Will Howard retweeted
When we started Elementl, one of the first things we tried to do was buy a coal plant to host a reactor.
It had a grid connection, water and a workforce. It looked obvious. But it wasn’t!
Siting a reactor carries technical requirements we didn’t fully appreciate: seismic, flooding, population, water, transmission. Any one of them can kill a site. Finding that out cost us months and hundreds of thousands of dollars.
So we built a tool. Today we announced Atlas as the first tool in Elementl Lattice, built with @googlecloud on Gemini. It checks hundreds of site characteristics across thousands of square miles, for multiple reactor designs at once.
A first screen that took us months now takes minutes. The full desktop study takes days or weeks. We kill bad sites early and spend real money only on the good ones.
A developer’s job is to buy down risk as quickly and cost-effectively as possible. Atlas is one of many initiatives at Elementl to do that. Meeting this moment takes more than better reactor technology. It takes better processes for siting, licensing, developing and building new nuclear.
We’re not alone in this. Westinghouse is using Google Cloud’s AI on AP1000 construction. NextEra is using it on grid operations. Our customers expect us to bring the best tools available to this work.
Credit to @Rob_Whelan, who leads our AI strategy, and to the technical team at Elementl, whose decades of siting and licensing experience went into the build. Rob is a former submariner and also widely known as the handsomest man in nuclear.
If you have a site you think could host a reactor, talk to Will (@NuclearBill1) on our team.
Will Howard retweeted
Thanks for the continued dialogue.
To clarify my earlier post: I think companies starting from scratch today are too far behind to warrant investment. The teams that recently reached criticality or won Janus awards have a fighting chance to win the military opportunity and to serve remote communities and mines, with two cautions. First, criticality proves the physics, not an operating power plant. Second, I think the market is much smaller, and the fixed capital needed to serve it much larger, than multibillion-dollar valuations imply.
My view would change when an investment-grade customer signs a sizable, firm PPA with upfront capital at risk. Not an MOU, not an LOI, not a preorder. A firm PPA. A customer investing directly in the supply chain would be an even stronger signal.
Rather than compare assumptions, let's take the view of a potential customer: a remote mine running on trucked-in diesel and weighing its alternatives.
- Keep running diesel and buy carbon credits. The gensets are paid for, the crews know them and they're easy to maintain. Low operating risk, though the mine stays exposed to fuel prices.
- Improve efficiency. Heat recovery, load management and better dispatch are the cheapest savings available. Low risk.
- Build a grid line. Usually the best answer when the grid is close, and out of reach for the most remote sites. Low risk where it's feasible.
- Switch to renewable diesel. It drops into the engines the mine already owns and shows a commitment to clean power without new equipment or capital. It can be blended to match what regulators require or what buyers will pay for a greener mineral, so the cost lands on the customers who value it. If the corporate goal is to check the box, it's a solid option. Its premium over regular diesel is the most the mine should pay anyone else to cut the same emissions. Low risk.
- Add wind, solar and batteries. Gold Fields' Agnew mine in Western Australia gets 50–60% of its energy this way, from a microgrid EDL has owned and operated since 2020 under a 10-year contract. Remote sites sometimes have space. Low risk, but rarely a full solution.
- Wait for orbital solar, which venture investors also fund. Overview Energy plans to beam power to existing solar farms and has scheduled its first orbital demonstration for 2028. (I'm being cheeky 🙂, but it's plausible.)
- Sign for a microreactor.
The customer weighs cost, commitment and risk, and the options don't compare neatly. Some decisions are reversible, and some are bridges that put off larger commitments. Site leadership can make some calls locally. Others need board approval, serious planning and capital, and nuclear is one of them.
The customer also has to underwrite to a different standard than a venture investor does.
So what has to be true for a mine to pick a microreactor? It has to be cost-competitive and reliable, and the community has to accept it.
Cost. Start with fuel. Cheaper HALEU TRISO needs large investments in enrichment and fabrication. Some of that is happening, but it isn't clear where all the money will come from or how far costs can fall. The problem is circular. The larger the firm order book, the faster costs come down. X-energy's deal with Amazon shows how it can get done. Amazon anchored the round that funds the first phase of X-energy's TRISO-X fuel plant. But few customers can play that role, and it's hard to know how much capacity Amazon's investment will fund beyond its own needs. The first plant will make enough fuel for up to 11 Xe-100s, and Amazon's own target of more than 5 GW implies about 60.
Then the site. Even a plug-in unit needs a pad, a security perimeter, roads, a hookup to the mine's system, an environmental review and, for some designs, added concrete shielding (5 ft?). Maybe a hole in the ground as well. In dollars, that's far less than an SMR or a large light-water reactor needs. Per megawatt, it's meaningful, because the cost is spread across one megawatt instead of hundreds.
Then operations. How many operators per site? How much security? What will maintenance cost? How will it affect the rest of the site?
Then capital. How much has to be committed upfront, and by whom? At a remote site with one offtaker, it's fair to ask the customer to put in capital or credit support. If it won't, third-party finance will be hard to attract, and the vendor ends up funding more of it. Even at $15 million a unit, a fleet of 100 ties up $1.5 billion.
Reliability. Lost production costs a mine far more than fuel, so the diesel fleet stays as backup and the reactor replaces only fuel. A 90% capacity factor is an assumption that will take years of operation to prove. There's also the risk of problems that only show up over time and force a fleet-wide standdown. When France found stress corrosion cracking, EDF halted about a dozen of its 56 reactors to replace cracked pipes, and its nuclear output fell to the lowest since 1988. For a vendor that owns the fleet, a common defect stops revenue from every unit at once.
Social license. Every host community asks what happens to the fuel. With no US repository, the vendor has to take the reactor back and defuel it, and defueling facilities take capital too. The NRC must certify a package to ship it. Part 71 wasn't written for a whole irradiated reactor, and the agency proposed changes only in July. The route needs NRC approval, with armed escorts through populated areas. Each state and tribe along the way gets advance notice, and many states charge fees. And it's more than one move. The fueled reactor goes back to the vendor, and its spent fuel and activated hardware later need their own trips to storage or disposal. At Radiant's stated 20–30¢/kWh, a 1 MW unit earns roughly $8–12 million over a five-year core, so each $1 million spent moving it takes about a tenth of that.
It's a tough sale. For a mine, or any other customer, choosing a microreactor is a bet on the site's power supply, so the threshold is higher and the tolerance for risk is lower. It also rests on lifecycle assumptions that are hard to validate, and with a reactor this small there isn't much juice to squeeze if any of them miss.
Management teams and boards who have to diligence all of this will ask, "Why should we even take time to look at this for 1 MW of power?"
Many vendors answer with an easy button. They offer to own and operate the reactor, or at least to handle its fuel from start to finish. Delivering that means funding factories and, depending on the company, part of the enrichment buildout, fuel fabrication, spent fuel logistics and defueling facilities. That may be the right answer for customers, but it puts much of the capital on the vendor's balance sheet. The more of it that comes from equity raised at the parent company, the harder today's valuations are to underwrite.
Again, a sizable, firm PPA with the customer's own capital at risk would show that someone has underwritten all of this. That's when multibillion-dollar valuations start to make sense.
Disclosure: I co-founded Elementl Power, an independent nuclear developer.
Even some of nuclear energy's biggest fans cannot figure out why investors like @nucleationvc are so interested in micro reactors. They believe the maths, as my British Commonwealth friends say, do not pencil out.
We believe that they do when the computations begin with a different set of assumptions. One can be exceptionally good at following math rules and still get a wrong answer by starting with an incorrect or incomplete set of assumptions.
Here are a few assumptions I've seen on well-crafted pitch decks that change the final results to the question - "Do micro reactors make economic sense?" We believe incorporating some or all of them into your calculations might change your attitude from dismissal to cautious interest and perhaps even acceptance.
1. Small reactors that produce power in the range of hundreds of kilowatts to tens of megawatts do not compete in the same market as extra large reactors connected to large power grids.
2. In the markets they address, some of which are much larger than many assume, the ONLY current competition is a combustion engine – either diesel or Brayton cycle – that consumes some kind of distillate fuel oil. Just in case you have not been paying attention, the current world price of distillate fuel - at the rock bottom wholesale level - is approaching $5.00/gallon. At 14 kWh/gallon, that's roughly $350/MWh. That is JUST the fuel cost.
3. Regulators have been making huge strides – often with a strong push from legislation enacted by overwhelming majorities – towards adapting rules that recognize the low consequences of events associated with small reactors.
4. Regulators are also enacting rules that recognize that reactors don't have to be tied to a single site, that identical reactors coming from the same factory do not need a completely new license review and that modern methods of risk analysis can help provide adequate assurance of public protection.
5. Micro reactors are not just scaled down large reactors. They have a completely different set of systems, structures and components. If those reactors are designed to be cost competitive, their parts lists, both safety and non safety, are far shorter than those of conventional reactors.
6. Many customers care deeply about cleanliness, CO2 emissions, independence from constant fuel deliveries, dependability and small power generation footprints. They want on-demand power in all seasons and in the worst of weather conditions.
The ways that these assumptions change the maths is challenging to express in a brief post or even a lengthy thread. But rest assured, there are a growing number of well-informed people who have done the calculations with enough detail to become excited.
Some have invested their careers. Others, like the team @nucleationvc, have invested their money and that of their limited partners.
The teams at @RadiantNuclear, @AntaresNuclear, @DeployableE, @AaloAtomics, @LastEnergy, @nano_nuclear, @CphAtomics, @ApolloAtomics, Blykalla and others are not ignorant, ill-informed or crazy. Some of them may not survive in their current form, but they are not chasing a chimera.
PS - In many decks making the case for smaller reactors, I have also seen some poor assumptions.
1. Military customers are not cost sensitive. (I spent 9 years doing military budget work. Cost is far more important than many think.)
2. Future fuel prices can be predicted by simply inflating today's prices.
3. A newly formed team without any members that have been working on new approaches for many years can catch today's leaders.
4. Small reactors can make lots of money selling medical isotopes while they are seeking entry into the power market.
Disclosure: Nucleation Capital is an investor in Radiant, Deployable Energy, Aalo, Blykalla, Copenhagen Atomics and Apollo Atomics. All of them are developing reactors with power ratings below 60 MWe. Accredited investors can join us as we continue finding new companies performing math based on well considered assumptions. Link in reply.
Will Howard retweeted
Here’s a local look at the nuclear project we’re developing in Meigs County, Ohio.
I appreciated Liv Jacobs taking time to chat with us!
thepostathens.com/article/20…
Will Howard retweeted
Building new nuclear in America is a project-delivery problem, not a technology problem. Today Elementl Power announced a preliminary agreement for a development partnership of up to $750 million built to solve it.
The partnership is with a consortium led by DS Investment Partners (DSIP), with Samsung C&T participating as a potential strategic investor. Once finalized, it would be the first private capital facility dedicated exclusively to developing multiple utility-scale U.S. nuclear projects. It would fund permitting, interconnection and long-lead procurement for up to three Elementl projects, including our 1.5 GW project in Ohio, and carry them toward final investment decision.
Development capital only works when technology risk is largely off the table. The @gvhnuclear BWRX-300 builds on decades of boiling water reactor operating experience, runs on commercially available fuel, and is already under construction at Darlington. That lets this capital do one job: make projects ready to finance and build.
Samsung C&T would bring nuclear construction experience and deep access to Korea’s supply chain and engineering bench. American trades will build these plants in Ohio and across our project communities: construction jobs now, operating careers for decades.
That’s what we mean by America’s Nuclear Prime: one repeatable model that brings development, capital and delivery together for successive waves of American nuclear construction.
There’s work ahead to finalize the partnership and deliver these plants. Grateful to DSIP and Samsung C&T, and proud of our team.
bloomberg.com/news/articles/…
Perks of austin is having a SpaceX electrical engineer and a robotics mechanical engineer help fix your oven
Elephant in the room in east austin:
Can’t get data / anything to load since the robotaxis rolled out
Love the product but something is going on - maybe they’re sending the camera footage through the local cell towers?
CC @robotaxi team
1
8
569
to all my austin deep tech / haha hidalgo homies: come check out my new show Tastemakers. It's like The Voice x Shark Tank x Kill Tony. Have a couple founders competing for $ live in Austin. @ Eberly on South Lamar at 6:30. Pull up!
luma.com/tastemakersshow
1
6
415
Will Howard retweeted
And to sprinkle some more Austin flavor onto Deep Tech Week, my nuke pal Will Howard (@NuclearBill1) is hosting a founder pitch comedy show.
Gotta love it, come through.
A Padel39 tourney during Austin Deep Tech Week is a vibe and a half.
Thank you @lacostejonathan @sawsanhaider @CaitlinCloud9 for bringing out the mathletes🙏🏾
Will Howard retweeted
Some refreshing news:
After years of stagnation and decline, nuclear output has reached a new record-high!
Will Howard retweeted
SemiAnalysis: the value of megawatts is so high that a power plant pays for itself in 20 days of inference revenue
"As a quick reminder, the value of megawatts for end-users is skyrocketing. A power plant supporting an islanded 1GW IT datacenter typically costs ~$5B."
"In today's environment, inference API revenue can yield $100B per GW per year, at 90%+ gross margins."
"Paying 2x more money or accepting 30% lower efficiency for faster speed of deployment becomes a no-brainer."
"Said differently, Anthropic and its peers can pay back the value of a power plant in 20 days of inference revenue."
_______
Link to article: newsletter.semianalysis.com/…
SemiAnalysis predicts OpenAI and Anthropic may stop selling API access to their best models
"Given recent security incidents and general worries about increasingly powerful AI, it is extremely likely that frontier model usage will only become more restricted from here."
"In fact, we believe it’s plausible OpenAI/Anthropic will eventually stop offering API access entirely for their most capable models."
"Open source seems like the obvious solution to all these dependency concerns, but it’s far from a silver bullet."
"First, all the “open source” licenses are starting to become increasingly restrictive. As just one example, any “model as a service business” making over $20M a year must negotiate a separate agreement with Moonshot to serve Kimi K3."
"Second, just because a lab open sources their models today doesn’t guarantee they’ll continue doing so in the future."
_____
Link: open.substack.com/pub/semian…
Will Howard retweeted
NextEra has closed a DOE loan of up to $1.9B to restart the 615 MW Duane Arnold nuclear plant, shut since 2020. Target: Q1 2029, pending approvals.
DOE debt helps fund the work upfront. Google’s 25-year power purchase agreement gives NextEra contracted revenue to recover its investment. CIPCO buys the remaining output on the same terms.
Google pays the production costs associated with its power, which NextEra says won’t fall on Iowa energy customers. The loan still needs repaying; electricity prices, financing terms and responsibility for overruns weren’t disclosed.
Cybercab release today?
->Uber “flash promo” to keep up. Golden age of cheap transportation coming to austin!!
6
686
Austin is ground zero: I bet we get <$5 rides for a while cause of this. Seems obvious that competition will drive the costs down to the price of electricity + maintenance + depreciation
2
4
290
Will Howard retweeted
this is unacceptable. @GregAbbott_TX we need your help.
AUSTIN MAN just had his murder charge dismissed by Travis County prosecutors, three years after he killed a man in a robbery.
APD charged Christian Alba, then 17, with capital murder and set bond at $1 million.
By October the capital murder was reduced to plain murder and the bond cut to $390,000. While in jail waiting on trial, Alba and four other inmates cornered a 17-year-old in a hallway at the Del Valle jail and held up a blanket to block the guards' view while they beat him. He was charged with assault and unlawful restraint. Both cases are still pending today, almost three years later.
He bonded out in February 2024. His murder trial in September 2025 ended in a mistrial. This past May, the court removed his curfew entirely.
Then on July 4, 2026, he filled a Nike backpack with 30 boxes of Winchester .300 Blackout ammunition at the Academy Sports on Brodie Lane and walked out. Surveillance caught the ankle monitor on his leg. He was under a court order not to possess firearms, of course. An APD homicide detective recognized him on sight. Arrested July 17, out on bond 2 days later.
Yesterday prosecutors cleared the board. Both murder charges dismissed. Aggravated robbery knocked down to robbery, 7 years probation. Unauthorized use of a vehicle, 180 days.
A man is dead, and the only jail time anyone will serve for any of it is 180 days for a stolen car.
Movies make people fear the wrong thing
Three Mile Island produced zero deaths, a chest X-ray dose, and a 30 freeze on new American reactors.
A stuck valve and a misread gauge caused a partial meltdown on March 28, 1979. “The China Syndrome”, a movie about a cover up at a nuclear plant, released in theaters 12 days earlier, so newsrooms had a field day with this one.
While Thornburgh advised pregnant women and small children only inside a five mile radius, national coverage turned it into a regional evacuation.
Panic eased when Carter walked the control room 4 days later, but the damage was done.
Columbia’s 1990 study of 32,000 nearby residents found no significant cancer increase. Nuclear stayed one of the safest energy sources ever measured. But the movie and the headlines didn’t allow a return to normalcy.
































