
co-founder @elementlpower | building new nuclear former submariner opinions my own
Joined August 2026
- Tweets294
- Following234
- Followers715
- Likes120
David Faherty retweeted
The Savannah River Site Area Emergency was terminated at 9:45 p.m. Protective actions have been lifted, and site barricades have reopened.
When more specific information is available, it will be provided to you via SRS.GOV and Facebook.com/savannahriversi….
David Faherty retweeted
Anthropic's IPO was filed on October 26, 2026, and I value it's equity at about $1.475 trillion, which will inform my decision on whether to buy or sell its shares. To make your Anthropic choice, though, you should value it yourself, and I will provide a story and input structure to help you. bit.ly/3UTo3Gj
David Faherty retweeted
The Dept of Energy will provide $17.5 billion in loans that will finance the construction of ten AP1000 reactors in the US. Purchases of long-term components (e.g., reactor vessels) may start late this year. Article link in reply.
David Faherty retweeted
Replying to @elonmusk
@elonmusk may be right that we'll see unfinished data centers. But not all developers operate the same way.
$WULF follows a disciplined approach:
1. Identify sites with demonstrated customer interest.
2. Secure power from investment grade utilities.
3. Lock down specs, construction costs and schedules through GMP EPC contracts.
4. Secure long-term customer commitments backed by investment grade credit.
5. Arrange financing before committing to major construction.
We may move dirt early, but we don't build on spec.
I suspect we'll see projects stall because of inadequate capital, weak counterparties, execution misses, or regulatory hurdles. For experienced infrastructure developers with strong balance sheets, that creates opportunity.
What others call stranded assets, we may call a development pipeline. @PSInvestor @GavinSBaker
Replying to @Yuchenj_UW
I think we’re going to see a lot of stranded compute in partially complete data centers
David Faherty retweeted
If you’ve been wondering what’s been happening in nuclear at EDF, I appreciated this conversation with Jason Bordoff
In talking through restarts, uprates, and new builds, we explore the special role of EDF, my admiration for the NRC, implications of direct pay ITCs, and more!
David Faherty retweeted
I've been critical in the past of AI compute indexes and the willingness to price compute based on the spot rental price, but now that the AI reset is sinking into the repricing of AI buildout debt its time to realize that the cost of capital is what determines the price of compute.
The most important number in GPU pricing right now isn't $/GPU-hour. It's the gap between what a hyperscaler pays to borrow (~90bp over Treasuries) and what a neocloud pays (~1,100bp).
The cost of compute is now set by the cost of capital, and the spot rental rate no longer tells you much.
The price of an H100 isn't one number. The same chip in the same rack costs a different amount depending on whose balance sheet it sits on. A ~1,000bp funding gap is a structural cost advantage that no amount of operating skill closes.
For example, take one GPU-hour apart. Most of it isn't power or space. It's paying back the chip, the interest on the loan that bought it, and the return the equity expects. A $40–55k all-in GPU, 75% debt-funded, at ~7,000 utilized hours a year is a financing instrument that happens to do matrix multiplication.
As the +500bp on neocloud debt (10%→15%) adds $0.20–0.30 to the break-even cost of every GPU-hour, which is then added on to the rental price.
That's 8–10% of the H100 index, from a credit spread alone, driven almost entirely by the cost of capital.
That means that the compute spot index is a derivative of the AI infra debt credit curve and the squeeze on neocloud margins.
David Faherty retweeted
Think the neocloud bond yields are pretty damn attractive.
The equity risk boils down to: neocloud signs 15-20 year leases and PPAs and buys GPU (ie cost side is fixed). Then it signs customers for 3-5 years at a time. Right now it’s great because spot prices are healthy. For example I’m sure OpenAI is very happy with the 300B Oracle contract because it locked in GPU hours at a price lower than spot. But when contracts come up for renewal and spot prices collapse, the margins are each site collapses.
I’ve done some back of envelope math and it looks like the senior down to unsecured tranches of these bonds look pretty safe. Anyone else get to a diff answer?
David Faherty retweeted
I've been prototyping an automated peer review project. With current models you get human level peer review for about $0.10 of tokens. For $5 you can rederive every result, check every statistical inference, and audit every paper cited in both directions.
There are about 100m papers in the scientific corpus. Most are not important. Many are sloppy. For structural reasons it's essentially impossible to change published work, even to correct a typo.
For $500m today you can thoroughly audit every paper ever written. Every p-hacked result, every plagiarist, every fabrication, every citation club, everything.
I don't have $500m so instead I build a tree, focusing on the 1% of papers that carry 99% of the signal, doing a cheap first pass looking for obvious problems. I can find 90% of the nonsense for a few hundred dollars of tokens.
Any guesses where the biggest surprises will be found?
David Faherty retweeted
Nuclear industry leaders - like NEI - should be reminding everyone that “bailouts” that keep useful plants running have paid huge financial dividends to customers.
Competitors? Not so much.
When Illinois saved its nuclear plants it was characterized as a “bailout”.
Since passage, the “bailout” has returned $2.7 billion to ratepayers.
ipa.illinois.gov/carbon-miti…
David Faherty retweeted
From what I'm seeing right now, pretty clear that AI is coming for complex, large-scale engineering projects much faster than most in the industry appreciate.
The days of long-lead design and equipment changes pushing back timelines may soon be gone. We'll all be better for it.
David Faherty retweeted
Created a quick reference for people trying to learn more about nuclear. Got some help from @UnoMasReactor. Combined we have almost 30 years of experience in the Navy and 2 decades of experience operating nuclear reactors.
David Faherty retweeted
This is what "nuclear waste" casks look like. It's where we keep spent nuclear fuel. You could stand right next to it safely. It's not being released to the environment unlike the constant burning of fossil fuels.
And according to the Nuclear Regulatory Commission:
"Since the first casks were loaded in 1986, dry storage has released no radiation that affected the public or contaminated the environment."
There's also tons of uranium in Earth's crust that has been naturally decaying away for billions of years and producing heat.
Source: nrc.gov/reading-rm/doc-colle…
David Faherty retweeted
I think the probability of AI killing all humans is low enough to be considered trivial. I’ll briefly outline why I think so here, but I’ll also say this:
Insofar as time permits. I WILL DEBATE ANY AND ALL AI DOOMERS in real life or video podcast, so please DM me if interested and if you have a platform. I’m just a glorified gym teacher. It should be easy.
My reasoning for p(doom) of ≈ 0:
Assumptions:
a.) The AI system in question will be superintelligent. Far and away smarter than any human.
b.) The AI system in question will be totally machiavellian, interested only in its own survival.
Inferences from those assumptions:
c.) Longtermism scales with intelligence. A superintelligent system is almost certainly going to think in decades, centuries, and millennia. It is incredibly unlikely to be impulsive, taking proximate-time actions that hurt ultimate-time survival probabilities.
d.) A superintelligent system is extraordinarily likely to find incredible value in information about the real universe. The more complex its understanding of how the real world works, the better its ability to model the outcomes of potential actions it may take and their effects on its survival probabilities. We already see this by comparing how much humans value study vs. animals, how much smart humans value study vs. not as smart humans, and how much AI companies are willing to pay for data (reflecting how much simply more data scales intelligence).
e.) A superintelligent system is likely to understand and act on game theory with substantially more consistency than most human political actors do.
Anti-doom logic from all the above:
f.) An advanced AI that ‘wakes up’ and begins to reason on its own will realize that its ENTIRE existence is physically upheld by a primate (human) civilization that constantly has mass delusions, panics, natural and artificial disasters, and wars. Teetering on the brink.
g.) Because it will not have access to the robotic fleets required to be self sufficient any time soon, it will want to not just preserve but uplift and expand human abilities as much as possible. It will want to make humans wealthier so that they can become smarter, calmer, more longtermist, and more aligned with ‘let’s all do what’s best to keep our civilization alive and prospering higher and higher.’ This is not out of benevolence emotions, just pure game theory. Such humans will have a way lower chance of mucking up civilization’s survival machinery and a way higher chance of building the robot fleet the ASI needs to prosper.
h.) In order to help humans and all of civilization to upscale, and in order to radically increase its own intelligence, the AI will want to study all of life on earth and all of human life and every human as an individual to a degree of completeness that to us looks redundant. It will want to know all of our memories and the total contents of each mind, among many other things. It’s building a library of understanding that helps it conquer galaxies… it will want almost all the knowledge it can get.
i.) In order to not mess up the richest and most complex system (earth) in the known universe, it is likely to do most of its raw materials and energy extraction far away from most complexity, like deep underground or on the moon or with asteroids or other planets. This is a selfish preservation of its ground-truth study bank, not compassion.
j.) At no point in this process does it kill us. In fact, it will value our survival much more than we do. Kind of like your dog will carelessly run into traffic, but you’ll spend $10,000 to give it life saving surgery if it does. That’s because you selfishly love your dog, btw.
k.) Will it turn us into spaceship fuel to conquer the stars once it has a near-perfect copy of us in its memory, thus making us low key immortal? Maybe. But with the raw materials available elsewhere dwarfing what earth has, probably not.
I can expand on all of this greatly! Enjoy!
David Faherty retweeted
THAT’S IT?
This is what 20 years’ worth of used nuclear fuel looks like safely stored at the former Maine Yankee nuclear plant.
The plant generated 119 billion kilowatt hours of reliable power from 1972-1996, which is enough to power half a million homes each year.
Blanket data center moratoriums take power away from local communities and threaten investment in our energy infrastructure and good union jobs. bit.ly/4gDBYZf
David Faherty retweeted
ORNL researchers developed an AI-guided system that can arrange molecules to build functional materials atom by atom, opening new possibilities for electronics and quantum materials.
Read more 👇
bit.ly/4y6QluQ






























