignore: certain infrastructures have lifespans or depreciation periods of several decades, but their construction and integration do not systematically take decades. Network transformation is often lengthy and can exceed ten years, but this timeline is not universally impossible. A nuclear plant takes between eight and twelve years from the investment decision to first electricity delivery to the grid. A hydrogen transport corridor requires a comparable or longer timeline.
Facing this reality, decarbonization targets with a 2030 horizon look less like industrial plans than declarations of intent. Yergin is not saying the effort is useless. He believes that costs, infrastructure, timelines, and permits make the transition more difficult and slower than expected. He also notes that recurring disappointment can erode confidence in institutions that carry climate policy.
Data Centers as a Case Study
The technology sector has made this thesis suddenly visible to a wider public. The IEA reports American agreements representing 26 GW, predominantly small modular reactors, recorded at end of 2024; Microsoft, Google, and Amazon announced major nuclear agreements or investments principally in 2024. This is a considerable sum. It represents, to give an order of magnitude, about ten standard power reactors, or several dozen small modular reactors.
These companies understood that variable renewable energies require flexibility and balancing means to ensure continuous supply: solar and wind are intermittent, and AI algorithms operate twenty-four hours a day. Nuclear provides controllable, decarbonized, dense power. The logic is impeccable. The timeline is problematic.
FERC reported approximately five years of waiting in interconnection queues for facilities built in 2022. These delays come from queues for impact studies, saturation of network engineering capacity, and regulatory procedures. Result: nuclear contracts can leave a transitional supply need for several years. In the short term, supplementary supply can combine gas, renewables, storage, grid, and sometimes coal depending on regions.
According to the IEA’s central scenario in its Energy and AI report, global electricity consumption by data centers would reach approximately 1,200 TWh by 2035, compared to approximately 430 TWh globally in 2024. This progression is considerable, equivalent to the electricity consumption of a large European country being added to existing demand within a decade. As analyzed in an article in the journal on the concentration of power among cloud giants, this accumulation of demand in the hands of a few actors also reshapes power relations on energy markets. Yergin, Orszag, and Arya write that the observed evolution is better described as “energy addition” than as energy transition.
The Essay’s Limitations
Yergin poses a lucid diagnosis. His blind spots deserve to be named.
The first concerns the learning speed of renewable technologies. Solar costs have fallen more than 90 percent in fifteen years. Stationary batteries follow a comparable curve. Some energy economists, notably in the tradition of industrial innovation studies, maintain that scale effects can compress deployment timelines well beyond what historical trajectories suggest. Yergin, precisely because he reasons from long history, may underestimate the rhythm breaks that massive markets and deliberate industrial policies provoke.
The second blind spot is geographic. The addition thesis is truer at the global level than at the national level. Some countries—Denmark, Portugal, Costa Rica—have actually achieved rapid substitutions in their electricity mix, even if their total primary energy consumption remains hybrid. The distinction between electricity and total energy is crucial, and Yergin sometimes treats it too hastily.
The third friction point concerns political agency. Yergin argues for pragmatism, which is right. But a soft version of pragmatism can become a justification for inaction: “timelines are long, so we don’t commit.” Economists like Philippe Aghion, who works on Schumpeterian innovation as a driver of growth, show that clear and stable industrial policy can accelerate creative destruction across entire sectors. The time frame is not neutral: announcing thirty years rather than ten can discourage private investment as much as it makes it realistic.
Two Horizons for One Constraint
Yergin’s essay points to a difficulty that climate policies tend to sidestep: the concrete consequences of unmet ten-year targets.
Two trajectories are emerging. In the first, governments and international institutions explicitly integrate the physical constraint into their timelines. Net-zero targets for 2050 fall within a normative scenario, distinct from exploratory scenarios based on current or announced policies, which allow managed addition financing simultaneous maintenance of existing grids and massive deployment of new capacities. This stance requires publicly accepting that under scenarios of current policies, fossils remain present after 2040 in most advanced economies, and much later in emerging economies where demand is growing sharply. It also requires financing mechanisms that survive electoral cycles: long-term state guarantees, perpetual green bonds, sovereign climate funds.
In the second trajectory, targets remain set at ten years because they serve a political mobilization function. Delays accumulate. UNFCCC reports and global stocktakes document persistent gaps between commitments, implemented policies, and trajectories compatible with climate objectives. According to the essay, shortages, disruptions, or marked energy price increases can provoke backlash against energy and climate policies, which can manifest electorally when energy security and financial accessibility are not assured. This dynamic is already visible in several European countries.
Yergin leans toward the first trajectory without imposing it as certainty. The signals that would distinguish which one is taking hold are observable: the actual gap between decarbonization targets and national trajectories measured year by year, average timelines for bringing new production capacities into service, and the share of fossils in the global mix despite record renewable investment. These indicators show the real speed of change, independent of press releases.
Can a climate policy recalibrated to actual physical timelines retain its mobilizing force? Yergin does not answer directly, but his text suggests that mobilization based on untenable promises is fragile. Mobilization based on an honest diagnosis, coupled with massive and patient investment, would be structurally more robust. This requires political leaders to announce to their constituents that the transformation will take an entire generation and therefore we must act now.
On AI and the energy needs it generates, the physical constraint Yergin describes is all the more pressing because demand is building quickly. The governance questions it raises—who pays for the networks, who decides interconnection priorities, how do we prevent data centers of large tech firms from crowding out other uses—exceed the scope of this essay but are its direct consequence.
Value of the Essay
The Troubled Energy Transition does not take long to read. It is a journal essay, dense and argued, not a book. Its value lies in the formulation: Yergin gives a precise name, addition, to a phenomenon many sense without knowing how to articulate it. And he anchors this name in flow data and infrastructure timelines that are independently verifiable.
For those working in energy, public policy, or finance, this essay offers a framework that recalibrates expectations. For those following AI and data center news, it explains why nuclear contracts signed today will not solve electricity problems in 2027. For those thinking about climate policy, it points to a difficulty to confront: if targets are structurally unattainable within announced timelines, it remains to define what we do with the gap.
Reading Yergin does not lead to fatalism. It leads to a form of uncomfortable honesty that is the condition for serious planning.
Bibliographic Information
Title: The Troubled Energy Transition: How to Find a Pragmatic Path Forward Author: Daniel Yergin Publisher: Foreign Affairs Publication Date: February 25, 2025 (online); March-April 2025 issue (print)
Sources
- Daniel Yergin, “The Troubled Energy Transition: How to Find a Pragmatic Path Forward,” Foreign Affairs, February 25, 2025 (online) and March-April 2025 issue, https://www.foreignaffairs.com/world/troubled-energy-transition-pragmatic-path-forward-daniel-yergin
- Energy Solutions Intelligence, State of SMR 2026, June 2026, https://smrintel.com/state-of-smr-2026/
- International Energy Agency (IEA), data center electricity demand projections, Electricity 2025, data available at iea.org
- FERC (Federal Energy Regulatory Commission), network interconnection queue data, 2026, data available at ferc.gov
