Nuclear, AI, and the fragility of markets

How real are all these announcements?

Team Coral is extremely pro-nuclear. Our 3rd post ever covered the Z-pinch innovation of Uri Shumlak at the University of Washington, the commercialization of which is the basis for the fascinating Zap Energy. We even proposed a GTM Framework for fusion, in what was the most widely-read post from year one of Coral, and one of our favorites to write. Among the core tenets of our writing and thinking is that degrowth is, to put it mildly, really dumb. Fully scaled, climate tech is the lever to create abundance across the world. This is going to require a lot of energy, delivered reliably and without emissions. Fission and fusion are the most obvious technologies to meet that need. 

All of which is to say that it brings us no joy, none whatsoever, to report that we’re getting really, really worried about the potential of either flavor of nuclear power as a public good and means of stabilizing our wobbly electric grid. 

You’ve probably seen the headlines; Microsoft signed an agreement to help restart reactor one at 3 Mile Island, and buy all of the electricity produced for the next several decades. Amazon is building a data center right next to a nuclear plant in Pennsylvania. Google contracted to buy power from a startup called Kairos, which will have small modular reactors online by 2030, unless of course they don’t. Credit where due; We applaud these companies for altruistically investing in renewable energy resources for the betterment of mankind engaging in an elaborate marketing exercise around buying energy for their AI data centers. 

And that’s what it is. Go back and read the second paragraph of the Google announcement linked above. They’re not even pretending. While most of the purported AI on the market is neither innovative nor useful, we’re not ones to begrudge commercial enterprises trying to make a buck. The issue here is in incentives and timelines. Demand for AI is growing really, really fast. The capacity to produce nuclear energy at sufficient scale to power the same is… not. The US sucks at building things. Fission is a mature, well understood technology, with 60 or so reactors under construction all over the world. None are in the US. Fusion, while theoretically a better technology, remains theoretical. The most optimistic estimates, primarily from fusion startups themselves, have these companies reaching interconnection in the mid 2030s. It will likely be later than that.

It is not in the nature of big companies, subject to quarterly earnings reports and analyst glares, to be patient with the development of new tech in order to power their highly profitable product lines. The data centers are still going to get built and switched on, and we’d bet virtually any amount of money that between few and none of them will be powered by the energy sources written about in press releases. The timelines simply don’t match. We are extremely interested in the specifics of the contracts. How many of them have out-clauses based on strict milestones? Few things are more devastating for a nascent industry than a big influx of promised capital, followed by a matching outflow before profits can be realized. 

Not too long ago, a number of big tech companies (hi Microsoft!) announced substantial investments in and commitment to purchasing carbon offsets via ocean-based carbon capture, which naturally led to an explosion of startups building to meet the advertised demand. It went well, and by well we mean that it is currently slightly easier to raise capital for a nascent Quidditch league than a VCM startup. We don’t begrudge the tech BigCos making money from AI, but they could perhaps learn to do no harm, and to make sure that their tech purchases are run out of technical teams, rather than by marketing or BD chuckleheads. 

So that’s the demand side, what about production?

In recent news, a 42 person startup which admits to being 15+ years from interconnection somehow raised $900m to do…. Something. Commonwealth Fusion Systems raised $1.8bn in funding last year, has lovely offices and conducts many heavily photographed events with local politicians, and has been 5 years away for the last decade. Perhaps the highest-profile fission startup, Oklo, went public earlier this year via SPAC (deep sigh), and counts Sam “AI” Altman as the chairman of the board (deeper sigh). If you’re wondering how these companies manage to attract such sophisticated investors, we suggest googling the term “secondary liquidity.” Paper markups are not difficult to monetize with the right network. 

We are not suggesting that these companies are wholly grifts, only partially. They have formidable technical talent, legitimate innovations in many aspects of their technology, and a path to make a lot of money in the fullness of time. Some are likely to reach technical viability and eventually power our homes. We hope they do. That said, their valuations are insane ($900m for 42 people!), the technical path is much longer than they wish to admit, and they are deeply vulnerable to regulatory disruption and market slowdowns. 

America doesn’t suck at building due to lack of brains, brawn, or willpower. We can’t build because the people regulating energy and the grid lack the knowledge and nuance to do so effectively, and because the regulations surrounding interconnection are slightly more byzantine than a Terry Gilliam bureaucracy. For more on these headaches through a nuclear lens, we wholeheartedly recommend the fantastic Titans of Nuclear podcast.

Do nuclear startups care? Perhaps not. Bloated funding rounds allow for high salaries, the secondaries market creates early liquidity, and very few people can even judge their achievement of milestones. It’s a great business if you can get it. Well, great for the startups, not so much for the public. 

It is tragic that nuclear power ever became controversial, but it did. We wrote about why that happened here, but tl/dr: Sensationalist news coverage of the very, very few safety incidents with older generations of reactors, fear mongering politicians looking for wedge issues, and lack of public knowledge around the fact that fission in particular is perfectly scalable and perfectly safe. We need education, pressure on lawmakers to pull their heads out of their asses, and the will to build things. 

Remember when there were huge tranches of money flooding into tech categories like one-click checkout, last-mile delivery, micro transport, and so much more? The investors and builders in those spaces forgot to develop real business models and solve regulatory hurdles, the tide of money went out, the companies crashed, and the public decided to ignore all of the above. Which is fine, honestly. We look forward to never having to step around an oversized Lime bike on the sidewalk again. Given the rush of dumb money, absurd valuations, long feedback loops, and misaligned incentives, we see the same dynamic playing out in this market. 

That sound you hear is the 2029 wave of shutdowns of nuclear startups, the public deciding the tech is unviable, and fossil fuel energy plants revving back up. This will be devastating. Nuclear has been so feared, for so long, that any further damage to public sentiment will likely be permanent, and result in the next round of moronic politicians promising to ban our best chance at energy for all. We’re typically capitalists, but not this time. Our current crop of nuclear startups is being deeply irresponsible in accepting commercial contracts that they may not be able to fulfill. Raise money, sure, but do it at a reasonable valuation, publish your scientific progress often, and commit to communicating with the public in plan, clear language. 

Sorry for all the gloom and doom. You may be wondering what we can actually do about this. A few things. Investors; We know it’s tempting to yolo money into deep-tech startups you don’t understand, on the potential of asymmetric returns. We wrote over a year ago that a fusion startup is likely to be a trillion dollar enterprise within the next decade. That prediction is aging poorly, but we get the hope. But, for fuck’s sake, slow down and remember your fundamentals. Price really, really matters when investing, as does having performance-based milestones to unlock funding, and having someone in the company whose job is solely to protect capital and the public. And most of all, stop letting companies sign contracts that they may not be able to fill. This does a disservice to your LPs and to the public. 

Builders and consumers, you can help too. First, be thoughtful and minimalist in your use of AI. Keep the bubble of demand small, and bigcos will have less opportunity to do stupid shit when finding electricity to power their servers. For the record, we like AI and use it daily, but look hard at each purchase to judge whether we’re buying a toy or a tool. Lest the previous sentence be misinterpreted, nothing in Coral has been or ever will be written by AI, but we dig it for various backend tooling. Second, vote for the energy transition. Pressure your lawmakers to get on board with a price on carbon, and to close the loopholes around offsets. Energy is an issue that requires the slow, grinding machine of government to do its’ thing, so lean in. 

For us, we hope we’re wrong about most of what we’ve written here. Energy abundance is the key to anything and everything. In the meantime, we’d be happy to invest and advise for anyone building in, let’s say, next-gen geothermal or hydroelectric power. Hit us up. 

Please consider forwarding this piece to a friend or tossing it on the social platform of your choice. Coral Carbon is free and always will be, and every subscriber helps us keep doing this. If you want to work with us directly, stop by our new online home;

https://www.coralcarbon.io/