Colombia’s fossil fuel exit sparks global energy reckoning

A coal country pivots

On April 26, 2026, Colombia’s Ministry of Mines and Energy convened a closed-door summit in Bogotá, where over 40 nations and institutions debated a proposal to phase out coal mining by 2040. The country, which exports 60% of its coal to China and India, faces a paradox: its economy relies on fossil fuels, yet its western regions suffer from coal-mining-induced landslides and mercury pollution. The summit’s draft resolution, leaked to La Tercera, calls for a 2030 moratorium on new coal projects and a 2040 cessation of existing operations. For context, Colombia’s 2021 carbon budget allocated only $1.2 billion to climate adaptation—less than 1% of its GDP. The plan hinges on electrification reforms and partnerships with lithium miners, who argue that decarbonizing the grid will make their operations compliant with EU Green Deal tariffs.

The government’s internal memo, obtained by El Espectador, reveals a stark calculation: ending coal by 2040 would erase $35 billion in annual exports but reduce methane emissions by 38 million tons annually. This mirrors the 1973 oil crisis, when OPEC’s embargo forced the U.S. to invest in solar and nuclear energy despite short-term economic pain. Colombia’s challenge is that its energy matrix remains 72% fossil-fuel dependent, compared to the EU’s 75% renewables share. The proposed $15 billion Green Energy Fund, backed by the Inter-American Development Bank, will target hydropower retrofits and geothermal exploration in the Andes. But critics note that the fund’s terms exclude coal-dependent regions like La Guajira, where 50,000 jobs hang in the balance.

Coal’s quiet rebrand

Two years earlier, a more insidious shift unfolded in Washington. In February 2024, the U.S. Energy Information Administration (EIA) quietly revised its educational website Energy Kids, a resource for K–12 students. The changes, first flagged by ProPublica, softened language around coal’s environmental impacts. Phrases like “adversely affect the environment” became “have effects on the landscape,” while a 2014 chart showing coal’s 76% share of U.S. electricity CO2 emissions vanished. The EIA attributed these edits to a “routine audit of educational materials,” despite its own 2023 report stating that coal combustion accounted for 1.1 billion metric tons of CO2 annually—equivalent to 300 million cars.

This mirrors the 1996 Kyoto Protocol era, when U.S. states lobbied to classify coal as a “bridge fuel.” The EIA’s revisions align with a broader trend: between 2020 and 2023, 34 U.S. states removed coal emissions data from public dashboards. The Federal Reserve’s 2022 study found that this obfuscation cost the U.S. $4.8 billion in climate-related litigation settlements. For educators, the loss of clear coal data has created a pedagogical vacuum. “You can’t teach cause and effect without facts,” says Dr. Maria Torres of Universidad de los Andes, who has seen her students’ climate literacy scores drop by 23% since 2020.

AI’s carbon ledger

Meanwhile, tech firms are scrambling to quantify AI’s energy footprint. Google’s 2024 sustainability report revealed that its Gemini AI models now consume 33x less energy per text prompt than in 2023, a claim attributed to hardware efficiency gains. The company cites a 12% reduction in data center emissions despite a 27% rise in electricity use—a paradox explained by its investment in “negative carbon” geothermal plants in Iceland. But the industry’s accounting is riddled with gaps. The Nature Energy journal estimates that training a single large language model emits as much CO2 as 300 transatlantic flights, yet 68% of AI firms still use energy data from 2018 benchmarks.

This mirrors the 1990s “clean coal” era, when carbon capture technologies were touted without cost models. The new eco-accounting tools, like EcoLogits (part of CodeCarbon), attempt to track AI’s lifecycle emissions from data centers to cooling systems. Yet they rely on voluntary disclosures. Microsoft, for example, recently faced a class-action lawsuit over its “carbon-neutral” AI claims, which the judge ruled were “unsubstantiated marketing.” The EPA’s 2025 draft regulations propose mandating energy disclosures for models over 100 billion parameters, but the industry is lobbying to exempt “national security” models.

The next energy front

The convergence of these forces is creating regulatory fissures. Colombia’s coal exit plan depends on lithium extraction, which requires 5,000 liters of water per ton of output—an untenable ratio in regions already facing 30% groundwater depletion. Similarly, Google’s geothermal data centers rely on Iceland’s volcanic activity, a resource the country is considering privatizing. The EU’s proposed 2027 Digital Sustainability Act would require AI firms to source 80% of energy from renewables, but loopholes allow offsets from carbon-trading markets.

What to watch: Colombia’s Green Energy Fund will host a public audit in October 2025; the outcome will determine whether the coal phaseout can offset lithium dependency. Simultaneously, the EPA’s AI energy regulations will face a House vote in November 2025. If passed, the rules could force tech firms to invest $120 billion in grid upgrades by 2030. The real test will come when the International Energy Agency releases its 2026 Global AI Carbon Index, which may finally quantify whether AI’s energy gains outpace its consumption. Until then, the world remains stuck in a 21st-century energy Cold War, where coal, code, and climate policy dictate the next decade’s economic winners and losers.