--
Paul Gunter, Director
Reactor Oversight Project
Beyond Nuclear
Story by Everett Sloane
• 1d •
4 min read
About 5,000 gallons of tritium-laced wastewater flowed into a storm drain pathway toward the Pacific Ocean at the shuttered San Onofre Nuclear Generating Station on Sept. 23, 2026, after a routine wastewater transfer ran well past its planned stop time. Southern California Edison detected tritium in the water at roughly 3,000 picocuries per liter, and the Nuclear Regulatory Commission logged the release the next day as a non-emergency event.
San Onofre stopped generating electricity more than a decade ago. What happened this week is a decommissioning-site accident, not a reactor incident, and it went to federal regulators through the same event-reporting system that tracks routine problems at retired plants across the country. Tritium, a radioactive form of hydrogen, is a routine byproduct of nuclear plant operations and shows up in low concentrations in groundwater and cooling-system water long after a reactor stops running.
Southern California Edison told the NRC that a planned transfer of mostly groundwater and concrete-cutting water continued between roughly 8 p.m. and 10 p.m. Pacific time on Sept. 23 after it should have stopped, sending the excess into the plant’s Unit 2 storm drain and intake conduit rather than completing the discharge path the company had planned to use. The utility notified the NRC’s event notification system the following afternoon, citing an earlier notice it had already made to the San Diego Regional Water Quality Control Board.
That sequencing matters under federal rules: NRC licensees report certain events to the agency once a separate regulator has already been told, rather than as the first call. Southern California Edison’s report reached the NRC at 2:57 p.m. Pacific time on Sept. 24, roughly a day after the discharge itself, and identified the pathway as leading from the Unit 2 conduit toward the ocean near San Onofre State Beach, on the San Diego County line in San Clemente.
The 3,000 picocuries-per-liter tritium reading is far below the EPA’s 20,000 picocurie-per-liter limit for drinking water, but the discharge broke a separate threshold: its pH measured 10.6, above the permit limit of 9 set for water leaving the site. Total suspended solids came in at 16 milligrams per liter, and results for oil and grease were still pending when the report was filed.
Wastewater leaving a nuclear site into public waters is governed by discharge permits the San Diego Regional Water Quality Control Board issues under the same National Pollutant Discharge Elimination System framework that covers ordinary industrial dischargers, not by NRC radiological limits alone. A pH violation on its own is a permit matter; combined with an off-path release of tritiated water, it drew reports to both agencies at once.
San Onofre’s two working reactors, Units 2 and 3, went online in the 1980s and supplied power alongside an older Unit 1 that dates the site’s operating history back to 1968. Southern California Edison permanently retired the plant on June 7, 2013, after replacement steam generators installed a year earlier developed unusual wear that the utility ultimately judged too costly to fix.
The site has been in an active dismantlement phase, known as DECON, ever since, according to the NRC’s decommissioning tracker for the site. Spent fuel finished moving into an underground dry-storage system on Aug. 6, 2020, and the NRC’s timeline calls for the site to reach unrestricted use by 2028, with full decommissioning — the point at which SCE’s oversight ends entirely — projected for 2051.
“Based on these preliminary analytical results and currently available information, Southern California Edison does not believe the release poses a threat to ocean water quality,” the utility told the NRC in its report on the Sept. 23 discharge. The agency’s own classification — non-emergency — reflects the same read: no evacuation, no public health order, no operating reactor involved.
It is not the only time SCE has had to explain an unplanned wastewater release at San Onofre. A 2020 discharge from the site’s sewage treatment plant, which sent partially treated wastewater more than a mile offshore after dilution problems, drew a sharper company response at the time: SCE called that earlier incident “unacceptable” and said it was reviewing whether its own procedures had failed to prevent it. Six years later, with the plant a third of the way through a 38-year decommissioning that has already outlasted its final decade of power generation, wastewater controls at the site remain the recurring weak point regulators keep coming back to.
Neither agency has said whether the Sept. 23 release will trigger a fine or a corrective-action order. The NRC’s report leaves that determination for later review, and Southern California Edison’s decommissioning timeline continues on the same schedule regardless — spent fuel stays in dry storage on site for decades longer than the reactors themselves ever ran.
This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.
https://iceberg-research.com/2026/09/24/deep-fission-fisn-a-money-pit/
September 24, 2026
Please refer to our disclaimer at the bottom of the report.
Main findings
Background and bull case
Deep Fission (NASDAQ: FISN) is a California-headquartered company founded in 2023. The company is developing a nuclear reactor called Gravity. The reactor is intended to be a miniature version of the conventional pressurized water reactor (PWR), using low-enriched uranium (LEU) as fuel. The company’s key innovation is that its reactor would be buried in a 40-60-inch water-filled borehole one mile (~1.6 km) underground. The reactor is designed with a power generation capacity of 5–15MWe.
Source: Deep Fission
Management claims that its reactor is cheaper to build because its innovative borehole design eliminates certain components required in above-ground PWR reactors. For example, the depth and surrounding geology provide natural containment, potentially eliminating the need for the reinforced concrete and steel containment structures typical of above-ground reactors.
FISN was founded by Richard Muller, a UC Berkeley physics professor, and his daughter Elizabeth, who serve as CTO and CEO of the company, respectively. The company became an unquoted public company in September 2025 through a reverse merger with shell company Surfside Acquisition, before listing on Nasdaq in June 2026.
FISN is participating in the Department of Energy’s (DOE) pilot reactor program. The company is seeking DOE authorization to build a pilot reactor at a site in Parsons, Kansas, by 1H 2027. Then, FISN plans to submit license applications to the Nuclear Regulatory Commission (NRC) to convert the pilot reactor into a commercial facility by the end of 2027, with high-volume deployment targeted to begin in 2028. Because its reactor is based on well-established PWR technology, its licensing pathway would be simpler than for novel advanced reactor designs.
The company claims that its reactor can then be commercially deployed in as little as six months, compared with three to four years for other small modular reactors (SMR).
FISN expects the surge in power demand from AI data centers will drive demand for its reactors and claims the largest pipeline of potential customers, representing 18.5GW of prospective capacity.
It’s time for small modular reactor supporters to admit it: natural gas, not SMRs, is the undisputed winner of the AI data center build-out
In early 2024, the market realized that AI would consume enormous amounts of energy. This kicked off a trading frenzy from conventional nuclear, to gas-turbine manufacturers, to the speculative (and often sketchy) SMR companies. Every player, including the most exotic concept, was supposed to win.
Fast forward two years, and the verdict is in. Gas-fired generation is the clear preferred solution for powering AI infrastructure. Data centers have booked 75GW of onsite generation capacity so far. Almost all of them will rely on natural gas.
Source: Semianalysis
While the largest and most efficient combined-cycle gas turbines face multi-year lead times, data centers have turned to smaller engines or aeroderivatives with lead times as short as twelve months. Gas generation is a proven technology and can meet AI power demand quickly, with the added advantage of cheap natural gas in the US. SMRs, on the other hand, sit at exactly the opposite end of the spectrum: they remain unproven, with commercialization not expected before the 2030s.
The political environment has also become more favorable to fossil-fuel-based generation, removing one of the key roadblocks to new gas development.
|
Google is investing $900M to boost Georgia Power’s nuclear capacity.
The goal: To fund upgrades that will raise the max power output of plants Vogtle and Hatch, adding 96 MW to Georgia's grid. The enhancements will also extend the life of the units and “help Georgia Power customers receive even more value from the existing nuclear fleet,” Joshua Peacock, communications lead at Georgia Power, told us.
A new structure: Google would make monthly payments under a proposed Georgia Power tariff and receive zero-emission credits in return. The value of those credits would be defined in a contract between the utility and data center customers.
The impact on bills: Georgia Power froze its base rates through 2028, and intends to ask the PSC to lower the base rates again in the future, according to Peacock. The utility has been able to keep its rates steady “because of all of the growth and revenue coming in from large energy users like Google,” he said.
Dear Decommissioning Working Group,
“Massachusetts blocks Holtec from discharging Pilgrim water”
Thu, Sep 17, 2026, 10:28PM
Pilgrim, shown here in 2021. (Photo: Holtec)
In 2024, the Massachusetts Department of Environmental Protection (MassDEP) denied Holtec International’s application to release treated water into Cape Cod Bay from the Pilgrim nuclear power plant site as decommissioning work continues.
Holtec appealed that decision, and on Tuesday, MassDEP announced that it has dismissed the appeal, leaving Holtec—for now—with fewer options for how to deal with the up to 1.1 million gallons of water it seeks to discharge.
The context: Pilgrim sits on Cape Cod Bay in Plymouth, Mass. It was shut down in May 2019, and ownership transferred from Entergy to Holtec for decontamination and decommissioning a few months later. At the time, Holtec planned to complete its work at the site within eight years.
In 2021, the company announced that it had placed the last of the plant’s fuel in dry storage. By 2022, all high-level waste was in storage. The ongoing decommissioning progress hit a significant roadblock in February 2024, however, when MassDEP denied Holtec’s request to discharge the plant’s treated wastewater into Cape Code Bay. MassDEP’s decision was made against the backdrop of significant public outcry regarding Holtec’s proposal.
The appeal: Holtec appealed MassDEP’s decision on the grounds of jurisdictional authority and technical veracity. To the former point, the company argued that the Atomic Energy Act bars MassDEP from entirely blocking the discharge of treated water from a nuclear power plant.
To the latter point, Holtec essentially rejected MassDEP’s presupposition that the water at issue is “waste,” as defined by the relevant law—in this case, the state’s Ocean Sanctuaries Act, which defines Cape Cod Bay as an ocean sanctuary subject to special environmental protections. One of those protections is a prohibition on “dumping or discharge of commercial, municipal, domestic or industrial wastes.”
The act defines wastes as “any unwanted, discarded, or environmentally harmful solid, liquid, or gaseous materials resulting from commercial, municipal, domestic, or industrial activities.”
Holtec argued that Pilgrim’s water does not fit into this definition, in that “the treated water to be discharged is not environmentally harmful,” the company said. The appeal further emphasized that the concentrations of the detectable pollutants in the released water would be both well within federal standards and actually “well below the concentrations detectable in the receiving waters.”
Nonetheless, on Tuesday, MassDEP announced that it has dismissed Holtec’s appeal and upheld its 2024 determination denying the company’s request to discharge the water at Pilgrim.
Next steps: The future of Pilgrim’s decommissioning is now somewhat uncertain. As Patrick O’Brien, director of government affairs and communications for Holtec, told Boston NPR station WBUR, the yearslong permitting process has already delayed the cleanup schedule, and the new decision may impose a further delay.
O’Brien added that the company is “reviewing the finalized denial and looking at what options may remain available.” MassDEP said its final decision can be appealed to the state’s Superior Court. Another solution advocated for by some in the state is shipping the water to an out-of-state facility.
Depending on the next steps Holtec takes, this regulatory back-and-forth ultimately may become moot. As deliberations continue, the water at the site is slowly evaporating. However, the rate of evaporation at the site has been variable and difficult to calculate thus far, and it is currently unclear how many years it would take for the remaining water at the site to evaporate.
Dr Ian Fairlie
An independent consultant on radioactivity in the environment.
This article requires an extensive introduction, so bear with me.
Current risk estimates for radiation are based by the ICRP principally on the Japanese survivors of the US bombs dropped on Hiroshima and Nagasaki in 1945. This is the Life Span Study (LSS). However it has several shortcomings including that it is based on instantaneous exposures (lasting less than a second) to high blasts of gamma radiation and neutrons. Other shortcomings were that the studies did not commence until 5 years after the explosions and many unrecorded deaths occurred during the interval. Also the translation of the estimated risks in the acutely-exposed Japanese population to people in different countries exposed to lower doses over lengthier time periods introduces large uncertainties in the LSS risk estimates.
Finally the LSS only covered ~80,000 people. A better, larger study covering ~300,000 workers – was the INWORKS study of nuclear industry workers in the US, UK and France which found radiation risks higher than the risks observed by the Life Span Study but broadly similar to it. (It is explained that the phrase “broadly similar to” means that the uncertainty ranges of the two studies overlapped.) The INWORKS study has been examined and discussed by me previously -see https://www.ianfairlie.org/news/important-new-bmj-article-increases-our-perception-of-radiation-risks/
A third, even larger, study Рthe US Million Person Study- has been underway in the US for several years. For its history, see РBoice Jr JD., Andr̩ Bouville, Lawrence T. Dauer, Ashley P. Golden, Richard Wakeford (2022) Introduction to the special issue on the US Million Person Study of health effects from low-level exposure to radiation. International Journal of Radiation Biology, Volume 98, 2022 РIssue 4.
The quantitative results of this US study have not yet been published formally, it seems, in any academic scientific article. However, rather strangely, an article containing its preliminary findings has been just published in the UK magazine of the UK Society for Radiological Protection. See https://srp-uk.org/professional-resources/radiation-protection-today
I say “rather strangely”, but perhaps it’s unsurprising for a US study to first release in the UK controversial findings on radiation risks, given the US President’s recent attacks on internationally-accepted cornerstones of radiation protection, such as the linear no threshold model of radiation risks and ALARA.
The US Million Person Study of Low-Dose-Rate Health Effects (https://www.millionpersonstudy.org/) is the largest radiation occupational epidemiologic study that I’m aware of. It eventually will cover about a million US nuclear workers and military veterans etc. It is examining cancer and non-cancer morbidity (cases) and mortality (deaths) after exposures to low rates of radiation. It includes nuclear industry workers who predominately received about 300 mGy over their working lives. In other words, it is not a low level study but a low rate study. It includes workers from Los Alamos National Laboratory, Oak Ridge Mound Nuclear Facility, Rocky Flats plant, civilian nuclear utilities, industrial radiographers, medical radiation radiographers, and workers at military nuclear weapons test sites.
Early findings based on ~ 600,000 US workers and US military veterans include the following
These findings are unsurprising and fit in with evidence elsewhere , especially the INWORKS studies. What is significant is that the study runs directly contrary to the views of the US President and can be viewed as a direct riposte to them. So far, no collective quantitative risk estimates from this study have been published in the US.
The Million Person Study continues at large US DoE facilities including Hanford and Oak Ridge, and US Navy personnel, including nuclear submarine sailors and naval shipyard workers. interestingly, is also re-examining the US radium dial workers from the 1920s and 1930s. The ongoing Million Person Study also covers
The SRPS article was written by two US academics, Dr Nicole Martinez and Dr Laurence Dauer. They conclude– somewhat courageously given the US President’s outspoken views –
“A better understanding of potential impacts of radiological exposure enables more informed decisions which in turn can support a more sustainable and prosperous future. See www.millionpersonstudy.org”
