On January 24, 1848, James Marshall spotted something shining at Sutter's Mill and recalled: "My eye was caught by something shining in the bottom of the ditch. I reached my hand down and picked it up; it made my heart thump, for I was certain it was gold. Then I saw another." His colleague James Brown tested the flake with his teeth, held it aloft, and shouted "Gold, boys, gold!" — at which point every man dropped their tools and gathered around.
Within weeks, San Francisco was a ghost town. Sutter's own laborers deserted him. First they left in small groups, then all of them, from the clerks to the cooks. The San Francisco Californian ran one final editorial lamenting that the whole country echoed with the cry of gold while fields sat half-planted and houses half-built. Then the newspaper's staff walked off the job too.
The economics explain the madness. A common laborer earned $1 per day in 1847. By 1849, that same laborer could earn $7 per day, a wage premium that made the gold fields an economically rational choice for almost anyone with functioning legs. Nearly 100,000 people set out for California in 1849 alone, roughly equivalent to 1.45 million people today.
I've been thinking about this a lot recently as I try to wrap my mind around the age of AI. Not because I'm a gold rush historian, but because I run a company and I recognize the feelings many of us share. Fear. Excitement. Disbelief. Astonishment. Incredulity. That something foundational is happening.
The ground under your feet is different than it was two years ago. Heck, even two months ago. And the most natural human response to that kind of shift is to stand still and wait for clarity, to tell yourself you'll adopt AI when the technology matures, when it's less risky, when everyone else has figured it out first.
I think that's the wrong move. It was the wrong move in 1848, and it's the wrong move now.
The man who sold the shovels
Samuel Brannan didn't grab a pan when gold was discovered. He grabbed the supply chain. He bought every pick, shovel, and mining pan in the region for about 20 cents each, then sprinted through Portsmouth Square waving a bottle of gold dust and shouting about riches on the American River. He sold those same pans for $15 apiece (a 7,400% markup) and his store peaked at $5,000 per day in sales (roughly $125,000 in today's dollars). He became California's first millionaire without mining a single ounce of gold.
What I love about Brannan's story is that it wasn't luck. He saw the information asymmetry and positioned himself in the path of everyone who would chase it.
And Brannan wasn't even the most interesting example, just the most brazen. Armour, Strauss, Stanford, a half-dozen others all made fortunes off the rush without mining a single ounce. Different plays and different industries but the same approach: instead of chasing the gold, position yourself in the path of the people who are.
The wealth distribution data tells the same story from the other side. Roughly 10% of gold-seekers realized any profit after accounting for costs. Only the top 4% made a profit worth mentioning. Just 1% became rich. While the "easy gold" of 1848 allowed miners to average $20 per day, by 1853 average daily yields plummeted to less than $6 per day.
The miners went bust when the easy placer gold dried up. The merchants built dynasties.
Finding color
There's a mining term from the diggings called "finding color" — when minute particles of gold appear in your pan. Not necessarily enough to celebrate, just the first sign that something valuable might be present in the ground you're working. That's where most of us are right now with AI. I call this stage "finding color" because it captures something the tech industry's language doesn't: that early moment where you can see it working but you can't yet prove it at scale. You've found color. Now what?
That's where most of us are right now with AI. We've found color. We're seeing the first glimmers. A task that used to take four hours now takes twenty minutes. A team of three doing work that used to need twelve. A founder building a product without a CTO. Not enough to declare victory. But enough to keep working the claim.
The question is what you do next. Because in 1848, finding color was the easy part. What separated the people who built lasting wealth from the people who went broke was never about who found gold first. It was about what they built around it.
But here's the thing about staking a claim in 1849; it didn't just mean panning for gold. The smartest claims weren't on the riverbed. They were on the supply route, the town square, the trail every miner had to walk through. Brannan staked his claim on the supply chain. Stanford staked his on the crossroads. The claim worth staking was the ground where you could build something that would still be standing after the rush moved on.
The modern shovel sellers
NVIDIA is this era's Samuel Brannan. The company reached $130.5 billion in annual revenue by fiscal 2025 which was up from $27 billion two years earlier. They achieved this by selling GPUs to everyone to build AI. They sell the picks.
The cloud providers such as AWS, Azure, Google Cloud are the railroads, the physical infrastructure on which everything runs. Satya Nadella admitted the bottleneck is electricity. They have chips sitting in inventory they can't plug in. The constraint has become physical. Just like building a railroad through the Sierra Nevada was physical.
Yet the democratization parallel is the one that matters most to anyone reading this — especially if you're running a small business and wondering whether AI tools are actually for you or just for the Googles and Amazons of the world. Just as any person could stake a gold claim in 1849, anyone can now build with AI. Solo-founded startups rose from about 22% of new companies in 2015 to 38% in 2024 and we're going to see that percentage explode in 2026. Small teams are shipping products that would've required fifty engineers a year ago. A solo founder with the right AI tools can now do what a ten-person team did in 2025. The tools for building a great company aren't locked behind capital anymore; they're behind a $20/month subscription and the willingness to learn.
But here's where the analogy becomes somewhat strained. The 1849 gold fields were radically decentralized. Any person with a pan could try their luck whereas the AI infrastructure layer is dominated by a handful of hyperscalers. The tools are democratized. The picks and shovels are affordable. But the mines themselves, the massive GPU clusters and energy contracts, are concentrated in ways the gold fields never were. You can build on top of this infrastructure, but you don't own it unless you move compute on premises and build agents with open source models. I almost left this part out because it complicates the narrative. The whole point of the essay is supposed to be "anyone can build" which is mostly true. But I'd be dishonest if I didn't flag that the infrastructure layer is concentrated in ways the gold fields never were.
The skeptics always show up
When the first gold reports reached San Francisco, citizens tested the ore with spyglasses and iron ladles. A contemporary account captured the psychology: many who couldn't confirm it through their own experiments declared the whole thing a humbug. The doubt persisted because people couldn't fathom that something so valuable had been sitting there undiscovered. Accepting the news meant accepting that they'd been standing on gold the whole time.
Sound familiar?
Every frontier generates the same three reactions. There are the people who rush in blindly. There are the people who refuse to believe. And then there are the people who pay attention, figure out what the new world actually rewards, and build for it.
The man who literally owned the land where gold was discovered is a cautionary tale. John Sutter clung to his old agricultural empire while the new reality dismantled it. Workers deserted. Squatters overran his land. He wrote that the gold discovery destroyed all his great plans. The man at the epicenter of history's greatest wealth event died penniless, petitioning Congress for money that never came.
Sutter failed because he couldn't let go of the model that used to work. And the most dangerous part? His business was still profitable when the shift started. The signs that you're falling behind don't show up in this quarter's revenue. They show up in the gap between what your team is doing manually and what your competitors are automating.
What gold rushes leave behind
The strongest argument for embracing any gold rush is what it builds.
Here's the detail that gets me: California agriculture. The Central Valley is now one of the most productive farming regions on earth. It was literally seeded by miners who failed at gold but stayed for the land. They came to extract minerals and ended up building farms. The rush pulls people in for one reason, and the ones who stick around create something nobody planned.
San Francisco went from 200 people to 57,000 in fourteen years. Wells Fargo exists because someone needed to run banking for gold camps. The Transcontinental Railroad was funded by merchant profits from the rush. None of that was the point, yet all of it outlasted the gold.
The one that really surprises people is Alaska. The U.S. bought it in 1867 and the country mocked the deal for thirty years. You may remember from high school history lessons the name it was given: "Seward's Folly." Then the Klondike happened and suddenly the gateway to the northern gold fields was American territory. The purchase price paid for itself many times over, but only because a gold rush nobody predicted vindicated a bet nobody understood at the time.
I think about that when people ask whether AI is a bubble. Maybe it is. I don't think that's actually the right question though... The dot-com bubble was a bubble too, but it left behind the fiber-optic cables that power the modern internet. Railroad speculation in the 1800s was reckless by any measure, and those tracks carried freight for a century after the speculators went broke. The frenzies often look stupid in real time. But the infrastructure they fund outlasts the speculation every time. So if AI is a bubble, the question isn't whether it'll pop, it's what gets built before it does.
Belinda Mulrooney, the Klondike's richest woman, understood this intuitively. She arrived in Dawson with goods she'd hauled over the Chilkoot Pass. Things like silk underwear, cotton cloth, hot water bottles. Once she arrived, she tossed her last coin into the river. Her goods netted a 600% profit. She built Dawson's finest hotel and brought the town its first telephone. She understood something the miners missed: the opportunity wasn't in the gold. It was in the people who came for the gold.
Stake your claim
Of the roughly 100,000 people who set out for the Klondike, only about 4,000 found gold and only a few hundred struck it rich. The individual outcomes were brutal, random, and often unfair. But the collective outcome was civilization: cities, railroads, universities, and industries that outlasted every individual prospector's story.
The fear around AI is real and I won't dismiss it. I've felt it myself. I think we're going to see a lot of creative destruction these next several years as entire business models are rethought and new nimble companies challenge massive incumbents.
Here's what the gold rushes actually teach us: most people lost. The miners went broke when the streams dried up, the skeptics showed up too late, and Sutter lost everything because he couldn't stop being a farmer long enough to notice the world had changed.
The people who built lasting wealth weren't smarter or luckier. They just asked a different question. Instead of "where's the gold?" they asked "what do the people chasing gold actually need?" Brannan figured it out first with supply chains. Mulrooney figured it out with hospitality. The others landed somewhere in between. But none of them mined gold.
The barrier to building something right now has never been lower. An API key, a laptop, and an idea worth testing. That's it.
But "build something" isn't enough. The gold rushes are specific about what works, and I'd argue they're the best AI business strategy framework nobody's using. Solve a real problem that the rush itself creates. Build the infrastructure that everyone else depends on. Serve the people who show up, not the resource they're chasing. And if you're already running a company, ask yourself the hardest question Sutter never asked: is the model you're protecting still the model worth building?
I think about this every day at my own company. We manage vacation rental properties, and honestly, this industry has operated like miners for as long as I've been in it. Chasing bookings. Clinging to manual processes. Treating every property as something to extract revenue from rather than something to build around.
We're trying to do it differently with AI-first operations, systems that actually scale and actively complete workflows, not just 70% of a workflow. Whether it's working fully as envisioned is a fair question! Some of it is, while some of it we're still figuring out. But the bet we've made is that the opportunity isn't going anywhere!
We believe the new limits in the age of AI are ambition, creativity, and the laws of physics.
We've staked our claim not on the gold, but on serving the prospectors. Where will you stake yours?