by Kyle Johnson

 

Uranium’s price is at a 12-month low. Lobo Tiggre, The Independent Speculator, says it’s taking all of his self-discipline not to go immediately all in. Here’s why.

 

Trump and AI

President Trump’s second term began with a flurry of executive orders and sweeping changes. This includes committing $500 billion to “Stargate”—a joint venture between OpenAI, Oracle, and SoftBank (among others) to build a network of AI data centers.

But here’s the most interesting bit: Trump acknowledged, “They have to produce a lot of electricity, and we’ll make it possible for them to get that production done very easily at their own plants if they want.” He has explicitly referenced coal and nuclear energy to meet this need.

Last September, I wrote about Amazon Web Services purchasing a 2.5-gigawatt nuclear facility.

Shortly after it was published:

• Microsoft announced an agreement with Constellation Energy to restart the reactors at Three Mile Island (which will be renamed the Crane Clean Energy Center).
• Amazon announced a deal with Dominion Energy Virginia to receive power from the utility’s nuclear reactors.
• Amazon also signed a contract with Energy Northwest and X-Energy to develop small modular reactors (SMRs).
• Oracle announced plans for an SMR-powered data center.
• Google announced it purchased six to seven SMRs from Kairos Power.
• Mark Zuckerberg’s plan to build a nuclear-powered data center for Meta’s AI was derailed because of a rare species of bee.

Meanwhile, Trump is defunding and cancelling “alternative” energy—especially windmills.

AI is going to need all the power it can get—even as it becomes more energy efficient—which makes nuclear a necessary part of the mix.

Love him or hate him, Trump seems to understand this. On January 20, Trump signed an executive order titled “Unleashing American Energy.” Section 9(c) reads:

The Secretary of the Interior shall instruct the Director of the US Geological Survey to consider updating the Survey’s list of critical minerals, including the potential of including uranium.

Bullish.

 

Not So Bullish

Lobo calls Trump an agent of change (if not chaos). Trump remains true to form. On the one hand, his actions seem likely to boost demand for uranium. And on the other hand, he threw a couple monkey wrenches into the mix.

In Trumpian fashion, he announced on Truth Social that his administration is authorized to immediately begin producing “Energy with BEAUTIFUL, CLEAN COAL” [sic].

Interior Secretary Doug Burgum wants to restart coal plants that were shuttered to due Obama and Biden’s regulatory actions. Energy Secretary Chris Wright has stated that the premature move away from coal has destabilized the energy grid and wants to pause the closure of coal-fired power plants.

At the very least, it seems that the Trump administration views coal as a stopgap—a fast and reliable way to meet energy demand in the short term. But it will likely come at the cost of slower nuclear deployment.

Long term, nuclear is still the best option. Although Trump has taken action to support uranium and nuclear energy, he continues sending shivers down the spines of uranium investors.

In mid February, Trump made several comments expressing interest in denuclearization talks with Russia and China. This startled many in the uranium markets, drawing comparisons to the past Megatons to Megawatts program.

From 1993 to 2013, the US purchased 500 metric tonnes of highly enriched uranium (HEU), the kind used for weapons, from Russia and downblended it to about 15,000 tonnes of low enriched uranium (LEU), the kind used to generate electricity. This is roughly the equivalent of about 153,000 tonnes of “natural” uranium—what mines produce. On average, the program brought 7,650 tonnes to the supply chain per year.

This is perhaps the key figure to focus on because most supply, demand, and consumption statistics are given in uranium’s pre-enriched or natural form.

To help put these figures into perspective, consider this: Viktor Mikhaylov, head of Russia’s Federal Agency on Atomic Energy, estimated that Russia inherited approximately 45,000 nuclear weapons after the fall of the USSR. The 500 metric tonnes of HEU purchased via the Megatons to Megawatts program was enough for 20,000 warheads.

Thankfully, the numbers are less absurd today. Estimates vary, but many analysts believe Russia currently has an inventory of 5,580 warheads (including 1,200 awaiting dismantlement). The US Defense Department estimates China has approximately 600 nuclear weapons.

We can run some back-of-the-envelope calculations about how much uranium would enter the supply chain under a new program. Admittedly, modern fusion bombs use significantly less uranium than fission bombs. But I think it’ll be hard to ascertain the exact number of fission vs. fusion warheads. For the sake of argument, let’s assume that Trump could convince both China and Russia to reduce their nuclear inventories by half. Let’s crunch some numbers using the same ratios as reported from the first Megatons to Megawatts program.

If 500 tonnes of HEU was enough for 20,000 warheads, then each warhead contained about 25 kilograms of uranium.

Let’s cut China’s 600 warhead inventory in half.

• 300 warheads x 25 kg of HEU = 7.5 tonnes of HEU from China

Now let’s cut Russia’s 5,580 warhead inventory in half.

• 2,790 warheads x 25 kg of HEU = 69.75 tonnes of HEU from Russia

Combined, this is about 77.25 tonnes of HEU.

Using the same ratios from the first Megatons to Megawatts program, 77.25 tonnes of HEU can be downblended to about 2,317.5 tonnes of LEU. This translates to about 23,683.5 tonnes of pre-enrichment uranium.

A significant sum. But only about 15.5% of the original Megatons to Megawatts program.

Here are some basic figures to grasp how the uranium market changed during the original 20-year program. When it began in February 1993, the price of uranium averaged about $10 per pound. That year, global consumption was 68,038 metric tonnes and production was just 39,916 metric tonnes. Despite this, prices slumped for more than a decade, dropping to about $7 per pound.

Then demand took off again, and spot uranium spiked to $143 per pound in June 2007. That surge wasn’t sustained: uranium had dropped to about $34.50 per pound when the program ended in December 2013. That year, global uranium consumption was 75,749 tonnes while production was 70,762 metric tonnes.

So would a new Megatons to Megawatts program oversupply the uranium market?

Absolutely not.

But that won’t stop markets from overreacting to news of new supply. If this happens—which Lobo says is a big “if”—uranium investors need to be careful. Uranium stocks would likely get clobbered. We’d expect them to rebound once reality set in, but there’d be a period of pain first.

Of course, nothing might happen at all. This is all based on a few comments made by Trump. But if you’re in uranium stocks (or thinking about it), this issue warrants your attention.

 

Software, Hardware, and Trillionaires

Markets tend to overreact to news. We’ve seen this recently with AI and energy demand.

For example, DeepSeek (a Chinese tech company) made waves with its cheap generative AI. The company claims it slashed application programming interface costs for the end user by up to 90%. Research by JP Morgan stated DeepSeek appears to have trained its models 45 times more efficiently than its competitors. DeepSeek claims training costs were just $6 million.

This was enough for some to claim the AI energy boom was over. But of course, the devil is in the details.

SemiAnalysis, an independent research company, reviewed DeepSeek’s claims and found that the $6 million figure did not include total server capital expenditures (an eye-watering $1.3 billion).

A study by the MIT Technology Review concluded DeepSeek could actually be more energy-intensive when generating responses than the equivalent-sized model from Meta. It noted, “The issue might be that the energy it saves in training is offset by its more intensive techniques for answering questions, and by the long answers they produce.”

MIT asked DeepSeek whether it’s acceptable to lie. DeepSeek spent 17,8000 joules to generate a 1,000-word response—41% more energy than Meta’s model when given the same prompt, according to MIT.

This doesn’t smell like the end of AI energy demand.

There have also been many hardware improvements. The release of Microsoft’s Majorana 1 chip also made quite a stir, with the company stating quantum computing is “years, not decades” away. Google released a chip last year, stating that quantum computing was “five years” away. IBM estimates large-scale quantum computers will be online by 2033.

At some point, there will be a truly revolutionary piece of hardware. But it won’t necessarily spell the end of the AI energy demand boom.

AI and quantum computing seem positioned to generate trillions in market cap for the leading companies. They might even produce the first US dollar-denominated trillionaires. Given the stakes, I suspect all efficiency gains made by software and hardware will simply reduce the time needed to achieve certain benchmarks, not actually reduce the energy consumed to get there.

Making AI more efficient just means we’ll use more of it.

 

Microsoft Pulling Back?

Microsoft recently cancelled leases related to AI data centers. The headlines say that Microsoft overestimated the capacity needed for its partnership with OpenAI. However, I suspect there’s more to it than that.

What if Microsoft’s executives reevaluated their relationship with OpenAI and their desire to better secure intellectual property rights? Consider that Microsoft’s agreement to restart Three Mile Island includes a 20-year power purchase agreement. A nuclear reactor cannot be turned on and off like a lightbulb. Microsoft’s actions with Three Mile Island are weightier than its decisions with AI data centers used for a partnership.

I appreciate that Bill Gates is no longer on Microsoft’s board, but he remains a technical advisor, and he’s strongly pro-nuclear. In a recent interview with the Wall Street Journal, Gates claimed he has a “very close relationship” with CEO Satya Nadella and spends 15% of his time on Microsoft. Don’t forget, Microsoft was subject to one of the largest antitrust cases in US history while Gates was CEO.

A tiger doesn’t change its stripes. Gates’ legacy is inextricably linked with the future of Microsoft. And he doesn’t seem like a guy who wants to share the spotlight. This is speculation, but I suspect he is advising Microsoft to tackle AI without involving a third-party AI company. Why make progress in AI with a partner when you can afford to do it on your own?

In part, analysts at TD Cowen seem to agree. They reason that Microsoft’s withdrawal was “largely driven by the decision to not support incremental OpenAI training workloads.” However, they believe that lease cancellations and deferrals of capacity “points to datacentre [sic] oversupply relative to the current demand forecast.”

Time will tell.

 

Restarts and Fresh Starts

The bulk of energy demand comes from non-AI sources—and it’s only increasing.

Federal grants have been approved to help facilitate the restart of the Palisades nuclear facility in Michigan. NextEra Energy has initiated talks with regulators about possibly restarting the Duane Arnold nuclear power plant in Iowa. Texas A&M dedicated land on campus to help four nuclear companies (Kairos Power, Natura Resources, Terrestrial Energy, and Aalo Atomics) build SMRs.

In the US alone, utility companies forecast the need for 38 gigawatts of new power production over the next five years to meet demand from data centers, electrification, and new industry. That’s the equivalent of about 34 new nuclear plants.

I’m happy to see this activity in the US, but nuclear energy is a global trend. Last October, the Onagawa power plant in northern Japan was restarted after being dormant since 2011. In December, Unit 2 of the Shimane nuclear power plant in western Japan was brought back online after being shut down in 2012. Japan has now restarted 14 reactors since 2015, with another 11 units pending while under regulatory review.

The World Nuclear Association reports that 440 nuclear power reactors are operational globally. There are about 65 reactors currently under construction and another 90 in planning.

The nuclear renaissance is picking up steam.

 

Litmus Test

The current uranium market is a good way to test your appetite for resource stock speculation. Uranium prices are down. Sentiment is down. It’s hardly in the news.

But long term, the bull case for uranium remains unchanged. If anything, it’s improving. The AI energy boom is real and growing. SMRs are being built, yet they are not included in most uranium supply/demand calculations.

There are bargains to be found, if you have the stomach for it.

KJ

 

PS: Lobo is always on the hunt for a good deal on uranium and other mining stocks. There’s only one place to get his latest thoughts on this and other resource markets: the Speculator’s Digest—our free, no-hype, no-spam newsletter.