BasicsCommercial aviation recorded 394 fatalities in 2025, up from 244 in 2024, across 38.7 million flights and eight fatal accidents, according to IATA’s annual safety report. Nuclear power carries a published death rate of 0.03 per terawatt-hour, according to Our World in Data. Set the two next to each other and the headline writes itself. It also falls apart, because one figure counts deaths per flight and the other counts deaths per unit of electricity.
This piece makes the comparison properly, then asks the question that matters to anyone buying or financing a small reactor. Does any operating data show that SMRs are safer than the large plants they are meant to replace? Not yet.
Two safety records that do not share a unit
IATA put the 2025 all-accident rate at 1.32 per million flights, an improvement on 1.42 in 2024. Across 2021 to 2025 the rate was one fatal accident per 5.6 million flights. The International Civil Aviation Organization uses a narrower scope (scheduled operations, aircraft above 5,700 kg) and counted 387 deaths, or 78 per billion passengers carried, and only four fatal accidents. A change in definition halved the number of fatal events. That is worth remembering before anyone quotes a single safety statistic with confidence.
Nuclear’s figure is built differently. Our World in Data puts nuclear at 0.03 deaths per TWh, against 24.6 for coal, 18.4 for oil, 2.8 for gas and 1.3 for hydropower. Its estimate includes modeled deaths from Chernobyl and Fukushima. It also places solar slightly lower, at 0.02, and warns that the uncertainty ranges overlap. Nuclear is not the single safest source. It sits in the same band as wind and solar, and far below anything that burns carbon.
No conversion factor turns deaths per TWh into deaths per flight. The honest reading is that each is near the bottom of its own category, and anyone claiming nuclear is safer than flying is selling a headline. If you see that claim in a pitch deck, check the denominator first.
What Chernobyl and Fukushima add to the count
The two accidents drive the tail of the nuclear number, so they deserve precision. Chernobyl produced 28 deaths from acute radiation syndrome, according to figures from UNSCEAR cited in a 2026 ICRP presentation. UNSCEAR’s 2018 review linked roughly one-quarter of attributed thyroid cancer cases to radiation, as summarized in Our World in Data’s analysis of the Chernobyl and Fukushima death tolls.
Fukushima Daiichi produced zero acute radiation deaths. UNSCEAR reported in 2013 that no radiation-related deaths or acute effects had been observed among nearly 25,000 workers on site. The earthquake and tsunami killed roughly 18,500 people. Mortality among evacuated elderly people in nursing care rose by an estimated 50 to 200 percent, and the ICRP presentation attributes that largely to the evacuation itself.
That last figure is the one SMR developers should care about, and it is an argument about emergency planning, not about radiation. If the harm from a bad day at a plant comes disproportionately from how people are moved, then the size of the evacuation zone is a safety decision with a body count attached.
What the operating SMR record contains
The verifiable operating list is short. Russia’s Akademik Lomonosov floating plant, with two 35 MWe KLT-40S pressurized water reactors, began commercial operation at Pevek on 22 May 2020, under a Rostekhnadzor licence running to 2029. China’s HTR-PM at Shidaowan, two 250 MWt pebble-bed modules driving one 210 MWe turbine, entered commercial operation in December 2023, as World Nuclear News reported. Its operators ran safety tests showing the modules cool down without an emergency core cooling system.
That is a handful of reactors and a few years of service. Some vendor-adjacent explainers say no SMR has ever had a safety incident. The statement is technically available and statistically empty. A fleet this small cannot confirm a safety advantage over plants that have logged decades of operation across hundreds of reactors. What SMRs have is a design case: passive cooling, small cores, factory-built modules. NuScale’s 77 MWe module, for example, sits in a water pool and shuts down without operator action. Those features are plausible and tested in the lab. They are not yet proven in a fleet.
Regulators are betting on the design, one stage at a time
Licensing is where the design case gets scrutinized, and the stages are not interchangeable. Reading the pipeline by milestone:
NuScale’s 77 MWe module received an NRC standard design approval, which confirms the design meets safety requirements but authorizes no plant.
TerraPower’s Natrium sodium-cooled reactor at Kemmerer, Wyoming, won an NRC construction permit in March 2026, and still needs an operating licence before its planned 2030 start.
Rolls-Royce SMR’s 470 MWe pressurized water design was in Step 3 of the UK Generic Design Assessment when last checked, with the Office for Nuclear Regulation still listing regulatory observations to resolve.
Holtec’s 320 MWe SMR-300 received its UK Step 2 statement on 31 March 2026, and the developer has not requested a detailed Step 3.
The NRC’s emergency preparedness rule for SMRs, effective 18 December 2023, lets applicants size the plume exposure zone to the facility. Existing plants carry a 10-mile zone. An SMR applicant still has to justify a smaller one with analysis, and the Union of Concerned Scientists’ Ed Lyman has pointed out that burden stays with the applicant. For a sense of how fast the permitting side is moving, TerraPower’s construction permit announcement is the benchmark so far. If you’re tracking the nuclear market professionally, SMRbrief Pro gives you the structured database to go deeper than any single article can.
Rolls-Royce’s status may have moved since the regulator’s last update, so confirm it against the ONR’s active assessments page before relying on it.
The open question for the next three years is whether regulators and communities will accept scaled-down emergency zones before any Western SMR has operated long enough to test them. Which would you want to see first: a smaller zone on paper, or ten years of operating hours behind it?