Kazakhstan is making a large bet on coal at exactly the moment it has committed to phasing carbon out of its economy. The government's new National Project for the Development of Coal-Based Power Generation lays out plans for roughly 7.6 GW of new coal-fired capacity, framed as a response to a real and immediate power deficit. At the same time, the country's Carbon Neutrality Strategy sets out a path to net-zero emissions by 2060. Read side by side, these two documents describe two different countries.

The instinct is to treat this as a straightforward contradiction: coal versus climate. That framing is too clean, and it obscures the actual policy problem. Whether Kazakhstan's coal buildout is compatible with its climate commitments splits into two genuinely separate questions, and they don't get resolved the same way. The first is whether the country's carbon market, KazETS, can credibly constrain these plants' emissions over multi-decade lifetimes. The second is what happens to local air quality in the meantime, regardless of how the carbon question is answered. Conflating the two is how policy debates on this end up going in circles.

1. What "clean coal" would actually require

The National Project leans on the language of "clean coal technologies," but that phrase is doing more work than it should. Efficiency improvements and conventional pollution controls, better boilers, flue-gas desulfurization, electrostatic precipitators, reduce fuel consumption and cut conventional pollutants. They do not meaningfully reduce CO2 output. The only measures that actually move the carbon needle are a narrow set: carbon capture and storage, co-firing with biomass or hydrogen at meaningful scale, or simply running less coal capacity than currently planned. None of these are cheap. CCS appears only in Kazakhstan's separate Carbon Neutrality Strategy, as a long-term aspiration for plants still running past 2035, with no binding regulatory framework.

So the first thing to be clear-eyed about is that "clean coal" as used in Kazakhstan's current policy language is mostly a conventional-pollution story, not a carbon story. That distinction matters because it means the National Project, as written, doesn't actually answer the question its own name implies. It modernizes coal. It doesn't decarbonize it.

2. Can KazETS actually hold the line?

That leaves KazETS to do the carbon work, and the mechanics of how it does that work are where the real risk sits. Launched in 2013 as the first emissions trading system in the CIS and Central Asia, KazETS now covers roughly 47% of national CO2 emissions across some 200-225 installations in power, centralized heating, extractive industries, and manufacturing. Allowances are currently allocated free of charge, not auctioned, based on historical emissions-intensity benchmarks by sector. The system moved into its sixth operating phase in 2026, following a fifth phase that ran 2022-2025. A World Bank-backed reform program, running through 2028, is working to enforce narrower benchmarks, the introduction of auctioning, and broader sectoral coverage.

But reform of the allocation formula doesn't solve the deeper structural problem, which is about time horizons rather than mechanics. Power plants commissioned today will operate for thirty to forty years. Kazakhstan's carbon neutrality target is 2060. That means the allowance trajectory has to do something allowance systems are generally bad at: commit credibly, today, to a tightening path that outlives several changes of government, several energy ministers, and at least one full commodity cycle in coal and gas prices. If early allocations stay generous enough to keep the new coal buildout economically comfortable at commissioning, and tightening is deferred to "later" without a binding, dated schedule, the risk isn't hypothetical. It's the standard mechanism of carbon lock-in: once the capital is sunk into a plant with a forty-year design life, the operator's incentive is to keep running it as cheaply as possible, and retrofitting or early retirement becomes a fight against sunk-cost economics regardless of what the allowance schedule says on paper today.

There's a second layer to this. Emissions-intensity benchmarking, if calibrated to the average of the existing coal fleet, effectively grandfathers the sector's current performance rather than pushing it toward the plants' theoretical best available technology. That's a reasonable transitional design choice, but if the benchmark isn't tightened on a pre-committed schedule tied to specific years rather than "future review," it becomes another way lock-in gets built into the system quietly, through the benchmark rather than through the allocation volume.

A well-designed, well-enforced KazETS is answering a CO2 question, on a decades-long clock. It has essentially nothing to say about the second problem, which operates on a completely different clock.

3. The problem KazETS doesn't touch: local air quality

Whatever happens with carbon accounting, these plants will burn coal, and burning coal produces particulate matter and other conventional pollutants that are a local, immediate public health issue, not a fifty-year accounting exercise. Karaganda is the clearest case study, and it's worth being precise about what's actually driving it: ranked the world's third most polluted city in 2024, with PM2.5 levels of 104.8 µg/m³[1], its emissions are dominated not by power generation but by metallurgy across the wider Karaganda industrial region, its largest named source, split between the Qarmet steel complex in Temirtau and Kazakhmys Corporation's copper mining and smelting operations. Thermal power and district heating plants make up a meaningful share of the remainder, alongside an estimated 25,000 residential coal stoves. This is a live, present-tense problem independent of how the carbon-neutrality debate resolves, and it needs its own regulatory answer, on its own timeline, one that reckons honestly with what's actually emitting.

That emissions mix matters for which trading design could plausibly work, and it's worth testing against the two conditions any cap-and-trade system actually needs: heterogeneous emitters with genuinely different abatement costs, so there's a real incentive to trade rather than a market of identical participants, and physically measurable point sources, so compliance can be verified rather than assumed. Karaganda's mix is a mixed result against both. On heterogeneity, steel, copper, and combustion are different enough industries that their abatement cost curves plausibly diverge more than a cluster of similar plants would, which is a genuine point in favor of tradability. On measurability, Qarmet and Kazakhmys are large, monitorable point sources, but the 25,000 residential stoves are exactly the diffuse, unverifiable source this condition is built to exclude. Any trading design proposed below covers the industrial emitters. None of them touch the stoves, which becomes the largest limitation.

There are three ways to approach the industrial share of the problem, and they all trade off differently.

Option A: Direct emissions standards

Mandatory scrubbers, electrostatic precipitators, and hard per-facility particulate limits, enforced through straightforward monitoring. The advantage is predictability: every facility knows the rule, compliance is binary, and enforcement doesn't require a functioning market or sophisticated monitoring infrastructure to work correctly. The disadvantage is inflexibility and, typically, a higher aggregate compliance cost, since every facility has to hit the same bar regardless of its specific abatement cost. For a mix as different as Karaganda's, an integrated steel complex, a copper mining and smelting operation, and a set of thermal plants, a uniform standard can mean some operators overcomply expensively relative to their actual abatement cost while others meet the letter of the rule with minimum effort, precisely the inefficiency a trading system is designed to avoid.

Option B: A localized particulate-matter market for Karaganda

I encountered the working precedent for this directly, through a climate and energy fellowship with the University of Chicago’s Institute of Climate and Sustainable Growth this past summer, where researchers involved in emissions-market design walked me through the architecture behind India's Surat Emissions Trading Scheme, developed by the University of Chicago's Energy Policy Institute, J-PAL, and the Gujarat Pollution Control Board. Launched in 2019, it started with 156 randomly enrolled industrial plants and has since scaled to 298, with a second market covering 120 more plants added in Ahmedabad in 2023. It pairs mandatory continuous emissions monitoring at each plant with a cap-and-trade permit system for particulate matter, allowing lower-cost abaters to sell excess permits to higher-cost ones. The results, evaluated by randomized controlled trial and published in the Quarterly Journal of Economics in 2025, are concrete: a 20-30% reduction in particulate emissions, more than 10% lower abatement costs, and 99% compliance, figures that make a far stronger case than a generic claim that trading "lowers costs."

The adaptation challenge is real, though. Surat's regulated units are textile-dyeing and printing plants, numerous, relatively small, and fairly similar to each other, which is precisely the kind of homogeneous-but-still-tradable base that makes monitoring straightforward. Karaganda's industrial emitters are the opposite in composition: a small number of very large, very different facilities, an integrated steel complex and a copper mining and smelting operation, each with its own emissions profile, abatement technology, and cost structure. That heterogeneity is arguably good for the "good guy, bad guy" trading logic, but it means the market design, permit denomination, and monitoring protocol would need to be built specifically for two or three large, dissimilar heavy-industry emitters, not adapted from a cluster of many small, similar textile plants. It is a smaller, more concentrated market than Surat's, which cuts both ways: easier to monitor comprehensively, harder to generate real liquidity with so few participants.

Option C: A broader, zone-based trading system

Modeled loosely on California's RECLAIM program, which covered nitrogen oxide and sulfur oxide trading across the Los Angeles basin starting in the 1990s. RECLAIM divided the basin into two zones, coastal and inland, and restricted coastal facilities to buying permits only from other coastal facilities, an explicit response to concerns that trading could otherwise let pollution concentrate in the more exposed, densely populated areas along the coast[2].

But it's a cautionary precedent, not a clean success story. The clearest documented failure is price volatility: Fowlie, Holland, and Mansur's study of the program found that NOx credit prices rose from roughly $2,000 per ton in January 2000 to more than $120,000 per ton by March 2001, driven by the California electricity crisis, a scale of instability that a smaller, thinner Kazakh market would be even more exposed to. The equity picture, by contrast, is more reassuring: the same peer-reviewed analysis found that emissions changes at regulated facilities did not vary significantly with neighborhood income or demographics. The honest lesson isn't that zone-based trading is inherently risky for equity outcomes. It's that this result wasn't automatic, RECLAIM's zone structure was deliberately designed to prevent pollution from concentrating in the more exposed coastal areas, and the empirical finding suggests that design choice worked, which is itself the argument for building similar safeguards into any Kazakh version rather than assuming a market will sort itself out fairly.

A Kazakh version of this approach would need geographically tighter zones than RECLAIM's basin-wide districts, exchange ratios that actively discourage trading pollution toward disadvantaged residential cores rather than merely tracking it, permit pricing explicitly weighted by population exposure, and price-stability mechanisms that RECLAIM itself lacked during its worst period. Those are all correctable design choices. The RECLAIM experience is useful precisely because both its clear failure (price volatility) and its contested outcome (equity) show what happens when they aren't addressed explicitly upfront.

None of these three is "the" answer for regulating particulate matter in Kazakhstan. The point is that this is a real, consequential design choice Kazakhstan should make deliberately, rather than one that gets resolved as an afterthought of carbon policy.

4. If the coal buildout goes ahead, renewables need to come with it, on the right terms

There's a fourth piece that follows directly from the first three, and it concerns what happens to renewable energy investment if the new coal capacity gets built as planned. If new coal generation moves forward, Kazakhstan needs a corresponding, deliberate push into renewables and the flexible or balancing capacity, storage, demand response, and grid-forming inverter technology that lets high shares of wind and solar actually displace traditional generation in the mix over time, rather than simply sitting alongside it as an addition to total capacity.

That only works, though, if the financial incentives are structured correctly. Carbon offsets generated by renewable energy projects are, functionally, a financial asset belonging to whoever generated them. If those offsets end up effectively absorbed by coal-based stations, used to offset the new plants' own emissions rather than retained and monetized by the renewable projects that generated them, renewable investment loses one of its clearest and most durable financial returns. The alternative is straightforward: offset ownership should sit with the renewable project and its investors, who can then choose to sell into KazETS directly or into a voluntary carbon market as one develops. That is what makes renewable investment attractive on its own underlying economics, rather than dependent indefinitely on subsidy, and it's what keeps open the possibility that renewables actually erode coal's share of the generation mix over time, rather than simply coexisting alongside an expanding coal base indefinitely.

5. Three questions, three frameworks

The debate over Kazakhstan's new coal generation shouldn't be framed as a binary choice between coal and decarbonization, because that framing collapses three problems that don't share a single solution or a single timeline.

The long-term question, whether new coal capacity is compatible with carbon neutrality, depends on KazETS allocation and benchmark design holding a credible tightening trajectory across a multi-decade plant lifetime, and on Kazakhstan being honest now about where lock-in risk actually sits in that system. The immediate question, what these plants and their industrial neighbors do to local air quality starting today, depends on a separate regulatory choice among direct standards, a Surat-style trading scheme for large point-source emitters, or a RECLAIM-style zone-based market, each with a different cost, speed, and enforcement profile, and none of which reaches the diffuse residential sources that make up a real share of the problem regardless of which is chosen. And running underneath both is a third condition: renewable investment needs its offset value explicitly protected, not quietly transferred to the coal plants it's supposed to be competing against and, eventually, displacing.

Kazakhstan doesn't have to choose between its coal buildout and its climate strategy in the abstract. It has to build three separate policy frameworks, on three different timelines, that don't quietly undermine each other in practice. That's a harder and more specific task than the current framing suggests, but it's also a solvable one, provided each piece gets designed on its own terms rather than assumed to fall out automatically from the other two.