The AI Wildcatters
How today's coding assistant gold rush mirrors the Pennsylvania oil boom—and why the infrastructure players could win
In December 2024, Cursor raised $100 million at a $2.6 billion valuation. By March 2025, the AI-powered coding editor was in talks for a $10 billion valuation. Three months later, they closed at $9.9 billion.
That's a 25x increase in 18 months.
Cursor isn't alone in this frenzy. Cognition Labs, makers of the Devin coding agent, doubled from $2 billion to $4 billion in less than a year. Windsurf was reportedly in talks with OpenAI for a $3 billion acquisition before everything fell apart. The pattern is everywhere: AI coding assistants are attracting massive valuations based on the promise that autonomous agents will transform software development.
Investors are betting billions that we're witnessing the birth of a new industry. The excitement is palpable, the growth metrics are staggering, and everyone wants in. It feels like anyone with a decent AI coding tool can strike it rich.
But this gold rush energy feels familiar. We've seen this story before, in a different time and place, with a different resource that promised to change everything.
The Original Wildcatters
In 1859, Edwin Drake struck oil in Titusville, Pennsylvania, launching America's first oil boom. Word spread quickly: there was black gold beneath the ground, and anyone with a drill could get rich.
The wildcatters came in droves. These were independent operators who leased small plots of land and drilled wells, hoping to strike it rich with a single lucky hit. Oil Creek Township had no population in 1860, over 5,000 in 1870. Pithole City exploded from non-existence in 1860 to over 10,000 residents in 1865, complete with hotels, theaters, and a post office that handled more mail than Philadelphia.
The parallels to today's AI boom are striking. Just as wildcatters rushed to drill wells wherever they could lease land, AI entrepreneurs are rushing to build coding assistants wherever they can access foundation models. The excitement is the same: a new technology that promises to transform an entire industry, with seemingly unlimited upside for anyone who can get there first.
But most wildcatters went bust as quickly as they boomed. Pithole City fell to just 237 residents by 1880. Oil Creek Township collapsed from 5,000 people to 526. The boom towns that had sprung up overnight became ghost towns just as fast.
The problem wasn't that oil wasn't valuable—it was incredibly valuable. The problem was that drilling wells wasn't where the real money was made.
The Infrastructure Play
While wildcatters were focused on drilling, John D. Rockefeller was building something different. Starting in 1863, he built his first oil refinery near Cleveland. By 1870, he had founded Standard Oil, but his strategy wasn't to drill more wells—it was to control what happened to the oil after it came out of the ground.
Rockefeller focused on refining, transportation, and distribution. He built pipelines, acquired railroads, and standardized the entire process from crude oil to kerosene. While wildcatters competed to find oil, Rockefeller built the infrastructure that made oil useful. He didn't need to strike oil himself—he just needed to control the systems that processed everyone else's oil.
The strategy worked. By the 1880s, Standard Oil controlled roughly 90% of U.S. oil refining and pipelines. The wildcatters who had struck it rich in the 1860s were now dependent on Rockefeller's infrastructure to get their oil to market. He had become the bottleneck that every oil producer had to go through.
But Rockefeller didn't just build infrastructure—he used it as a weapon. In 1872, during what became known as "The Cleveland Massacre," Standard Oil acquired 22 of its 26 Cleveland competitors in just six weeks. Rockefeller's method was simple: he controlled the railroads through rebate agreements, which meant he could ship oil cheaper than anyone else. When competitors refused to sell, he would undercut their prices until they faced a choice—sell to Standard Oil or go bankrupt.
Those who refused to sell fell from the heights of Cleveland society into bankruptcy. Control of the infrastructure meant control of who survived.
History Rhymes
In June 2025, Windsurf was riding high. The AI-powered coding environment had built a loyal user base and was reportedly in talks with OpenAI for a $3 billion acquisition. Then Anthropic sent a letter.
The AI company cut off Windsurf's access to Claude 3.5 Sonnet and related models with less than a week's notice. Anthropic's reasoning was straightforward: they couldn't justify supplying their largest competitor, OpenAI, with access to their models through a middle layer. As Anthropic co-founder Jared Kaplan put it, "It would be odd for us to sell Claude to OpenAI."
The effect was immediate. Windsurf had to reroute traffic through third-party providers, restrict access for free users, and scramble to find alternatives. More importantly, the OpenAI acquisition talks collapsed.
Like a wildcatter who had struck oil but didn't own the railroad, Windsurf discovered they had built their business on someone else's infrastructure. The company that had been positioning itself as a unified $3 billion acquisition suddenly found itself fragmented: Google hired the founders and research team for $2.4 billion, while the remaining team and product were acquired by Cognition Labs.
Sure, $2.4 billion isn't exactly poverty—it's all funny money at these valuations anyway. But the point isn't the dollar amount. The point is that Windsurf's future was decided by someone else's letter. They had built a valuable product, attracted users, and generated revenue, but when it mattered most, they didn't control their own destiny.
Today's Standard Oil
While AI coding assistants compete for users and valuations, the cloud providers are building the infrastructure that all of them depend on. Amazon Web Services holds 33% of the cloud market, Microsoft Azure has 20%, and Google Cloud has 11%. Together, they control the compute, storage, and APIs that power the AI boom.
Follow the money. Amazon is spending $105 billion in 2025 on AI infrastructure alone—more than most countries' entire GDP. Cloud infrastructure spending hit $90 billion in Q4 2024, up 22% year-over-year. These companies aren't just providing servers—they're building the railroads and refineries of the AI age.
Just like Rockefeller, today's infrastructure players are pursuing vertical integration. Amazon controls the infrastructure through AWS, invested in Anthropic for the models, and launched Kiro to compete in applications. Google provides the cloud infrastructure, hired Windsurf's founders for the talent, and builds its own AI tools. Microsoft runs the infrastructure through Azure, partners with OpenAI for models, and competes directly with GitHub Copilot (Another fact in the deal falling apart was that Microsoft wanted access to the Windsurf IP). They don't need to build the best AI coding assistant—they just need to control every layer of the stack.
And now they're coming for the wildcatters directly. Cursor, with its $9.9 billion valuation, is watching the same infrastructure freight train that flattened Windsurf bearing down on them. Anthropic launched Claude Code, an agentic coding tool with flat-rate pricing and an API that lets others build competing applications. Google rolled out Gemini CLI, a free open-source coding agent that brings AI directly to the terminal.
It's too early to tell which AI applications will survive the current gold rush. But it's not too early to bet on the infrastructure players. History doesn't repeat, but it often rhymes—and right now, it's rhyming with a familiar tune about railroads, refineries, and the companies that control the pipes.