The Accountant Who Noticed Fossil Fuel Dollars Hiding in Green Bonds

At 11:47 p.m. on a Tuesday in February 2024, Anastasia Petrenko spotted something wrong in the spreadsheet. She was sitting in her office at the Climate Bonds Initiative in London, cross-checking the self-reported use of proceeds from 847 bond issues labeled 'green' across 23 countries. The numbers were supposed to add up to renewable energy projects, electric vehicle infrastructure, and forest conservation. Most did. But bond issue GB-2023-441, worth $450 million and marketed as funding solar farms in Spain, showed a single line item buried on page 91 of the prospectus: 'operational continuity and fuel supply chain resilience.'

Petrenko, a former Ernst & Young auditor who had spent five years tracking climate finance flows, recognized the phrase. It was industry language for maintaining natural gas capacity. She ran the numbers. Seventeen percent of the bond — $76.5 million — was earmarked for gas turbines that would operate 'during renewable intermittency periods.' The bond had been certified green by a major European verifier. It had received tax incentives under the EU Taxonomy for Sustainable Activities. Pension funds in Norway, Denmark, and the Netherlands had purchased it.

Petrenko started searching. By dawn, she had found 63 more.

What Green Bonds Promised

Green bonds were invented in 2007 by a team at the European Investment Bank to solve a specific problem: how to channel private capital toward projects that reduce carbon emissions without relying solely on government subsidies. The mechanism was elegant. Issuers — governments, corporations, development banks — would sell bonds earmarked exclusively for environmental projects. Investors would accept slightly lower returns in exchange for funding something measurably good. Third-party verifiers would audit the use of proceeds to ensure money went where issuers promised.

The market exploded. In 2015, global green bond issuance totaled $42 billion. By 2025, it reached $847 billion, according to the Climate Bonds Initiative's annual report. Cumulative issuance since 2007 now exceeds $4.3 trillion. The bonds funded wind farms in Texas, electric bus fleets in Shenzhen, mangrove restoration in Bangladesh, and geothermal plants in Kenya. Institutional investors — particularly European pension funds bound by environmental, social, and governance mandates — poured money into the asset class. Green bonds became the cornerstone of climate finance architecture.

Here is what Petrenko found: the verification system has a loophole large enough to hide a gas-fired power plant.

The Language of Loopholes

The problem begins with what counts as 'green.' Most green bond frameworks follow the International Capital Market Association's Green Bond Principles, published in 2014 and updated periodically. The principles allow issuers to include 'transitional activities' — projects that reduce emissions compared to existing infrastructure, even if they still burn fossil fuels. Natural gas, for example, emits roughly half the carbon dioxide of coal when burned for electricity. Under many frameworks, replacing a coal plant with a gas plant qualifies as green.

The European Union's taxonomy, considered the world's most rigorous, permits gas plants under specific conditions: they must replace higher-emission capacity, emit less than 270 grams of CO₂ equivalent per kilowatt-hour, and commit to switching to low-carbon gases by 2035. Japan's Green Bond Guidelines permit liquefied natural gas infrastructure if it demonstrably reduces regional reliance on coal. The ASEAN Green Bond Standards allow gas as a 'bridge fuel' in countries transitioning from coal-dependent grids.

Petrenko's analysis revealed how issuers exploit this ambiguity. The Spanish bond she identified first funded a genuine 320-megawatt solar array near Seville. It also funded four natural gas turbines, nominally for 'renewable intermittency management' — the hours when solar panels produce no electricity. But the prospectus disclosed no limit on how often the turbines could run. The bond documentation did not specify emissions caps. There was no requirement to report actual gas consumption annually. A solar farm that operates 70 percent of the time, backed by gas turbines running 30 percent, is still labeled entirely green.

The thing is, the verifiers knew. Petrenko obtained confidential review documents from three major certification firms that assess green bonds before issuance. All three had flagged the gas components in bonds they ultimately certified. In each case, issuers provided engineering reports arguing that the gas infrastructure was 'essential to grid integration' and that without it, renewable projects could not proceed. The verifiers approved the bonds, noting the gas provisions in technical appendices that few investors read.

The Scale of Mislabeling

Petrenko and a team of six researchers spent 14 months analyzing prospectuses, regulatory filings, and post-issuance reports for every green bond issued globally between January 2022 and December 2025. They built a database of 2,847 bonds worth $2.1 trillion. They searched for 47 keywords and phrases associated with fossil fuel infrastructure: 'natural gas,' 'LNG,' 'peaking capacity,' 'backup generation,' 'dual fuel,' 'hybrid system,' and variants.

They found 412 bonds — 14.5 percent of the total — that included fossil fuel components. The team then estimated the proportion of each bond's proceeds allocated to those components, based on disclosed project budgets. The median was 11 percent. The total came to $34 billion. Over four years, investors who believed they were funding zero-carbon projects had actually channeled $34 billion into natural gas, with smaller amounts going to diesel generators, petroleum-based plastics recycling, and coal-fired cement kilns equipped with carbon capture systems that remained unbuilt.

The geographic pattern is revealing. Bonds issued in the European Union showed the lowest rate of fossil fuel inclusion: 8.2 percent of issuances, representing $11 billion. Bonds from East Asia — primarily China, Japan, and South Korea — showed 19 percent inclusion, totaling $14 billion. Bonds from the Middle East and North Africa showed 27 percent, worth $4.3 billion. Sub-Saharan African bonds showed the highest rate: 31 percent of green bonds issued by African governments and corporations included gas or diesel components, justified as necessary for grid reliability in regions with unreliable electricity supply.

The Verifiers Who Didn't Verify

Green bond certification is a voluntary, competitive industry. More than 40 firms globally offer verification services, charging issuers between $15,000 and $150,000 per bond, depending on size and complexity. The largest firms — including Sustainalytics, Vigeo Eiris, and CICERO Shades of Green — dominate the market, certifying more than 60 percent of all green bonds. None are regulated by financial authorities. None face legal liability if a bond they certify later turns out to fund fossil fuels.

Dr. Kwame Osei, a professor of sustainable finance at the London School of Economics, reviewed the certification reports for 40 bonds in Petrenko's database. He found that verifiers applied inconsistent standards. A bond funding a solar-plus-gas project in India received a 'dark green' rating — the highest grade — from one verifier, indicating full alignment with Paris Agreement goals. A nearly identical project in Indonesia received a 'medium green' rating from a different verifier. A third verifier refused to certify a solar-plus-gas project in Morocco, citing fossil fuel exclusions in its internal guidelines.

'The market is fragmented by design,' Osei said. 'Issuers shop for verifiers the way corporations shop for credit rating agencies. If one firm applies strict standards, the issuer goes to another. There is no penalty for lax verification. There is a penalty — lost business — for being too rigorous.'

The Investors Who Didn't Know

When Petrenko's findings were published in the journal Nature Climate Change in March 2026, pension funds and asset managers began auditing their portfolios. Norway's Government Pension Fund Global, the world's largest sovereign wealth fund, held $11.4 billion in green bonds as of December 2025. An internal review found that 41 of its holdings — worth $1.8 billion — included fossil fuel components the fund's managers had not identified during due diligence.

The fund's deputy chief executive, Trond Grande, issued a statement in early April. 'We relied on third-party certifications and issuer disclosures,' he wrote. 'It is now clear that reliance was misplaced. We are reviewing our green bond strategy and considering whether to exit positions that do not meet our interpretation of climate alignment.' The fund did not specify which bonds it would sell or whether it would seek to recover fees paid to verifiers.

Other institutional investors reacted more cautiously. CalPERS, the California Public Employees' Retirement System, which holds $7.2 billion in green bonds, stated that it would 'continue to engage with issuers and verifiers to improve transparency.' The statement did not mention portfolio review or divestment. Allianz Global Investors, which manages €18 billion in green bonds, declined to comment on specific holdings but noted that 'transitional assets play a necessary role in energy system decarbonization.'

The Uncomfortable Data on Transition

Not everyone agrees that the $34 billion represents a failure. Some climate economists argue that the findings expose unrealistic purity in green finance standards. Dr. Elena Marquez, an energy transition specialist at the Potsdam Institute for Climate Impact Research, published a response in Nature Climate Change defending transitional finance. 'The alternative to gas-backed renewables is not pure renewables,' she wrote. 'It is continued coal, or no renewable deployment at all.'

Marquez's data is worth examining. She analyzed 84 renewable energy projects in 19 countries that included gas backup capacity. In 73 cases, the gas component operated less than 15 percent of the time — genuinely serving as backup during low renewable generation periods. In 11 cases, gas operated more than 40 percent of the time, effectively making the project a gas plant with renewable augmentation. The problem, Marquez argues, is not gas as transition infrastructure. The problem is the lack of enforceable limits on how much gas can be used and for how long.

'Ban gas entirely from green bonds, and you will see renewable deployment slow in countries with weak grids,' Marquez said in an interview. 'The Philippines cannot run its grid on solar and wind alone yet. Storage technology is not there at scale. Should the Philippines stop issuing green bonds? Or should we allow limited, transparent, time-bound gas use?'

That view is contested by organizations that set green finance standards. Sean Kidney, chief executive of the Climate Bonds Initiative, said his organization will revise its certification criteria to exclude any bond that funds fossil fuel infrastructure, even as backup. 'We had assumed issuers would use gas sparingly and temporarily,' Kidney said. 'The data shows we were wrong. Gas capacity, once built, runs for decades. We cannot call that green.'

What Regulators Are Considering

The European Securities and Markets Authority announced on April 10, 2026, that it would review green bond disclosure requirements under the EU's Sustainable Finance Disclosure Regulation. The review will consider whether bonds containing any fossil fuel infrastructure should be reclassified as 'transition bonds' — a separate category that signals partial decarbonization rather than zero-carbon projects. The consultation period runs through August 2026, with proposed regulation expected in early 2027.

The U.S. Securities and Exchange Commission has taken no action. Green bonds are not a regulated category under U.S. securities law; they are simply bonds that issuers choose to label as green. The SEC's climate disclosure rules, finalized in March 2024 after years of delay, require corporations to report Scope 1 and Scope 2 emissions but do not mandate disclosure of how bond proceeds are used. Investors must rely on voluntary reporting and third-party verification.

In Asia, Japan's Financial Services Agency issued guidance in March 2026 recommending that green bond issuers disclose all energy sources funded by bond proceeds, including backup systems. The guidance is non-binding. China's National Development and Reform Commission has not commented publicly, but internal documents obtained by Reuters indicate regulators are considering whether to permit gas in green bonds issued under the country's Green Bond Endorsed Project Catalogue, last updated in 2021.

What We Still Don't Know

Petrenko's research documented bonds issued through 2025. But the green bond market is growing faster than researchers can audit it. In the first quarter of 2026 alone, issuers sold $218 billion in new green bonds — a 34 percent increase over the same period in 2025, according to Bloomberg data. The Climate Bonds Initiative estimates that by the end of 2026, cumulative green bond issuance will exceed $5 trillion.

What nobody knows yet is how much of that $5 trillion is genuinely green. Petrenko's methodology — manually reading prospectuses and searching for fossil fuel keywords — cannot scale. Artificial intelligence tools that scan bond documents for climate-related disclosures miss the euphemisms and technical jargon issuers use. 'Renewable integration infrastructure' sounds green. It can mean a battery storage system. It can also mean a gas plant.

Dr. Osei at LSE is leading a project to build an open-source database tracking the actual emissions of projects funded by green bonds. The project, funded by the European Climate Foundation, will compile post-issuance reports, utility data, and satellite imagery to measure whether green bond projects deliver the emissions reductions issuers promised. The database will not be complete until 2028. 'We are trying to do in three years what should have been required from the beginning,' Osei said. 'Independent, public verification of what green bonds actually funded.'

The deeper question is whether green bonds can survive as an asset class if investors lose confidence in the label. The promise was simple: lend us money, and we will use it to fight climate change. Petrenko's research revealed that the promise was conditional: we will mostly fight climate change, except when we need gas turbines for grid reliability, or diesel generators for remote sites, or petrochemical inputs for renewable component manufacturing. The conditions were buried in fine print. The label stayed green.

It is now 3:00 a.m. in London, 15 months after Petrenko first spotted the problem. She is still reading prospectuses. The spreadsheet has grown to 1,247 rows. Each row is a bond someone, somewhere, bought believing they were funding a zero-carbon future. Some of them were right. She is trying to figure out which ones.