# Notes on Catalyst's insider playbook for new nuclear: the layer that jams is behind the reactor Source: Realpha Blog (blog.getrealpha.com) Original article and charts: https://blog.getrealpha.com/en/blog/catalyst-2026-11-12-the-insider-s-playbook-for-new-nuclear-projects/ > In the November 12, 2026 episode of Catalyst with Shayle Kann, 'The insider's playbook for new nuclear projects,' host Shayle Kann and Elementl Power co-founder David Faherty walk every gate a new US nuclear plant has to pass: development capital wants 25-30% annual returns, equity drops to around 11% unlevered after final investment decision, and the cost of capital settles near 10% once the plant runs. Note the vantage point: Kann's fund is an Elementl investor and he sits on the board, and the Ohio project discussed has not broken ground. Educational listening notes, no investment advice, no price targets, and my reading may be wrong. Published: 2026-11-13 Locale: en Tags: Catalyst, podcast-notes, nuclear, energy transition, project development TL;DR: Money for the same nuclear plant demands 25-30% during development and about 10% once it operates; the developer earns the spread for retiring risk one gate at a time. To price a nuclear headline, first ask which gate it sits at. ![A riverside construction pad at dusk: fresh concrete foundations and mats of rebar in the foreground, long shadows from survey stakes, a line of transmission towers stepping away one by one into the river valley and dissolving into blue haze](/covers/catalyst-2026-11-12-the-insider-s-playbook-for-new-nuclear-projects-cover.png) > Building a mound nine fathoms high can fail for want of one last basketful of earth. > > —— *Book of Documents, "Hounds of Lü"* (pre-Qin era; my own translation) In the November 12, 2026 episode of *Catalyst with Shayle Kann*, "The insider's playbook for new nuclear projects," host Shayle Kann asked David Faherty, co-founder of the nuclear developer Elementl Power, to lay out every gate between a bare field and an operating new nuclear plant in the United States. The prices Faherty gave that day: development capital wants 25-30% annual returns and three to five times its money; once a project clears final investment decision with debt and equity in place, unlevered equity returns fall to roughly 11%; by commercial operation the blended cost of capital is around 10%. The same plant asks the market for money at two or three times the price depending on the stage, and the developer's business is the spread from retiring risk gate by gate. Keep the vantage point in view: Kann's fund invested in Elementl and he sits on its board, and the Ohio project Faherty describes has not started construction on a single unit. ![Three bars descending from left to right: the tallest is the 25–30% return demanded during development, the middle one drops to 11%, and the shortest operating-stage bar sits near 10%, so the height gap shows at a glance that capital in the development stage costs two to three times what it costs later.](/figures/nuclear-three-stage-cost-of-capital-en.svg) ## For years I watched nuclear with my eyes on the reactor My filter for nuclear news used to be narrow: whose small modular reactor — the compact, factory-fabricated new generation of designs — looked cleanest, who would get a license from the Nuclear Regulatory Commission first, whose fuel cycle was cleverer. Read enough of it and you arrive at a tidy conclusion: once the technology matures, the plants get built, and the rest is a matter of time. Kann punctured that in his opening. A better reactor does not produce a project by itself. He then listed what somebody still has to do: find a site that can actually host the plant, permit it with the local community on side, sign a power buyer whose credit holds up for decades, get through the interconnection queue and the regulator, assemble vendors, an engineering and construction contractor and an operator, and allocate risk across a capital structure worth billions. In gas, wind, solar, storage and geothermal, one role does all of that. The role is called the developer. In nuclear it is mostly missing. Utilities will not carry the early-stage risk, cloud customers want the electrons, and reactor vendors get pushed into developing their own projects when their job is to finish a product. ![Five power sources each sit above a filled-in developer box, except nuclear on the far right, whose box is an empty dashed outline, so the gap is visible at a glance.](/figures/nuclear-missing-developer-slot-en.svg) ## Assembling a plant into something a pension fund can own On siting, Faherty's order is customer first, technical screen second. The reason is the premium on the first wave: few buyers will pay extra for technology still climbing its learning curve, so find those buyers, then draw your circle on the map. Elementl paid tuition here. Their first attempt was a retiring coal plant — on a river, with transmission and distribution already in place — and diligence turned up too little buildable acreage, a floodplain, and liability for the coal ash sitting on site. They later plugged into Oak Ridge National Laboratory's OR-SAGE, a tool that cuts the country into hundreds of millions of small cells and screens them for power plant siting with geographic data. Conditions split into two tiers: exclusion criteria kill a site outright, such as seismic hazard or a community that does not want you; avoidance criteria get costed, such as soil composition, bedrock, logistics, and distance to the grid. A screen that once took six months and several subject matter experts now returns a first answer from a circle drawn on a map. His most practical aside: the land also has to be for sale, and a utility that knows a site is a crown jewel will keep it. ![A funnel narrowing from top to bottom: hundreds of millions of grid cells across the United States enter at the top, exclusion criteria throw whole blocks of them aside as dead, the survivors pass through avoidance criteria that only add cost, and a small circle of investable sites falls out at the bottom.](/figures/nuclear-siting-two-layer-filter-en.svg) Interconnection was the most counterintuitive stretch of the episode. Other generators complain about queue length because they build fast and cannot wait; nuclear comes online slowly, so a three to five year queue should be digestible. Yet projects have died at this gate. His answer is to keep the project simple. Some developers argue for stacking several generation sources at one interconnection point, and he grants the appeal, but new nuclear already carries enough complexity, and another layer means the regulator and every other stakeholder has to learn one more thing. On customers he reached for liquefied natural gas terminals as the analogue: those projects got financed on creditworthy buyers who put some money in early, plus a long contract. Nuclear needs a harder spec — a 20 to 25 year power purchase agreement with an investment grade counterparty — before Department of Energy lending or private debt will show up. What the customer wants is an easy button: sign, walk away, and have electrons arrive five or seven years later while somebody else handles the middle. Operations stay out of house too. The US has utilities with strong nuclear run records, training pipelines and a reputation the regulator recognizes; financing requires a credible capacity factor — the share of full output a unit actually delivers over a year, and he uses 92% as the benchmark — and half of that number comes from the operator, half from the technology. Technology selection therefore becomes subtraction. They picked GE Vernova Hitachi's BWRX-300: a boiling water reactor, the X standing for tenth generation, fewer components than a pressurized water design, passively safe, and already under construction as a first-of-a-kind at Darlington in Canada. Insurance tightens the choice further — insurance comes before debt, debt before equity, and pressurized and boiling water reactors are known quantities to underwriters, while a new design needs the insurance market educated first. For that new design, beyond the roughly one billion dollars of design and licensing, large pieces of the supply chain may have to be stood up too; Faherty says de-risking technology is not his job and his investors do not expect it of him. A human detail slipped out when he talked fuel: you can walk into the Wilmington plant today and watch the fuel rods being fabricated. Touching them takes certification. The capital formation answer stayed with me the longest. Development carries the most risk and demands the highest return, yet a project investor has no hundred-times upside to compensate for it, so the money jams right there. Elementl signed a preliminary term sheet with a Korean group led by DS Investment Partners and including Samsung C&T for a development capital facility that would fund the development of three grid-scale projects. The key design is portfolio-level exposure: returns are measured across the whole portfolio, so if one project gets paused or killed, those dollars can still earn on the ones behind it. Sometimes, he said, solving a big problem means making it bigger. The structure unlocks other doors at the same time: customers see a funded pipeline, operators see a fleet to run, and the supply chain and trade unions see a career. ![Insurance, debt, and equity form three interlocking links, but the new-technology link on the left is a broken dashed gap, leaving the other two spinning idly when the chain breaks.](/figures/nuclear-insurance-debt-equity-chain-en.svg) For construction he gave one rule above the rest: do not break ground before the design is complete. His Vogtle example is specific. The license application pinned rebar spacing to a tolerance that the field could not hold — the steel expanded in the day's heat and contracted at night, drifting inside what was technically acceptable but outside what the license said — and the decisions that followed went wrong in sequence. The root cause was licensing, the constructor and the vendor never sitting together during design. The same project paid a second price in people: five or six technicians trained for every one retained in the field, with attrition burning skills and cash together. So his order of operations is to pull the constructor in early for constructability reviews, stand up training facilities and trade school pipelines early, and reach a class 2 cost estimate, which requires a finished design plus paid quotes from the supply chain that a lender will underwrite. Barakah in the United Arab Emirates slipped in part because training started too late. The spoiler at the scaling gate surprised me: personal preference. Standardization gets hard when you sell projects to different utilities, because one utility's chief nuclear officer has had trouble with a valve type and the next one has not, and the specs fork. An integrator sitting across several projects is what holds the line, and standardization has to reach past the nuclear island — roughly a third of a project — into the balance of plant and the balance of site. The thickest learning comes from going from one unit to five at the same location, because the local workforce and local suppliers get fluent. ## Where I land now, and what I am unsure of You may have had the same puzzle I did: a nuclear company puts out a big announcement, the stock pops for a day, sometimes it goes the other way. This episode gave me a ruler. Ask which gate the announcement sits at. A letter of intent or a term sheet means someone is willing to talk, with disbursement still waiting on definitive documents and regulatory approval; final investment decision is the one that means money is being spent to build. The boxes left to fill can be named one by one: a 20-year-plus investment grade contract, an operator, insurance, a finished design, an underwritable cost estimate, interconnection. One box filled is worth one box of reaction. ![The same row of six boxes appears twice: in the top row only the first box is filled, standing for a letter of intent, while in the bottom row all six are filled, standing for a final investment decision, so the share of filled boxes is the weight of the announcement.](/figures/nuclear-announcement-how-many-boxes-filled-en.svg) The second shift is where I point my attention along the chain. I used to treat reactor design as the valuable link. Now I ask who turns the next tranche of money into the next milestone, and what they post as collateral. That layer's skill is slicing risk and selling each slice to whoever will carry it, and the barrier is relationships and track record, with patents no help at all. Looking back at reactor vendors through that question, the ones getting pushed into developing their own projects are doing cash-hungry work they are not built for. Three things I am unsure of. Standardization may not hold, and Faherty concedes it is the hardest part while the cost-decline story rests on it. The premium the first customers will pay may not last — the willing payers are AI data centre operators in a hurry, for whom power cost is a small line, and nobody knows whether that urgency survives to 2034. And the incentives inside portfolio-level funding point somewhere worth watching: if returns are measured across the portfolio, a developer feels less pressure to kill a bad project early, and the episode did not discuss what keeps that discipline. ## Further reading - *Catalyst with Shayle Kann*, November 12, 2026: "The insider's playbook for new nuclear projects," with guest David Faherty. - Oak Ridge National Laboratory's OR-SAGE siting screen, which the lab describes as a screening aid rather than a substitute for boots on the ground. - The cost and schedule record of Vogtle units 3 and 4, the source of every construction lesson in this episode. - Progress on the first-of-a-kind GE Vernova Hitachi BWRX-300 at Darlington in Canada. - The Elementl Power and DS Investment Partners preliminary development capital agreement, as a way to see the distance between a term sheet and a final investment decision. ## The one thing to take away One idea: change the things that will change before you break ground. The Vogtle rebar went wrong because a number that five minutes on a drawing could have loosened was carried out to the field instead, where every day costs money and amending that number means reopening a license. That is where the developer role earns its keep: it sells the service of moving hard-to-reverse decisions earlier. Here is something I have tried, offered as an invitation. Pick one thing you plan to start within three months — a job change, a renovation, enrolling in a programme, a conversation you have been putting off — and write three lines on paper: if this goes wrong in six months, which question that I have no answer to today is the likeliest cause; who holds that answer; what date I will go ask. Then put the date in your calendar, and on that day come back to the paper.