Nobody Wins Overnight. They Just Wait Long Enough for the World to Catch Up.

Silicon Valley has sold us all on the origin myth. Two founders, a garage, a demo that goes viral, and eighteen months later there's a glass tower and a Nasdaq ticker. It's a clean story, and clean stories travel well. It's also almost never how it actually happens.

The companies I find most interesting aren't the ones that got lucky fast. They're the ones that spent years (sometimes decades) doing something that looked, from the outside, like it wasn't working. Then the market moved, or the technology matured, or a distribution channel opened up, and the same company that looked stuck and stale suddenly looked inevitable. The "overnight success" wasn't overnight. It was the moment the world finally caught up to a capability someone had already built.

Here's three examples that paint a picture of an interesting pattern.

StackBlitz: Seven Years of Infrastructure, Two Months of Liftoff

Eric Simons and Albert Pai founded StackBlitz in 2017 to build a browser-based coding IDE on top of VS Code. The idea was to do to dev tools what Figma had done for design… The interesting part was never the IDE, it was their real technical bet, something called WebContainers: a way to run full Node.js applications entirely inside a browser tab, with no server behind it. That's a genuinely hard infrastructure problem, and it took them more than four years to get right.

For six years, StackBlitz built a loyal (but niche) following among developers and never found a way to turn that following into meaningful revenue. By late 2023, the company was in real trouble.

In early 2024 they prototyped an AI app-builder called Bolt, tried it against the models available at the time, and shelved it. The models simply weren't good enough to make the idea work. Then in June 2024 they got early access to Anthropic's Claude 3.5 Sonnet, the first model capable of actually delivering on the concept. They committed to building it that July and launched Bolt.new in October 2024.

It hit $20M in annualized revenue within two months. By March 2025 it was at $40M, and StackBlitz raised a $105M Series B shortly after.

Seven years of unglamorous, load-bearing infrastructure work sat there quietly until a model showed up that was finally good enough to stand on it.

Keurig: The Unsexy Strategy Was the Whole Strategy

John Sylvan and Peter Dragone, college roommates, founded Keurig in 1992 to solve a genuinely small problem: brewed coffee going stale and bitter as it sat on a burner. The early prototypes were rough. Bulky machines that practically needed a plumber to install, leaky pods, grounds in the cup, and by some accounts K-Cups that failed spectacularly during investor demos. A seed round in 1994 and a later $1M investment were what kept the company alive through that period.

The strategy that eventually worked was patient to the point of looking boring: give away the machines and make money on the pods, starting with office coffee service in 1998.

It took until 2004 (twelve years after founding) for a home brewer to launch, followed by years of in-store demos and giveaways before Keurig became synonymous with the single-serve category itself. Green Mountain Coffee Roasters acquired the company in 2006, growth accelerated from there, and the arc eventually ran through a $14B buyout by JAB Holding in 2016 and the 2018 merger that created Keurig Dr Pepper.

Nothing about "give away the razor, sell the blades" is a novel insight. What's notable is how long the company had to grind on distribution (one office at a time) before that ordinary idea had enough surface area to become a category.

Nvidia: A Sixteen-Year Bet Nobody Asked For

Jensen Huang, Chris Malachowsky, and Curtis Priem founded Nvidia in 1993 to build 3D graphics chips for gaming. Their first product flopped badly enough that the company had to lay off half its staff, and it was a $5M investment from Sega (persuaded rather than pitched) that kept Nvidia from folding entirely.

The company recovered with the GeForce line in 1999 and spent more than a decade being, essentially, a very good gaming graphics card company. They helped folks like me get Quake up and running properly. That was amazing on a personal level, but not world-changing.

Their second act started with a bet almost nobody was asking for: CUDA, launched in 2006, which let GPUs handle general-purpose computation instead of just rendering graphics.

For six years, that sat as a niche feature with a small, patient following. Then in 2012, researchers used CUDA and two Nvidia GPUs to train AlexNet. This proved that GPUs could train neural networks dramatically faster than CPUs could. Nvidia leaned into AI computing from there, and by the 2022–2023 generative AI boom, it had become one of the most valuable companies in the world. Almost thirty years after founding, on a bet it had placed sixteen years earlier.

The Common Thread

None of these companies got lucky in the way the mythology implies. The secret that each of them had was they simply did not die. Each one survived a long, unglamorous middle period (infrastructure nobody could see, distribution that looked plodding, a chip feature with no obvious market) and each one's "overnight" moment was really the outside world arriving at a capability that already existed.

I think about this in terms I picked up in Little League. The kid who suddenly looks like a natural at 14 usually isn't having a breakthrough season. They're the one who's been doing the boring reps (the tee work, the footwork, the stuff with no highlight reel) since they were nine, and this is just the year their body and the level of competition finally lined up with the work already done. Nobody claps for the reps. They only clap for the dingers.

This matters for how I think about advising founders and the risk-averse organizations I work with on AI adoption as well. The founders I respect most aren't the ones chasing the moment. They're the ones building the unglamorous thing that would matter if the moment arrived. And the organizations that get AI adoption right aren't the ones waiting for a perfect, obvious use case. They're doing the quiet, bite-sized work now. They're cleaning up a workflow, understanding their own data, running a small pilot nobody will notice, so that when the capability catches up to the ambition, they're already standing on something.

The lesson isn't "be patient", patience alone isn't a strategy. It's this: build the thing that's actually hard and actually valuable, even when (especially when) nobody outside is paying attention yet. The market will eventually notice what you built. Whether that's true depends entirely on whether you built something real during all the years no one was watching.