Gold has nearly kept pace with the biggest tech boom of our lifetime. Since ChatGPT launched in November 2022, gold has underperformed the NASDAQ by only 15%, and that gap comes after gold itself fell 15% over the last five months. So over the past four years you could have bought gold and roughly matched this revolutionary technology. Before you even reach the part where the money gets printed to backstop the whole AI buildout, gold has already held its own. If that printing comes, gold stands to outperform tech by a wide margin.
The dollar hides what is really happening
Markets have been strikingly resilient in dollar terms. If you had shorted stocks when the Fed started hiking rates, with the NASDAQ at its late-2021 highs, you would have lost: stocks fell but came all the way back. The S&P 500 is about 80% higher than it was at the start of January 2022, when hikes began.
Measured in gold, the picture flips. In gold terms the S&P was down about 40% earlier this year. Consumer confidence and the S&P used to track each other closely, then they split in 2022. Overlay the S&P priced in gold against consumer confidence and the two charts match again. The S&P increasingly just measures dollar debasement, the loss of the dollar's buying power.
Real estate tells the same story. Home prices in dollars are off the charts, but priced in gold they sit at 1980 levels, back when mortgage rates were 15% to 20%. A rising number in dollars can hide a portfolio that buys less and less.
Gold's plunge as a warning
Gold peaked near 4,600 in August and dropped to about 3,850 by Friday. That last leg down is worrying. Gold tends to move first. When it acts this way it signals that something bad is coming: that the Fed made a mistake by hiking, that the AI trade is breaking, that there is a liquidity problem. Gold warns of these before the broader market sees them.
Is the warning about the economy and markets, or about geopolitics? Both.
Silver and the electric grid
Silver has lagged badly and needs to rise, because none of the new buildout happens without far more silver. The electrical infrastructure names such as PAVE and GRID also look richly valued, though nowhere near AI levels. The United States added no new electricity generation capacity for 20 years, from 2004 to 2024. Even if AI collapses, the country still has to build a lot more grid. That demand does not stop.
Critical minerals and the China trap
The wars are a bigger waste of capital than AI. The only thing burning money faster than AI is the United States in Iran and the other pointless wars of the past 25 years. The Iran conflict was supposed to last three or four weeks; it is now in month nine and going nowhere.
Michael Every's idea runs like this: because the US handed off the processing and production of critical metals over recent decades, it now faces a choice. Either force other countries to hand over the metals it needs, or create an economic environment so painful worldwide that others suffer far more than America does. Pick your poison - hand it over, or life gets miserable, worse for you than for us. This strategy is mistaken, another faulty plan of the same kind that produced the Iran miscalculation. Military spending cannot quickly rebuild decades of lost industrial capacity.
China likely welcomes all of this, because it comes out far ahead in the long run. China just sold about 38% more electric scooters year over year. Every time Trump talks tough on Iran, China sells more solar panels and more BYDs. It works like gun sales under Obama, when every speech about guns sent gun stocks soaring and people bought more rifles. Ordinary people do not care about the elites' power games. They want to drive to work cheaply. No one will bleed for this.
The split is simple: America makes debt and paper, China makes things. The result is intense deflation inside China. China has the lowest bond yields of any major country by far. Part of that is its closed capital account, but a big part is that the world sends China $1.2 trillion a year. China no longer buys US bonds. It recycles that money into factories, gold, and commodities, and it is shifting how things get priced.
What happens when the US runs out
The question no one answers: what happens when America runs out of the munitions and defense tech it needs? The answer is that the whole world learns the US military cannot go to war. The world moves deeper into chaos. The United States becomes safer in its countryside than in its cities.
Then China steps in: things are rough over there, would you like some drones and equipment to make your country safer? Sure. How do you pay? In yuan, or you can reserve gold and settle in gold. Fine. Once allies see the US defense umbrella is dead, the reason to hold dollars disappears. You go to gold and you buy directly from China. China provides your protection, so attacking China is off the table - that would be the end of the world. If the US cannot field enough missiles and defense tech against Iran because China controls the inputs, then a fight with China would cut America off completely.
The dollar's strength rests on confidence in American protection, something markets rarely price correctly. If allies doubt that umbrella, they rush to diversify reserves and trade settlements. China's grip on defense supply chains makes that shift more dangerous. Monetary trust and military credibility are now tied together.
People in the military knew about these supply chain problems in early 2023. This was no surprise. So two possibilities: is the Trump administration weakening America on purpose, or is this a future Harvard Business School case where bad news never reaches the top because the boss yells at anyone who brings it, so he thinks things are fine when they are not? Multiple people, chapter and verse, say it is the latter. A third option: bad math - a bet that the enemy could not withstand a shock and a barrage, so take them out, with enough supplies on hand.
The AI bubble against a broken balance sheet
AI capex is the largest in history by a wide margin, bigger than the railroads, growing exponentially larger than anything before. Every earlier infrastructure mania ended in a secular bear market that caused enormous pain.
In gold terms, in real terms, this likely ends as one of the great bear markets in history. In dollar terms, maybe not. Remember 1929: the dollar was backed by gold, the Dow fell 90% in gold terms, and the moment the dollar was devalued against gold, the Dow took off. The same dynamic holds from 2022 to now - the S&P is up 80% in dollars but still down in gold. Even after the gold selloff, tech is still down about 15% in gold terms five years later.
A Bloomberg article from a few weeks ago lists the five prior cases: the canal bubble of the 1840s, the railroad bubble of the 1870s and 1880s, the electrification bubble of the 1920s, the highway bubble of the 1950s (possibly running into the 1970s), and the telecom bubble of the late 1990s and early 2000s. This AI boom is by far the biggest, so its burst should be the biggest too, in real gold-measured terms.
One thing breaks the comparison. Debt to GDP at the federal level was tiny in every earlier bubble. It is 120% now. The federal deficit as a share of GDP was tiny then. It is 7% and rising now. Entitlements were tiny or cash-flow positive during the telecom bubble and did not really exist before that. So the AI boom unfolds against a fiscal picture that makes every past bubble look mild. The real danger is not just falling stocks. It is how the government pays for the aftermath.
