BasicsNuclear power has avoided roughly 1.5 gigatonnes of global CO2 emissions a year since it started running at scale, according to the IEA, and something close to 70 gigatonnes cumulatively since 1971. Those numbers come almost entirely from large conventional reactors built decades ago. The question hanging over small modular reactors is whether they can add a meaningful second act, not just as a cleaner way to keep AI data centers running, but as a tool that reaches parts of the emissions problem that big plants never could.
The honest answer is mixed. Some of the five pathways below are already under construction. Others depend on cost curves that haven’t bent yet. Here’s where SMRs are actually positioned to cut carbon, ranked by how close each one is to real deployment rather than by how good it sounds in a pitch deck.
Retiring coal plants without retiring the emissions math
The most immediate carbon case for SMRs isn’t glamorous. It’s plugging a small reactor into the exact grid interconnection a coal plant is about to vacate. A 2022 Department of Energy study, still the reference document everyone in this space cites, screened 394 coal sites and found 80% of them have the basic characteristics needed to host an advanced reactor. The case study result is the number worth remembering: swapping a large coal plant for a similarly sized nuclear plant on the same site cuts regional greenhouse gas emissions by 86%, the equivalent of pulling more than 500,000 gasoline cars off the road.
TerraPower’s Natrium project in Kemmerer, Wyoming is the clearest real-world test of this idea. The 345-MWe sodium-cooled fast reactor, sited next to a retiring coal unit, began construction this spring after the NRC issued its first commercial construction permit in nearly a decade. Coal-to-nuclear conversion also solves a problem carbon math alone doesn’t capture:
Preserves the existing grid interconnection, cutting years off transmission permitting
Retains a trained energy workforce, since DOE estimates roughly 80% of coal plant jobs transfer directly
Reuses site infrastructure, which DOE says can cut construction costs by up to 35%
Keeps the tax base intact for host counties that would otherwise lose it overnight
None of that guarantees the economics work everywhere. Plenty of the 394 screened sites will stay coal, gas, or nothing at all. But of the five pathways here, this is the one with an actual reactor pouring concrete right now.
Feeding AI’s power habit without adding more gas turbines
This is the pathway getting all the headlines, and for good reason: it’s where the money and the megawatts are moving fastest. On August 4, Florida’s Holtec International signed a non-binding agreement with Louisiana’s Entergy Services and South Korea’s Hyundai Engineering & Construction to evaluate a dual-unit SMR-300 project capable of delivering 680 MW of carbon-free power to Gulf South data centers. NuScale’s separate deal with the Tennessee Valley Authority targets 600 MWe of capacity that the company says would cut emissions by more than 1 million metric tons of CO2 annually compared with an equivalent coal plant.
The climate logic here is straightforward but easy to miss amid the AI hype: without firm, always-on nuclear capacity, most hyperscalers filling the demand gap are turning to natural gas turbines instead, not more solar and batteries. Every megawatt of SMR capacity that displaces a planned gas peaker, rather than simply adding to total generation, is a real emissions win. Every megawatt that just meets demand growth that would have gone unmet otherwise isn’t. Tracking which is which, deal by deal, is exactly the kind of granular work that separates a real climate contribution from a marketing claim, and it’s why SMRbrief Pro tags every announced data center offtake agreement by whether it’s displacing fossil capacity or simply adding new load.
Decarbonizing the heat that renewables can’t touch
Solar panels and wind turbines are good at making electrons. They’re much worse at making the high-temperature steam that refineries, chemical plants, and steel mills need around the clock, which is why industrial heat remains one of the hardest emissions sources to clean up. This is where SMRs have a genuine structural advantage over renewables, not just a cost one.
The clearest current example is X-energy’s planned deployment of four Xe-100 units at Dow Chemical’s Seadrift, Texas facility. Each unit is designed to deliver 80 MW of electricity and 200 MW of process heat, aimed at replacing fossil-fired steam generation and eventually supporting on-site hydrogen production. NuScale is running a parallel effort with Ebara Elliott Energy, using heat exchangers and steam compressors to boost reactor-generated steam to the 500°C range that chemical plants actually need, with field testing targeted for after 2027.
The emissions upside varies a lot by end use:
Chemical sector cogeneration: up to 60% carbon reduction versus gas-fired steam, per recent industry analysis
Steelmaking via nuclear-produced hydrogen: roughly 50% reduction versus a conventional blast furnace, per RMI research, though direct gas-based reduction can achieve similar cuts more cheaply today
Petroleum refining: the single largest near-term opportunity, since refinery steam and hydrogen demand is almost entirely gas-fed right now
This pathway is real but slower than the coal or data center stories. High-temperature designs capable of the 700-900°C heat that primary steelmaking needs aren’t commercially mature yet, and most near-term industrial projects, including Dow’s, are targeting lower-temperature steam rather than the hardest-to-abate processes.
Making renewables’ worst days less carbon-intensive
Solar and wind are cheap when the sun shines and the wind blows. On the days they don’t, grids fall back on whatever’s fastest to ramp, which today usually means natural gas. SMRs don’t compete with renewables for the same hours; they compete with the fossil backup that renewables currently need.
Sweden’s Vattenfall and the industrial consortium Industrikraft, whose members include ABB, SSAB, and Volvo, moved their planned SMR project at the Ringhals site into joint development late last year, shortlisting either five GE Vernova-Hitachi BWRX-300 units or three Rolls-Royce SMR units for a combined 1,500 MW addition. The project exists specifically because Swedish heavy industry needs firm, fossil-free power to hit its own decarbonization targets, and intermittent renewables alone can’t guarantee it.
A few reasons this pathway matters for the emissions math specifically, not just the reliability story:
Every hour a firm nuclear unit covers is an hour a gas peaker doesn’t fire up
Grid operators need less battery overbuild to manage renewable variability when nuclear provides a steady floor
Industrial offtakers with hard decarbonization deadlines, like the Ringhals consortium members, are willing to fund firm capacity that pure-play renewables developers won’t build alone
This is probably the least visible of the five pathways in day-to-day coverage, but it’s arguably the one that determines whether high-renewable grids actually hit net zero or just get most of the way there and stall on the last, hardest hours.
Replacing diesel where the grid doesn’t reach
The smallest pathway in total emissions terms, but the most literal fit for what “small” modular actually means, is replacing diesel generation in places a transmission line will never reach. Russia’s Akademik Lomonosov floating nuclear plant has been supplying power and heat to the remote Arctic port of Pevek since 2020, running on the same reactor technology used in Russia’s icebreaker fleet. Rosatom is now building at least four more floating plants, with one or more slated to power the Baimskaya copper mine in the Chukotka region, a site with no realistic grid connection and no alternative to diesel or heavy fuel oil.
This use case won’t move the needle on global emissions totals the way coal-to-nuclear conversion or AI data center displacement will. But per unit of capacity, it may be the cleanest substitution on this list, since diesel generation is both carbon-intensive and expensive to fuel in remote locations, meaning the economic case and the climate case point in the same direction without much subsidy required.
Which of these five pathways actually delivers the most avoided emissions by 2035 is still an open question, and it’s one worth revisiting project by project rather than taking any single company’s roadmap at face value. The coal conversions and data center deals are the ones with steel in the ground today. The industrial heat and remote diesel plays are the ones worth watching for the next reactor vendor that actually breaks ground rather than just signs another memorandum of understanding.