What would it take for the energy system to stop changing the climate?

Why the question matters

An emissions target names an outcome; it does not name a mechanism. Because nearly all CO2 emissions come from energy use, every target is in practice a statement about how much carbon-free energy must be built and how quickly, which is a far more concrete question than the target itself.

Treating the existing capital stock as given changes the diagnosis. Much of the emissions expected over the coming decades is already embodied in power plants, factories, and vehicles that exist, so the decisive moment is the investment decision rather than the combustion that follows it.

Where emissions are counted determines who appears responsible for them, and a country can reduce its measured emissions simply by importing what it used to manufacture.

What our research finds

• Because tracking systems need substantially more land per unit of capacity, land costs raise their levelized cost of electricity by 20% against 8% for fixed-tilt, so fixed-tilt is cheaper in real Chinese conditions and China installs 18 to 26% more panels for the same electricity output (Chen et al., 2026).
• Electricity that European wind and solar farms currently curtail could make 1.9 million tonnes of hydrogen a year, replacing 30 per cent of the continent's fossil-derived hydrogen and cutting ammonia and refinery emissions by 20 million tonnes of CO2 (Ganter et al., 2025).
• Net-zero scenarios lower overall trade-related energy security risk in 70 per cent of countries by cutting fossil fuel imports, but raise risk to electricity or transport in 82 per cent of the countries that become more dependent on imported materials (Cheng et al., 2025).
• Dietary fats can be synthesized chemically for under 0.8 g CO2-equivalent per kilocalorie against more than 1.5 g for palm oil, with large reductions in land and water use besides (Davis et al., 2023).
• Less than half of projected 2050 hydrogen demand could be produced domestically without running into land or water scarcity, making Africa, South America, Canada and Australia the likely exporters (Tonelli et al., 2023).
• By 2018 the picture had changed: existing infrastructure will emit about 658 GtCO2 if operated as historically, and proposed power plants would add roughly 188 GtCO2 more, jeopardizing the 1.5 °C target (Tong et al., 2019).
• Aviation, long-distance transport, steel, cement, and reliable electricity are the genuinely difficult parts of the system, and addressing them requires integrating sectors that are currently discrete (Davis et al., 2018).
• Deep decarbonization of United States electricity at high reliability requires substantial dispatchable capacity, because daily and seasonal variability of solar, wind, and demand do not coincide (Shaner et al., 2018).
• Building the replacement emits and the old coal plants go on emitting as they retire, so a 40-year transition of 1 TWe away from coal halves the coal-caused warming only after 43 to 53 years for solar photovoltaics, and a switch to natural gas needs a century or more to cut it even by a quarter (Myhrvold and Caldeira, 2012).
• 37 per cent of global CO2 comes from internationally traded fuels and a further 23 per cent is embodied in traded goods, which makes the wellhead, mine mouth, or refinery the point where fewest parties would need to be regulated (Davis et al., 2011).
• If existing fossil infrastructure ran out its lifetime and were then replaced with non-emitting infrastructure, CO2 would stay below 430 ppm and warming would reach about 1.3 °C — so what was already built did not, in 2010, foreclose the targets (Davis et al., 2010).
• In 2004, 23 per cent of global CO2 emissions were traded internationally, with more than 30 per cent of consumption-based emissions imported in several wealthy European countries (Davis and Caldeira, 2010).
• Even a factor-of-three uncertainty in climate sensitivity leaves the conclusion intact — unless sensitivity is low and tolerable change is high, stabilization demands a massive transition to carbon-free energy (Caldeira et al., 2003).
• Carbon-free primary power required by mid-century could be several times the roughly 10^13 watts now supplied by fossil fuels, and every candidate technology surveyed has severe deficiencies at that scale (Hoffert et al., 2002).
• The no-policy IS92a baseline already assumes 10 terawatts of carbon-emission-free power by 2050 — as much as all of today's energy sources combined — and stabilizing CO2 requires tens of terawatts more (Hoffert et al., 1998).