Will AI Crash The Economy?

Authored by Charles Hugh Smith via OfTwoMinds blog,

The lines of dominoes being toppled run through every nook and cranny of the economy.

As we all know, the problem with euphoria is the inevitable collision with reality and the resulting disillusionment. But wait–it gets worse.

The new love of your life, your savior who is going to make everything right again, is not just impossibly flawed–they’re a con artist. Now that really hurts. They not only stole your heart, they stole your money.

Which brings us to the AI Boom / Bubble. The euphoria is literally immeasurable, but the disconnect from reality is easily visible and can be broken down into measurable bits:

1. AI revenues are orders of magnitude lighter than the sums being invested (capex, i.e. capital investment). The euphoria is based on the idea that revenues will catch up, but the second date is raising doubts about Prince Charming’s non-flim-flammed revenues and prospects.

This report has raised eyebrows, and the real question is: OK, so let’s say it underestimates revenues by 50%. That means we’re at 3% of revenues needed to justify the capex rather than 2%. Maybe this is why Prince Charming invites his amour to poorly lit bistros–he’s had, um, work done and he’s wary of bright lighting.

$2 trillion in new revenue needed to fund AI’s scaling trend (Bain & Company)

2. AI tools are inherently untrustworthy and lend themselves to generating “going through the motions” slop that gives the superficial appearance of value but actually has negative value as it’s incomplete, misleading and/or incoherent. Sorting the wheat from the chaff actually takes more time because AI is so adept at generating a superficial gloss. In other words, AI generates time sinks rather than productivity.

AI-Generated ‘Workslop’ Is Destroying Productivity (Harvard Business Review)

People Overtrust AI-Generated Medical Advice despite Low Accuracy.

Add in that AI slop looks similar to authentic research and that AI tools have a measurable preference for AI-generated content (i.e. AI slop), and we have a toxic cocktail of untrustworthy output.

3. The rate at which major companies are adopting AI is rolling over. This chart reflects the peak of euphoria has been reached by those with the most resources to figure that out and the real-world utility of AI tools is yet to be determined.

The claim making the rounds is that it’s not Prince Charming’s fault that he’s disappointed his enamored amour; she’s making unrealistic demands on poor PC. In other words, it’s the companies’ fault that AI is underperforming. Is this the great promise of AI, to blame the mark and not the con-artist?

4. AI data centers are competing with other users for electricity, water and capital. The apologists’ claim is that AI data centers are only a tiny little straw sipping on the grid’s total energy, but this overlooks that price is set on the margins and demand for electricity and water by those with unlimited bank accounts will push prices up at rates far above the total additional consumption of AI data centers.

This reality is reflected anecdotally in household complaints that their utility bills have shot up from $250 to $800 a month. Yes, there are other factors at work–the need to invest in grid upgrades, higher insurance rates for catastrophic weather events, etc., but to ignore AI data centers’ insatiable demand for water and electricity is like seeing Prince Charming palm your wallet and then making excuses for him.

How Can We Meet AI’s Insatiable Demand for Compute Power? (Bain & Company)

By all means fact-check that 60% of Santa Clara’s electricity goes to AI data centers: I did. It’s true.

This is not to say there isn’t a use-case for AI. The point here is the excesses of capital and resources heedlessly thrown at AI in the frenzy of euphoria will crash the economy. I know it seems like there are endless trillions to toss around, but back in the real world, capital isn’t infinite, and capital squandered in mal-investments that have little to no real return is capital that could have been invested more productively elsewhere.

The same is true for water and electricity / energy. These resources are not infinite, and when someone with a bottomless bank account enters the market, prices will rise, which means consumers will be devoting scarce income to utilities, leaving them less to spend on other goods and services.

Companies spending scarce capital on AI will be forced to assess the actual financial costs and return on capital invested, and they will pull back. This retrenchment will reverse the parabolic rise of spending on AI, and that will deflate the AI Bubble that has inflated the entire stock market into a euphoric bubble that has now exceeded the extreme euphoria of the dot-com era bubble that popped 25 years ago.

So will AI crash the economy? Malinvestment on an unprecedented scale, disappointing revenues, soaring costs for utilities stripping discretionary income from consumers and the inevitable reversal of investment euphoria and the crash of stock market bubbles bursting–these are not drivers of positive economic development.

Once the stock market euphoria bursts, the wealth effect reverses, and since people feel poorer (and are poorer), they slash borrowing and spending. Those who’ve maxed out their credit have no choice: stop paying the car loan or rent to keep the lights on.

One domino falling, OK, no big deal. This is different: the lines of dominoes being toppled run through every nook and cranny of the economy. What’s been untouchable will be touched–by a hammer.