Research and Writing

The US labor market today puts that on full display. As the name says, labor is a “market” subject to the rules of supply and demand. When there are too many workers and not enough jobs, the price of labor - wages - falls. And vice versa. The chart below from the Atlanta Fed shows the picture: wage growth has fallen from the super inflationary levels in 2022, but is still well above most of what we have experienced since the GFC. The price of labor is telling you the market is fine.

But how can that be? Hasn’t demand for labor obviously fallen? We get headlines every week with layoff announcements, and the labor stats say we aren’t adding many jobs every month!

Yes, that is true - and callously, we are lucky it is true. Because if it weren’t true, we’d have an inflationary mess on our hands. Why? Because while the demand for labor has fallen, the supply of labor has also fallen a lot. If we lost all of those workers without losing some jobs, the supply/demand balance for labor would cause wages to rocket. Our labor comes from two sources. We still grow some the ol’ fashioned way, though not as many as we used to because of our demographics. Also, workers migrate here...or at least they used to. The chart below shows our best estimate of net migration into the US. As Sesame Street taught us, one of these things is not like the others.

The net impact of all of this is that the labor market is starting to look more and more like Japan’s. Japan’s work force doesn’t grow. Unemployment is super low, but if you do end up unemployed it takes forever to find a new job because there is no dynamism. No one is hiring, just like no one is firing. It creates a new normal where the 100k+ monthly job creation numbers we have grown accustomed to are a thing of the past. And unless lowering interest rates indirectly causes a bunch of babies to be conceived, it isn’t going to do anything helpful.

The Fed is going to ease this week on the pretense that the labor market is not ok and inflation is ok. We see conditions as the opposite, and believe this policy mistake will put a nice tailwind behind commodities and a headwind to the dollar.

But what is “counting cards” and how does it work? The simple version is that as cards get used up from a deck, the remaining deck now has different characteristics and probabilities. In blackjack, a deck with more high cards in it favors the player more, and a deck with more low cards in it favors the casino more. So players sit at the table and keep track of which cards have been used since the last time the deck was shuffled. When the deck favors them, they bet a lot, and when it doesn’t, they bet the minimum they can to keep their seat. In other words, they print money by recognizing when the conditions warrant taking more or less risk.

Looking across financial markets at the moment, we see a deck full of low cards. So making small bets and waiting for better opportunities looks like the right plan. In our commodity strategy, we see the oil market as torn between oversupply and unpredictable geopolitical events. The copper market is torn between tariff fears, mine accidents, and mediocre demand. The nuclear complex is torn between the hype and the reality of how long new cash flows will take to materialize. Outside of commodities, the Fed is torn between inflation that is too high and sticky, and a labor market that doesn’t fit neatly into their frameworks. The US stock market is torn between fears of a bubble and the river of liquidity flowing from fiscal deficits and capex. There are still some opportunities out there, as always - many EM assets, pockets of European equities, TIPs, some refineries and natural gas producers, some pockets of the vol market, and we still like gold. But our view is that now is not the time for big swings…it’s better to be boring, stay diversified and stay global, and react to how conditions evolve.

Point 1: Nuclear investments are very long-term in nature, so they require confidence that both political parties are on board. And they are.

In 2024, the Biden administration passed the ADVANCE Act via a nearly unanimous vote in both branches of Congress. The act laid important groundwork to reduce the burden and regulatory risk around building new nuclear capacity. Since then, the Trump administration has continued and even accelerated that push, ensuring the Act is implemented, adding to the government support of nuclear through executive orders and signing a deal with Westinghouse with the ambitious goal of quadrupling US nuclear output by 2050. It’s interesting to contrast that with the uptake (or lack thereof) from major oil companies in terms of expanding in Alaska. The oil majors are concerned that the next time the Democrats are in charge, their permits will be yanked and their investments will be wasted, which is a reasonable thing to be concerned about!

Point 2: Despite the government tailwind, US nuclear expansion will be far too slow to align with AI power needs.

Despite all the talk, there are still no new reactors under construction in the US. Half of global construction is in China, the rest is sprayed between Europe, Asia, and Egypt. The map below from The World Nuclear Industry Status Report 2025 shows the picture.

And even if we did start building reactors, we are out of practice and will probably stink at it for a while. The only plant we’ve built recently was Vogtle, which was a spectacular disaster - it took 11 years and cost over $30bn. Meanwhile, China pumps these things out in 6 years at a fraction of that cost. That means we are probably looking at a best case scenario of 2035 before meaningful new nuclear capacity can hit in the US, waaaay too slow for AI ambitions.

Point 3: That includes Small Modular Reactors, known as SMRs, which for all of the hype are not close to being useful.

There is a huge difference between building a demonstration reactor that doesn’t malfunction and large scale deployment at high capacity factors. The chart below is from a survey of industry insiders conducted by UxC. The fact that the timeframe they are asking about is “by 2050” tells you most of what you need to know about SMRs.

Point 4: We’ve added some uranium mining capacity, but still not nearly enough.

While it is impossible to be too precise or confident with these numbers, our best estimate is that the world uses up 180mm lbs of uranium a year and pulls 160mm lbs out of the ground. The gap is filled by drawing down inventories (which built up massively post-Fukushima) and a smattering of recycling. There is a fascinating dynamic in the market where the utilities believe that the new mines in development (from Denison, NexGen, and Paladin) will deliver on schedule and reduce the pressure in the system, so they are preferring to burn down inventories rather than pay what they perceive as high prices. Meanwhile, actual miners are like “lol, that’s adorable that you think a mine can be completed on time.” Relevant UxC survey below. Of course everyone talks their book to a degree, but in this case their actions are consistent with their words.

Point 5: The largest public nuclear mining companies are richly priced, so we are positioned in physical uranium and a smattering of lottery-ticket junior miners.

“Bubble” is too strong a word, but overall the capital that has flowed into the public market space has been too big to all be efficiently allocated. Cameco does have a decent chance of growing into its valuation and is a reasonable long-term holding, otherwise we are staying away from larger and medium-sized players. Most of our exposure at this point is in the physical uranium trusts, which could benefit from squeeze-like dynamics if/when news comes out that those mining projects aren’t all smooth sailing.

But long before Trump was re-elected and the trade war ignited, China was already working to reduce its biggest point of vulnerability. China is – by far – the largest importer of oil in the world at a staggering 11 million barrels every single day. 11 million barrels, laid end to end, would stretch from Beijing to London. Yikes.

So how do you reduce such a staggering reliance on foreign countries? To start, find suppliers that can’t weaponize their oil, because they actually need you more than you need them – i.e. sanctioned countries. The table below shows where China gets their oil. It’s not a surprise that Russia is number 1 and the US – despite being the largest oil producer in the world by far - doesn’t crack the top 10. The only real surprise on the list is “Malaysia.” I use quotes because Malaysia’s exports to China are 4 times the volume of oil they pull out of the ground. There is absolutely no doubt that some savvy Malaysian businessmen are making a pretty penny laundering Iranian and Venezuelan oil.

Another way to reduce vulnerability is to grow domestic production, and China has been on one heck of a spending spree to that end. In the last 6 years China has spent half a trillion dollars on exploration and drilling and clawed themselves up to 5th place in the world in total oil production. Meanwhile, they’ve been socking oil away for a rainy day – adding roughly a million barrels a day to their storage this year. At this point their reserves are 4 times bigger than the US strategic reserve, tipping the scales at around 1.2 billion barrels – over 100 days of coverage if there were a total stop to their ability to get oil from abroad.

The last leg of the stool is reducing demand. The nice thing about this goal is that you can make pretty good headway just by going about your normal day - humanity is constantly improving the energy efficiency of our economic activity through the normal mechanisms of technology advancement and capitalism. The chart below shows how much oil it takes to generate a dollar of real GDP for the US and China. Since the year 2000, the US has knocked about a third off this metric, and China is close to half.

The other way to reduce demand is substitution – swap out oil for some other energy source. No matter how hard that is to do in your head, you are underestimating how hard it is. This chart is from the IEA. The way you read this chart is that the total height of each bar is how much oil they would have used if there was no substitution away from gasoline and diesel transportation. The checkered parts of the bar are areas where oil product demand has been displaced. Half of EVs in the entire world are in China, and you still need a magnifying glass to see the impact on their oil demand.

For the last decade, China has been the primary engine of global oil demand growth. If Beijing has its way, that will no longer be the case, and the evidence points to them having their way. Meanwhile, global investment in oil supply has been woefully inadequate for many years, creating concerns about maintaining and/or growing global production. How will those two secular trends net out? Our guess is “poorly,” and that betting against a repeat of the relatively placid price action in oil over the last two years has a nice expected value.

Meanwhile asset managers like us tell people they should be diversified, and folks nod as they press the button to buy more NVDA. Their brain is on board with diversification but it’s not what their gut actually wants.

I am often reminded of a haunting conversation I had many years ago with a senior person at a large sovereign wealth fund. To paraphrase: “If we have 10 managers and 9 make money in a quarter, we celebrate. No one is concerned that all of our managers make (and presumably lose) money at the same time. Instead, the concern is for manager 10…why couldn’t they make money this quarter? Should we put them on a watch list?”

I find it fascinating how intuitively humans accept diversification in other parts of their life, just not with their money. A sports team is just a diversified portfolio of athletes who shine at different times. No one watches a soccer game and says “who’s that bum in the back with the colorful shirt, he hasn’t scored a goal all season.” Yet when bonds do poorly during economic booms, that’s pretty much exactly the reaction.

Making money feels good. It’s a dopamine hit that we want again and again. And because of that, most investors tend to overweight their personal experience of what has made and lost money versus considering the wider range of outcomes that history teaches us.

The table below is that history. I made it a while ago, so the numbers for the current decade have changed, but the point hasn’t changed. The assets that do the best and the worst are constantly changing. Asset class outperformance tends to make an asset expensive, which sows the seeds for poor future returns. The opposite is also true. Unloved assets get sold, which depresses their prices relative to the cash flows they produce and increases their future returns. The big takeaway is that unless you own a crystal ball you are usually better off owning a bunch of different stuff than gambling on one.

Why am I saying this now? If you are ~40 years old or less, almost your entire adult life has been marked by the greatest bull run in a single market in all of financial history. Congrats! But now it is time to hold your nose, trust your brain over your gut, and take some chips off the US equity table. Not because some imminent crash is coming or something like that, but because it is always the right thing to do, and it is especially the right thing to do if you haven’t done it in a long time and your portfolio is concentrated in a small number of highly correlated stocks that just rallied a zillion percent.

Before getting into anything AI-specific, it's worth stepping back and considering humanity's track record when functionally-infinite money is thrown at a problem. When measured apples to apples, the amount of money spent on powering AI will dwarf the amount of money it took to figure out how to put a man on the moon with "computers" that were closer to abacuses than to the computer in my dishwasher. We ran highways and railroads across the US for much, much less. The Manhattan Project was a drop in the bucket by comparison. We developed a vaccine to COVID in record time, and the cost could have been covered by the loose change in Sam Altman's couch. To solve the AI power problem, we need some combination of 1) producing more of a thing - electricity - that we are already good at producing, and 2) improving the energy efficiency of technology, a thing we are also very good at. To think THIS is the problem that will ultimately thwart our species fails the smell test.

But let's get specific. Here is a chart where a bunch of smart people sharpen their pencils and project how much juice all of these AI data centers will need. The estimates are comically different from each other: a ~5x range over the next 5 years.

For the sake of this analysis, we will pick the highest line from the chart: BCG's "high" estimate at ~1000 Terawatt-hours (TWh). We'll use it because it is the most difficult bar to clear, even though I will eat my hat if the real number ends up anywhere close to that. Current data center demand is about 200 TWh, so they think it'll add an incremental 800 TWh. Current US total electricity demand is about 4200 TWh. So if you do the math, they are projecting an electricity demand growth rate of about 3.5% over the next 5 years. That is high relative to recent history, but in a longer context it is completely normal - in fact, since 1950 US electricity demand has grown at an average of 3.2% per year.

Where can we find 800 TWh? To start, let's look at where electricity comes from now: the chart below from the EIA is labeled as "billion kilowatt-hours" instead of "terawatt-hours" but those units are the same thing.

Some highlights from this chart:

- Natural gas is now our largest source of electricity, and has grown at ~3.5%/year for the last decade. If it continues at that pace it will add ~300 TWh on its own, and that pace could certainly be dialed up. For all intents and purposes we have infinite natural gas, and it is super cheap. We think this will be the largest contributor to the solution.

- Nuclear is too slow to move the needle. I wrote a piece titled "Calling Homer Simpson" that outlines why...scroll down to March on this page to read more.

- Renewables have some issues as it relates to intermittency (that are tractable) but could easily provide most or all of the 800 TWh. As a point of comparison, China added 1400 TWh of renewables over the last 5 years.

- Coal is an interesting wild card. It has gone from about 2000 TWh at the peak to about 700 now. Of that ~1300 TWh reduction, we could get almost all of it back on a timeline that works. About half of the reduction came from just lowering the output of still operational plants, most of which could be dialed back up. The other half came from plant retirements. How long it would take to unretire those plants is largely a function of political will. At the extreme, if Trump signed an executive order waiving permitting/studies/paperwork to do this (and it stood up in court), you are talking about a couple of years, maybe 5 years tops. Fast enough for the problem at hand. As an aside, for years we've been hearing from ivory-tower types that unemployed coal miners should be re-educated to learn how to code. Maybe the universe has a dark sense of humor and AI will take the coding jobs, and then coders will have to be re-educated to learn how to mine coal!

When you put it all together - even if you assume that the BCG high estimate is completely correct and the army of PhDs who are trying to lower the energy cost of AI are toiling away fruitlessly, we could probably cover the incremental demand by 2x if we went all hands on deck. And spending trillions of dollars tends to get a lot of hands on the deck.

Now...saying we can do this doesn't mean there won't be big hiccups. It will be incredibly HARD and will require little to no wavering in financial resources and political will. There are a zillion articles out there about the literal bottlenecks in materials and components, and they aren't wrong - it is basically guaranteed that this won't be a smooth process. Some part of the supply chain will get snarled, some protests will pop up and complicate permitting, there will be a blackout in some mid-sized city and associated outrage, etc etc. But when we compare those issues to the financial rewards for clever humans to solve them - we see a footnote in history, not the main story.