Everyone Is Digging for AI Gold

For those reading this on the web, those that recognize the .gif know ball.

At a panel recently, the host (and for some reason also a panelist?) described the AI space as a gold rush. They meant it as a compliment. The room nodded along. Good energy, all of that.

Then someone else on the panel pushed back. Paraphrasing: yes, and gold rushes also mean holes everywhere. Piles of upturned soil. A landscape that looks like it got hit by something. People falling in before they even see what they're falling into.

Back to the original analogy… If I had a dollar for every time someone in the AI space uses a metaphor or analogy that is more poignant and AI critical than they intend, I'd have my own trillion dollar valuation at this rate. Reminds me of when I watched someone argue against the Colorado AI bill by saying AI should be treated like when a dog bites someone. So close. Genuinely almost there. And yet completely missing it, probably because his company had a stake in the outcome.

That second framing stuck with me more than the first. Not because the gold rush comparison is wrong. It's actually one of the more accurate things I've heard said about the last few years of AI. But because every time someone uses that analogy, they tell the fun half and skip the rest.

One thing to keep in mind before we go further: gold rushes aren't defined by the gold. They're defined by what people do because they think the gold is there. So let's talk about the rest.

Everyone starts with the investment numbers, so let's get that out of the way.

Foundational AI companies raised $1.4 billion total in 2022. In just the first quarter of 2026, that same category pulled in $178 billion. By 2025, AI was capturing close to 50% of all global venture funding.

Cool. Super relevant if you're on a cap table.

Counting AI tools is a losing game. Depending on which directory you check, you'll get five different answers. But the direction is pretty clear. According to Stack Overflow's 2025 Developer Survey of over 49,000 developers across 177 countries, 84% are now using or planning to use AI tools in their development process, up from 76% the year before. More people are building more things faster than at any point before. Whether a meaningful chunk of what gets built solves a problem anyone actually has is a separate question.

Here's the part that doesn't make the announcement posts though. Trust in AI accuracy among those same developers fell to 29% in 2025, down from 40% the year prior. Adoption is up. Trust is down. Make of that what you will.

Building has increased fast. Lovable, a platform where non-developers can build functional software through plain language, hit $100 million in annual recurring revenue eight months after launching, with about 45 employees. By late 2025, they were closing in on 8 million users and over 100,000 new projects created every day. Anyone can ship something now. Whether what gets shipped matters is a separate conversation most people aren't rushing to have.

Talking about the Gold Rush is awesome. Unlike some administrations, I love revisiting history and comparing to where we’re at today.

The basic story here isn't really contested. Gold was discovered at Sutter's Mill in 1848. Hundreds of thousands of people rushed in. As the easy deposits ran out, individual prospectors got replaced by organized capital and machinery. Most people who came went broke. The ones who got rich were mostly not miners.

Sam Brannan heard about the discovery early. Before telling anyone, he quietly bought up every pickaxe, shovel, and pan he could find in the region. Then he walked through San Francisco waving a bottle of gold dust and started shouting the news. He became California's first millionaire. Not from mining. From selling miners their tools.

Levi Strauss made durable pants. Wells Fargo moved money and mail. The merchants in Sacramento charged whatever they wanted for flour and rope because the miners had no other options and needed to eat. The miners were the story everyone told. The merchants were the ones who actually won.

So who are the merchants now? Nvidia's data center revenue grew 279% year over year in Q3 2023. By Q1 2024, it was up 427% year over year. They didn't build an AI product. They sold the infrastructure everyone else needed to build theirs. The cloud providers did the same. The API is the modern pickaxe. You don't need to strike gold if everyone pays you every time they swing.

Meanwhile, the actual miners are falling into their own holes: startup failures tracked by Carta rose 25.6% from 2023 to 2024, with AngelList tracking a 56.2% increase over the same period. At Series B specifically, closures rose 133% from Q1 2023 to Q1 2024. Largely the same story. Different branding.

There are people raising serious concerns about all of this. They keep getting buried under the next product announcement.

Daron Acemoglu, an MIT economist who won the Nobel Memorial Prize in Economic Sciences in 2024, said it plainly: "These models are being hyped up, and we're investing more than we should. Much of what we hear from the industry now is exaggeration." That's not some tech-skeptic blogger. That's one of the most recognized economists in the world.

The enterprise data backs him up. McKinsey surveyed nearly 2,000 organizations for their 2025 State of AI report and found that 88% report using AI in at least one business function, but only 5.5% are seeing more than 5% EBIT impact from it. Two-thirds are stuck in what the report calls "pilot purgatory," running AI experiments that never graduate to production.

To be clear: none of that means the technology doesn't work. It means most organizations are paying for something they haven't figured out how to actually use yet, while the people selling them the infrastructure keep building more of it. Right now, AI is a capital expenditure story. The productivity story hasn't arrived for most people.

Gold rushes don't just change who gets rich. They change what gets dug up, what gets burned, what gets drained, and what gets left behind.

The environmental damage from hydraulic mining in California took decades to address. Rivers got redirected, landscapes torn up, towns built on boom economics that collapsed when the gold ran out. The land absorbed the cost of everyone else's ambition. The AI version of that is happening now, and it's not just a metaphor anymore.

A June 2026 report from the United Nations University puts real numbers on it. By 2030, global data centers powering AI are projected to consume 945 terawatt-hours of electricity annually, nearly triple the combined electricity use of Pakistan, Bangladesh, and Nigeria. The water footprint will equal the basic annual domestic water needs of all 1.3 billion people in Sub-Saharan Africa. The land footprint will exceed 14,500 square kilometers. On top of that, hardware refresh cycles driven by AI competition could generate up to 2.5 million tonnes of electronic waste per year by 2030.

And that's just the infrastructure. The cost is also landing on specific people.

The IMF estimates that nearly 40% of global employment is exposed to AI, rising to about 60% in advanced economies where more jobs involve cognitive work than physical labor. That's exposure, not guaranteed displacement. But Goldman Sachs data shows workers aged 22 to 25 in AI-exposed roles have already seen a 16% employment drop, with software developers in that age range clocking nearly a 20% decline from their late-2022 peak. US customer service employment dropped by about 80,000 positions between 2022 and 2024. Not projections. Jobs that existed and then didn't.

The noise is a cost too. Ahrefs analyzed nearly a million new web pages published in April 2025 and found that 74.2% contained detectable AI-generated content. Spotify removed 75 million spam tracks in a single year as AI tools made it trivially easy to generate fake music for streaming revenue.

Who's falling into the holes? Both the people digging and the innocent bystander going to the market.

The Gold Rush peaked around 1852. By 1855, the easy surface gold was gone. The individual prospector got replaced by industrial mining operations, better capitalized, more methodical, built for a problem that turned out to be harder than it looked from the outside.

That consolidation is happening in AI now. Median late-stage AI deal sizes went from $48 million in 2023 to $327 million in 2024. Money is concentrating, not spreading out. Vendors are already quietly scaling back and discontinuing AI products as results take longer to pan out than expected. The announcements about that tend to be a lot quieter than the launch events were.

Those attuned to the AI space may remember the surge of “wrapper apps”, or basically an app that the value proposition is calling an AI model. The wrapper app era didn't go out with a bang. It just got ignored until it was gone.

Between the history books and the maps, we can surmise safely that the Gold Rush ended. Meanwhile, California (and several other things started during that period) didn't.

The people who stayed built farms, banks, law practices, newspapers, and eventually universities. I believe that Stanford was funded largely through the railroad as a result of this rush. The long-term value wasn't in the gold. It was in everything built while everyone was looking for it, by people who were honest about what they were actually doing.

The AI companies that matter in 2035 probably aren't the ones that moved fastest in 2023. They're the ones that had a real answer to "why does this need to exist" and didn't confuse the rush with the destination. And also the providers giving you the AI pickaxe, but I digress.

The gold is (probably) real. The rush is definitely real. So are the holes and the piles of soil to wade through.

When is this dog not sleeping???