Seth Cohen

🇺🇸 industrialist // ⚛️ // “a bull in a china shop” — @secretarywright// Former @DOE @NRC @NASA @DOGE

[Seth Cohen @OverReactor1776]

Now that I’m out of government, I can finally respond for myself: Get bent, soyboy. We didn’t do this for “Silicon Valley . . . companies.” We did this for you, for your family, your community, your state, your nation, and your species.

Nuclear energy provides the safest, highest density, reliable power available on our planet. My career colleagues at DOE and NRC inspired me to think about nuclear as a way to forge American steel and electrolyze aluminum without releasing particulate matter, to desalinate water in the Middle East and save humanity from resource wars. By rejecting the false narratives and Cold War hysteria, we can secure the next American century while raising whole countries out of poverty.

Do you really think I left an incredible career at Kirkland, paid out of pocket for an apartment in DC and dozens of cross-country trips, and left my family on the west coast because I wanted to enrich people I never met before taking this job? I came to D.C. to do something that mattered, to satisfy a driving curiosity (more on that later), and, most importantly, to serve.

As I learned more about nuclear energy and its history, I developed a conviction that one nuclear’s biggest issues was a culture of cynicism: nothing new or exciting could happen because it would end in disappointment, and that militated against rocking the boat even a tiny bit. The career staff in government and their industry counterparts lived through dark winters before and stopped believing that warm springs could bloom into summers.

I have two core philosophies. First, I believe in ruthless optimism. Rational decision making requires detached risk analysis. But we also cannot win if we believe we can lose. Merging the two requires orienting teams around driving missions. That way, when a real opportunity presents itself, you can take a huge swing.

If I take credit for anything—honestly, almost all of the success belongs to the incredible and dedicated people at @ENERGY and @NRCgov—it’s countering the cultural rot and morass that risked forfeiting American excellence. My colleagues and I gave cover to the scientists and engineers, which freed them up to focus on delivering safe power. And, as success materialized, they started to dream again. That’s why the pilot program succeeded, and why I feel confident about the future of NLICs and NRC reform. Nobody needs me anymore because they can innovate on their own.

My second core philosophy is to assume positive intent. Avi, I know that you heard about my real motivations from multiple people you interviewed when preparing your hit piece on me. Rather than telling that story, one which could help inspire another generation of people to use their talents for the greater good, you ignored them. Instead, you implied that Peter Thiel recruited me for nefarious purposes. (I’ve never met him, but, @peterthiel, if you’re reading this, I’m a huge fan!)

Nuclear regulation starts and ends with safety. I promised everyone I worked with that I would resign before doing or pushing for anything that could compromise public safety. But I also distinguished between real safety and performative bullshit. That’s what the careers came to embrace, too. We love nuclear, why would we do anything that could risk threatening its future?

America faces a crossroads. We can either trod a road of cultural decay or hike our way back to the peak of global innovation. Join me on the latter path. Correct the fear mongering and conspiracies and tell the story of America’s great reindustrialization. Tell the story of our public servants, our great entrepreneurs, our scientific dominance. Tell the real story about how DOGE went nuclear.

[Avi Asher-Schapiro @AASchapiro]

Seth Cohen—the DOGE official who was a driving force behind much of the Trump admin's nuclear energy policy—resigned. He was one of the key figures reshaping the regulatory environment to benefit Silicon Valley-backed advanced reactor firms. https://t.co/54gjD0CCPT

[Seth Cohen @OverReactor1776]

@qcmacdonald @dasutherland @MTSlive Sadly I have not had time to play a single round :(

[Seth Cohen @OverReactor1776]

Seth Cohen: Stellaris and deadlifting

[Sheel Mohnot @pitdesi]

Famous hobbies outside work:

Mark Zuckerberg: Brazilian jiu-jitsu, MMA, surfing, hydrofoiling
Bill Gates: Bridge, tennis, pickleball, reading
Steve Jobs: Zen meditation, music, long walks, Japanese art and design
Sergey Brin: trapeze, gymnastics, diving, burning man
Jeff Bezos: Sailing, flying helicopters, ocean exploration
Jensen Huang: Cooking, eating adventurously, was a nationally ranked table tennis player!
Morris Chang: Bridge, classical music, Shakespeare
Warren Buffett: Bridge, ukulele, reading
Elon Musk: Video games, spending hundreds of millions of dollars trying to influence elections

Obsession with your company is great! But the idea that the best founders have no interests outside work is bullshit hustle-porn.

Having hobbies isn’t the opposite of being obsessed with building something great.

[James E. Thorne @DrJStrategy]

The Data Center Buildout Is Not Dead

The Senate’s 73-25 vote to advance the Bipartisan American Affordability and Jobs Act undercuts the claim that rapid data center expansion is politically finished.

With White House support and President Trump publicly backing the measure as a way to unleash energy production and protect ratepayers, the bill both raises costs for operators and clears the largest practical obstacle to new capacity: the inability to build power and transmission on a workable timeline.

Under the legislation, large loads of roughly 20 megawatts or more must cover the full incremental cost of the generation, transmission, and distribution upgrades built to serve them. Ratepayers are shielded from cost shifting, and states gain additional tools to impose collateral or other requirements. Modeling suggests this could meaningfully increase the retail power price faced by data centers and add roughly 10 percent to the annualized cost of a gigawatt-scale facility. Those numbers are real, and some marginal projects will be repriced or delayed.

They are not fatal. The same bill shortens federal environmental reviews, limits the window for legal challenges, strengthens project certainty, and expands federal transmission siting authority after a defined state review period. The binding constraint on data center growth has not been the willingness of capital to fund servers; it has been the multi-year wait for reliable power. Faster permitting of generation and transmission addresses that bottleneck directly.

Higher power costs and clearer cost allocation will push some developers toward co-located generation or more selective site selection. That is an adjustment, not a shutdown. The combination of ratepayer protections, accelerated infrastructure deployment, and explicit Trump administration support is more likely to sustain the buildout than to end it. The political signal is that Washington is prepared to let the projects proceed, provided the incremental costs fall on the operators rather than on everyone else.

[Seth Cohen @OverReactor1776]

@NotASithLord @sthenc Dm me

[Seth Cohen @OverReactor1776]

Who should I meet in Miami?

[Jordan Bramble @jordanbramble]

This was obvious for years for anyone paying close attention.

[Seth Cohen @OverReactor1776]

@cremieuxrecueil Let’s talk!

[Seth Cohen @OverReactor1776]

@AlexCKaufman I once litigated a case about certain precursor chemicals flowing through pipelines. Industrials rock

[Seth Cohen @OverReactor1776]

@dasutherland @MTSlive Sure. As long as there’s a Stellaris LAN party involved.

[Seth Cohen @OverReactor1776]

@dasutherland I tried to explain on @MTSlive that Stellaris’s use of energy as an input credit (pre-Cygnus update. Nothing has caused me more heartburn than my failure to correct the record on the show) is the best mechanism for understanding energy poverty versus abundance.

[Seth Cohen @OverReactor1776]

@whatisnuclear @DougBernauer Boys, boys, there’s no need to fight. I can buy electricity from everyone!

[Seth Cohen @OverReactor1776]

@nukebarbarian @maxbodach @poastmilque @WSJ @mavriaX @JoinFAI Come to the holiday party and we can chat

[Seth Cohen @OverReactor1776]

Thanks for telling our story, @WSJ!

[The Wall Street Journal @WSJ]

Exclusive: Three veterans of President Trump’s Energy Department founded Mavria over the summer to mass-produce data centers, recently hitting a $1 billion valuation after raising $130 million in total funding https://t.co/my2UIJDWLz

[Seth Cohen @OverReactor1776]

@whatisnuclear @WSJ Don’t threaten me with a good time, @whatisnuclear

[Seth Cohen @OverReactor1776]

@jordanbramble @NotMattGeorge

[Seth Cohen @OverReactor1776]

@LSJRobinson @eventidia @mavriaX We’ve learned from the best!

[Lee Robinson @LSJRobinson]

This makes too much sense. @mavriaX is shifting the economies of scale from the production of bits to the production of machines that make bits. There are ~20 great companies using the same thesis on nuclear.

[Seth Cohen @OverReactor1776]

@dr_chrisjones @MTSlive @sophiadew Thanks, Chris!

[Seth Cohen @OverReactor1776]

Productizing data centers will enable us to deliver faster, more consistent compute wherever power is available today. Thanks for having me, @MTSlive @sophiadew

[MTS @MTSlive]

Mavria cofounder @OverReactor1776 warns America could lose its AI lead not because of weaker models, but because it can't build the infrastructure fast enough:

His solution: move data center construction into factories.

"You have a K-shaped deployment model for AI."

"On one hand, you have the world's best researchers constantly releasing frontier technology out of the Valley. And then the deployment of physical infrastructure underneath it, whether that's power transmission or the data center, is significantly slower."

"If we aren't able to keep our lead, the reason isn't going to be because we’re not doing enough on the model side, it's because we’re not doing enough on the physical deployment side."

"One of the key lessons from nuclear is this. You actually can see massive improvements in construction projects by taking all of these different pieces and shrinking them down so that they're more manageable and putting as much of that into the factory as you possibly can."

"We know that by bringing data center construction into the factory, we increase reliability and speed and can also scale the amount of data center capacity that America can deliver pretty significantly."
@mavriaX

[MTS @MTSlive]

Mavria cofounder @OverReactor1776 warns America could lose its AI lead not because of weaker models, but because it can't build the infrastructure fast enough:

His solution: move data center construction into factories.

"You have a K-shaped deployment model for AI."

"On one hand, you have the world's best researchers constantly releasing frontier technology out of the Valley. And then the deployment of physical infrastructure underneath it, whether that's power transmission or the data center, is significantly slower."

"If we aren't able to keep our lead, the reason isn't going to be because we’re not doing enough on the model side, it's because we’re not doing enough on the physical deployment side."

"One of the key lessons from nuclear is this. You actually can see massive improvements in construction projects by taking all of these different pieces and shrinking them down so that they're more manageable and putting as much of that into the factory as you possibly can."

"We know that by bringing data center construction into the factory, we increase reliability and speed and can also scale the amount of data center capacity that America can deliver pretty significantly."
@mavriaX

[Seth Cohen @OverReactor1776]

Thanks for having me, @MTSlive!

[MTS @MTSlive]

America once added 10 GW of nuclear power every year, more than its largest proposed AI data center would need.

Mavria cofounder @OverReactor1776 predicts the US could scale to dozens, if not hundreds, of gigawatts annually over the next decade:

"We have tremendous capability to unlock energy both terrestrially and in orbit. A lot of the things that I feel are threatening America's progress come from short-sighted decisions, some of which are well-intentioned."

"There was a time when America was adding 10 gigawatts a year of nuclear power. The largest data center in the United States that's envisioned is eight gigawatts in Ohio. And we were adding more than that just from nuclear every single year. Think about that."

"I think we are able to harness terawatts worth. Using modern technology and scaling up, I think we could be adding dozens if not hundreds of gigawatts a year over the next decade."

"But we have to want it."
@mavriaX